2018-04-20 11:40:15 -04:00
|
|
|
#define Simd
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
using ARMeilleure.State;
|
|
|
|
|
2018-04-20 11:40:15 -04:00
|
|
|
using NUnit.Framework;
|
|
|
|
|
2019-04-12 12:14:16 -04:00
|
|
|
using System;
|
2018-09-17 00:54:05 -04:00
|
|
|
using System.Collections.Generic;
|
2018-05-11 19:10:27 -04:00
|
|
|
|
2018-04-20 11:40:15 -04:00
|
|
|
namespace Ryujinx.Tests.Cpu
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
[Category("Simd")]
|
2018-04-20 11:40:15 -04:00
|
|
|
public sealed class CpuTestSimd : CpuTest
|
|
|
|
{
|
|
|
|
#if Simd
|
|
|
|
|
2019-06-12 08:03:31 -04:00
|
|
|
#region "Helper methods"
|
|
|
|
private static byte GenLeadingSignsMinus8(int cnt) // 0 <= cnt <= 7
|
|
|
|
{
|
|
|
|
return (byte)(~(uint)GenLeadingZeros8(cnt + 1));
|
|
|
|
}
|
|
|
|
|
|
|
|
private static ushort GenLeadingSignsMinus16(int cnt) // 0 <= cnt <= 15
|
|
|
|
{
|
|
|
|
return (ushort)(~(uint)GenLeadingZeros16(cnt + 1));
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint GenLeadingSignsMinus32(int cnt) // 0 <= cnt <= 31
|
|
|
|
{
|
|
|
|
return ~GenLeadingZeros32(cnt + 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
private static byte GenLeadingSignsPlus8(int cnt) // 0 <= cnt <= 7
|
|
|
|
{
|
|
|
|
return GenLeadingZeros8(cnt + 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
private static ushort GenLeadingSignsPlus16(int cnt) // 0 <= cnt <= 15
|
|
|
|
{
|
|
|
|
return GenLeadingZeros16(cnt + 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint GenLeadingSignsPlus32(int cnt) // 0 <= cnt <= 31
|
|
|
|
{
|
|
|
|
return GenLeadingZeros32(cnt + 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
private static byte GenLeadingZeros8(int cnt) // 0 <= cnt <= 8
|
|
|
|
{
|
|
|
|
if (cnt == 8) return (byte)0;
|
|
|
|
if (cnt == 7) return (byte)1;
|
|
|
|
|
|
|
|
byte rnd = TestContext.CurrentContext.Random.NextByte();
|
|
|
|
sbyte mask = sbyte.MinValue;
|
|
|
|
|
|
|
|
return (byte)(((uint)rnd >> (cnt + 1)) | ((uint)((byte)mask) >> cnt));
|
|
|
|
}
|
|
|
|
|
|
|
|
private static ushort GenLeadingZeros16(int cnt) // 0 <= cnt <= 16
|
|
|
|
{
|
|
|
|
if (cnt == 16) return (ushort)0;
|
|
|
|
if (cnt == 15) return (ushort)1;
|
|
|
|
|
|
|
|
ushort rnd = TestContext.CurrentContext.Random.NextUShort();
|
|
|
|
short mask = short.MinValue;
|
|
|
|
|
|
|
|
return (ushort)(((uint)rnd >> (cnt + 1)) | ((uint)((ushort)mask) >> cnt));
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint GenLeadingZeros32(int cnt) // 0 <= cnt <= 32
|
|
|
|
{
|
|
|
|
if (cnt == 32) return 0u;
|
|
|
|
if (cnt == 31) return 1u;
|
|
|
|
|
|
|
|
uint rnd = TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
int mask = int.MinValue;
|
|
|
|
|
|
|
|
return (rnd >> (cnt + 1)) | ((uint)mask >> cnt);
|
|
|
|
}
|
|
|
|
#endregion
|
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
#region "ValueSource (Types)"
|
2018-08-04 15:58:54 -04:00
|
|
|
private static ulong[] _1B1H1S1D_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x000000000000007Ful,
|
|
|
|
0x0000000000000080ul, 0x00000000000000FFul,
|
|
|
|
0x0000000000007FFFul, 0x0000000000008000ul,
|
|
|
|
0x000000000000FFFFul, 0x000000007FFFFFFFul,
|
|
|
|
0x0000000080000000ul, 0x00000000FFFFFFFFul,
|
|
|
|
0x7FFFFFFFFFFFFFFFul, 0x8000000000000000ul,
|
|
|
|
0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-04-21 15:15:04 -04:00
|
|
|
private static ulong[] _1D_()
|
2018-04-20 11:40:15 -04:00
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7FFFFFFFFFFFFFFFul,
|
|
|
|
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-04-29 19:39:58 -04:00
|
|
|
private static ulong[] _1H1S1D_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x0000000000007FFFul,
|
|
|
|
0x0000000000008000ul, 0x000000000000FFFFul,
|
|
|
|
0x000000007FFFFFFFul, 0x0000000080000000ul,
|
|
|
|
0x00000000FFFFFFFFul, 0x7FFFFFFFFFFFFFFFul,
|
|
|
|
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-11-17 21:41:16 -05:00
|
|
|
private static ulong[] _1S_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x000000007FFFFFFFul,
|
|
|
|
0x0000000080000000ul, 0x00000000FFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
|
|
|
private static ulong[] _2S_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7FFFFFFF7FFFFFFFul,
|
|
|
|
0x8000000080000000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2019-07-08 10:55:37 -04:00
|
|
|
private static ulong[] _4H_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7FFF7FFF7FFF7FFFul,
|
|
|
|
0x8000800080008000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-04-29 19:39:58 -04:00
|
|
|
private static ulong[] _4H2S1D_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7FFF7FFF7FFF7FFFul,
|
|
|
|
0x8000800080008000ul, 0x7FFFFFFF7FFFFFFFul,
|
|
|
|
0x8000000080000000ul, 0x7FFFFFFFFFFFFFFFul,
|
|
|
|
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-06-18 13:55:26 -04:00
|
|
|
private static ulong[] _8B_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7F7F7F7F7F7F7F7Ful,
|
|
|
|
0x8080808080808080ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-04-21 15:15:04 -04:00
|
|
|
private static ulong[] _8B4H_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7F7F7F7F7F7F7F7Ful,
|
|
|
|
0x8080808080808080ul, 0x7FFF7FFF7FFF7FFFul,
|
|
|
|
0x8000800080008000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-04-20 11:40:15 -04:00
|
|
|
private static ulong[] _8B4H2S_()
|
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7F7F7F7F7F7F7F7Ful,
|
|
|
|
0x8080808080808080ul, 0x7FFF7FFF7FFF7FFFul,
|
|
|
|
0x8000800080008000ul, 0x7FFFFFFF7FFFFFFFul,
|
|
|
|
0x8000000080000000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
|
|
|
|
2018-04-21 15:15:04 -04:00
|
|
|
private static ulong[] _8B4H2S1D_()
|
2018-04-20 11:40:15 -04:00
|
|
|
{
|
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7F7F7F7F7F7F7F7Ful,
|
|
|
|
0x8080808080808080ul, 0x7FFF7FFF7FFF7FFFul,
|
|
|
|
0x8000800080008000ul, 0x7FFFFFFF7FFFFFFFul,
|
|
|
|
0x8000000080000000ul, 0x7FFFFFFFFFFFFFFFul,
|
|
|
|
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
|
|
|
|
}
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2019-07-08 10:55:37 -04:00
|
|
|
private static uint[] _W_()
|
2019-06-12 08:03:31 -04:00
|
|
|
{
|
2019-07-08 10:55:37 -04:00
|
|
|
return new uint[] { 0x00000000u, 0x7FFFFFFFu,
|
|
|
|
0x80000000u, 0xFFFFFFFFu };
|
2019-06-12 08:03:31 -04:00
|
|
|
}
|
|
|
|
|
2019-07-08 10:55:37 -04:00
|
|
|
private static ulong[] _X_()
|
2019-06-12 08:03:31 -04:00
|
|
|
{
|
2019-07-08 10:55:37 -04:00
|
|
|
return new ulong[] { 0x0000000000000000ul, 0x7FFFFFFFFFFFFFFFul,
|
|
|
|
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
|
2019-06-12 08:03:31 -04:00
|
|
|
}
|
|
|
|
|
2019-05-30 18:51:39 -04:00
|
|
|
private static IEnumerable<ulong> _1H_F_()
|
|
|
|
{
|
|
|
|
yield return 0x000000000000FBFFul; // -Max Normal
|
|
|
|
yield return 0x0000000000008400ul; // -Min Normal
|
|
|
|
yield return 0x00000000000083FFul; // -Max Subnormal
|
|
|
|
yield return 0x0000000000008001ul; // -Min Subnormal
|
|
|
|
yield return 0x0000000000007BFFul; // +Max Normal
|
|
|
|
yield return 0x0000000000000400ul; // +Min Normal
|
|
|
|
yield return 0x00000000000003FFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000000000001ul; // +Min Subnormal
|
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x0000000000008000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0x000000000000FC00ul; // -Infinity
|
|
|
|
yield return 0x0000000000007C00ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
|
|
|
yield return 0x000000000000FE00ul; // -QNaN (all zeros payload)
|
|
|
|
yield return 0x000000000000FDFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x0000000000007E00ul; // +QNaN (all zeros payload) (DefaultNaN)
|
|
|
|
yield return 0x0000000000007DFFul; // +SNaN (all ones payload)
|
|
|
|
}
|
|
|
|
|
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
|
|
|
{
|
|
|
|
ulong grbg = TestContext.CurrentContext.Random.NextUShort();
|
|
|
|
ulong rnd1 = GenNormalH();
|
|
|
|
ulong rnd2 = GenSubnormalH();
|
|
|
|
|
|
|
|
yield return (grbg << 48) | (grbg << 32) | (grbg << 16) | rnd1;
|
|
|
|
yield return (grbg << 48) | (grbg << 32) | (grbg << 16) | rnd2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-10-23 10:12:45 -04:00
|
|
|
private static IEnumerable<ulong> _4H_F_()
|
|
|
|
{
|
|
|
|
yield return 0xFBFFFBFFFBFFFBFFul; // -Max Normal
|
|
|
|
yield return 0x8400840084008400ul; // -Min Normal
|
|
|
|
yield return 0x83FF83FF83FF83FFul; // -Max Subnormal
|
|
|
|
yield return 0x8001800180018001ul; // -Min Subnormal
|
|
|
|
yield return 0x7BFF7BFF7BFF7BFFul; // +Max Normal
|
|
|
|
yield return 0x0400040004000400ul; // +Min Normal
|
|
|
|
yield return 0x03FF03FF03FF03FFul; // +Max Subnormal
|
|
|
|
yield return 0x0001000100010001ul; // +Min Subnormal
|
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x8000800080008000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0xFC00FC00FC00FC00ul; // -Infinity
|
|
|
|
yield return 0x7C007C007C007C00ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
|
|
|
yield return 0xFE00FE00FE00FE00ul; // -QNaN (all zeros payload)
|
|
|
|
yield return 0xFDFFFDFFFDFFFDFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x7E007E007E007E00ul; // +QNaN (all zeros payload) (DefaultNaN)
|
|
|
|
yield return 0x7DFF7DFF7DFF7DFFul; // +SNaN (all ones payload)
|
|
|
|
}
|
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
2018-10-23 10:12:45 -04:00
|
|
|
{
|
2019-05-30 18:51:39 -04:00
|
|
|
ulong rnd1 = GenNormalH();
|
|
|
|
ulong rnd2 = GenSubnormalH();
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
yield return (rnd1 << 48) | (rnd1 << 32) | (rnd1 << 16) | rnd1;
|
|
|
|
yield return (rnd2 << 48) | (rnd2 << 32) | (rnd2 << 16) | rnd2;
|
2018-10-23 10:12:45 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
private static IEnumerable<ulong> _1S_F_()
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x00000000FF7FFFFFul; // -Max Normal (float.MinValue)
|
2018-09-17 00:54:05 -04:00
|
|
|
yield return 0x0000000080800000ul; // -Min Normal
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x00000000807FFFFFul; // -Max Subnormal
|
|
|
|
yield return 0x0000000080000001ul; // -Min Subnormal (-float.Epsilon)
|
|
|
|
yield return 0x000000007F7FFFFFul; // +Max Normal (float.MaxValue)
|
2018-09-17 00:54:05 -04:00
|
|
|
yield return 0x0000000000800000ul; // +Min Normal
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x00000000007FFFFFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000000000001ul; // +Min Subnormal (float.Epsilon)
|
2018-09-17 00:54:05 -04:00
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x0000000080000000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0x00000000FF800000ul; // -Infinity
|
|
|
|
yield return 0x000000007F800000ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x00000000FFC00000ul; // -QNaN (all zeros payload) (float.NaN)
|
|
|
|
yield return 0x00000000FFBFFFFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x000000007FC00000ul; // +QNaN (all zeros payload) (-float.NaN) (DefaultNaN)
|
|
|
|
yield return 0x000000007FBFFFFFul; // +SNaN (all ones payload)
|
2018-09-17 00:54:05 -04:00
|
|
|
}
|
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
2018-09-17 00:54:05 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong grbg = TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
ulong rnd1 = GenNormalS();
|
|
|
|
ulong rnd2 = GenSubnormalS();
|
2018-09-17 00:54:05 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
yield return (grbg << 32) | rnd1;
|
|
|
|
yield return (grbg << 32) | rnd2;
|
2018-09-17 00:54:05 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-04-12 12:14:16 -04:00
|
|
|
private static IEnumerable<ulong> _1S_F_W_()
|
2019-04-03 08:21:22 -04:00
|
|
|
{
|
|
|
|
// int
|
2019-04-12 12:14:16 -04:00
|
|
|
yield return 0x00000000CF000001ul; // -2.1474839E9f (-2147483904)
|
|
|
|
yield return 0x00000000CF000000ul; // -2.14748365E9f (-2147483648)
|
|
|
|
yield return 0x00000000CEFFFFFFul; // -2.14748352E9f (-2147483520)
|
|
|
|
yield return 0x000000004F000001ul; // 2.1474839E9f (2147483904)
|
|
|
|
yield return 0x000000004F000000ul; // 2.14748365E9f (2147483648)
|
|
|
|
yield return 0x000000004EFFFFFFul; // 2.14748352E9f (2147483520)
|
|
|
|
|
|
|
|
// uint
|
|
|
|
yield return 0x000000004F800001ul; // 4.2949678E9f (4294967808)
|
|
|
|
yield return 0x000000004F800000ul; // 4.2949673E9f (4294967296)
|
|
|
|
yield return 0x000000004F7FFFFFul; // 4.29496704E9f (4294967040)
|
2019-04-03 08:21:22 -04:00
|
|
|
|
|
|
|
yield return 0x00000000FF7FFFFFul; // -Max Normal (float.MinValue)
|
|
|
|
yield return 0x0000000080800000ul; // -Min Normal
|
|
|
|
yield return 0x00000000807FFFFFul; // -Max Subnormal
|
|
|
|
yield return 0x0000000080000001ul; // -Min Subnormal (-float.Epsilon)
|
|
|
|
yield return 0x000000007F7FFFFFul; // +Max Normal (float.MaxValue)
|
|
|
|
yield return 0x0000000000800000ul; // +Min Normal
|
|
|
|
yield return 0x00000000007FFFFFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000000000001ul; // +Min Subnormal (float.Epsilon)
|
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x0000000080000000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0x00000000FF800000ul; // -Infinity
|
|
|
|
yield return 0x000000007F800000ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
|
|
|
yield return 0x00000000FFC00000ul; // -QNaN (all zeros payload) (float.NaN)
|
|
|
|
yield return 0x00000000FFBFFFFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x000000007FC00000ul; // +QNaN (all zeros payload) (-float.NaN) (DefaultNaN)
|
|
|
|
yield return 0x000000007FBFFFFFul; // +SNaN (all ones payload)
|
|
|
|
}
|
|
|
|
|
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
|
|
|
{
|
|
|
|
ulong grbg = TestContext.CurrentContext.Random.NextUInt();
|
2019-04-12 12:14:16 -04:00
|
|
|
|
|
|
|
ulong rnd1 = (uint)BitConverter.SingleToInt32Bits(
|
|
|
|
(float)((int)TestContext.CurrentContext.Random.NextUInt()));
|
|
|
|
ulong rnd2 = (uint)BitConverter.SingleToInt32Bits(
|
|
|
|
(float)((uint)TestContext.CurrentContext.Random.NextUInt()));
|
|
|
|
|
|
|
|
ulong rnd3 = GenNormalS();
|
|
|
|
ulong rnd4 = GenSubnormalS();
|
2019-04-03 08:21:22 -04:00
|
|
|
|
|
|
|
yield return (grbg << 32) | rnd1;
|
|
|
|
yield return (grbg << 32) | rnd2;
|
2019-04-12 12:14:16 -04:00
|
|
|
|
|
|
|
yield return (grbg << 32) | rnd3;
|
|
|
|
yield return (grbg << 32) | rnd4;
|
2019-04-03 08:21:22 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
private static IEnumerable<ulong> _2S_F_()
|
|
|
|
{
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0xFF7FFFFFFF7FFFFFul; // -Max Normal (float.MinValue)
|
2018-09-17 00:54:05 -04:00
|
|
|
yield return 0x8080000080800000ul; // -Min Normal
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x807FFFFF807FFFFFul; // -Max Subnormal
|
|
|
|
yield return 0x8000000180000001ul; // -Min Subnormal (-float.Epsilon)
|
|
|
|
yield return 0x7F7FFFFF7F7FFFFFul; // +Max Normal (float.MaxValue)
|
2018-09-17 00:54:05 -04:00
|
|
|
yield return 0x0080000000800000ul; // +Min Normal
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x007FFFFF007FFFFFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000100000001ul; // +Min Subnormal (float.Epsilon)
|
2018-09-17 00:54:05 -04:00
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x8000000080000000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0xFF800000FF800000ul; // -Infinity
|
|
|
|
yield return 0x7F8000007F800000ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0xFFC00000FFC00000ul; // -QNaN (all zeros payload) (float.NaN)
|
|
|
|
yield return 0xFFBFFFFFFFBFFFFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x7FC000007FC00000ul; // +QNaN (all zeros payload) (-float.NaN) (DefaultNaN)
|
|
|
|
yield return 0x7FBFFFFF7FBFFFFFul; // +SNaN (all ones payload)
|
2018-09-17 00:54:05 -04:00
|
|
|
}
|
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
2018-09-17 00:54:05 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong rnd1 = GenNormalS();
|
|
|
|
ulong rnd2 = GenSubnormalS();
|
2018-09-17 00:54:05 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
yield return (rnd1 << 32) | rnd1;
|
|
|
|
yield return (rnd2 << 32) | rnd2;
|
2018-09-17 00:54:05 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-04-12 12:14:16 -04:00
|
|
|
private static IEnumerable<ulong> _2S_F_W_()
|
2019-04-03 08:21:22 -04:00
|
|
|
{
|
|
|
|
// int
|
2019-04-12 12:14:16 -04:00
|
|
|
yield return 0xCF000001CF000001ul; // -2.1474839E9f (-2147483904)
|
|
|
|
yield return 0xCF000000CF000000ul; // -2.14748365E9f (-2147483648)
|
|
|
|
yield return 0xCEFFFFFFCEFFFFFFul; // -2.14748352E9f (-2147483520)
|
|
|
|
yield return 0x4F0000014F000001ul; // 2.1474839E9f (2147483904)
|
|
|
|
yield return 0x4F0000004F000000ul; // 2.14748365E9f (2147483648)
|
|
|
|
yield return 0x4EFFFFFF4EFFFFFFul; // 2.14748352E9f (2147483520)
|
|
|
|
|
|
|
|
// uint
|
|
|
|
yield return 0x4F8000014F800001ul; // 4.2949678E9f (4294967808)
|
|
|
|
yield return 0x4F8000004F800000ul; // 4.2949673E9f (4294967296)
|
|
|
|
yield return 0x4F7FFFFF4F7FFFFFul; // 4.29496704E9f (4294967040)
|
2019-04-03 08:21:22 -04:00
|
|
|
|
|
|
|
yield return 0xFF7FFFFFFF7FFFFFul; // -Max Normal (float.MinValue)
|
|
|
|
yield return 0x8080000080800000ul; // -Min Normal
|
|
|
|
yield return 0x807FFFFF807FFFFFul; // -Max Subnormal
|
|
|
|
yield return 0x8000000180000001ul; // -Min Subnormal (-float.Epsilon)
|
|
|
|
yield return 0x7F7FFFFF7F7FFFFFul; // +Max Normal (float.MaxValue)
|
|
|
|
yield return 0x0080000000800000ul; // +Min Normal
|
|
|
|
yield return 0x007FFFFF007FFFFFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000100000001ul; // +Min Subnormal (float.Epsilon)
|
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x8000000080000000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0xFF800000FF800000ul; // -Infinity
|
|
|
|
yield return 0x7F8000007F800000ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
|
|
|
yield return 0xFFC00000FFC00000ul; // -QNaN (all zeros payload) (float.NaN)
|
|
|
|
yield return 0xFFBFFFFFFFBFFFFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x7FC000007FC00000ul; // +QNaN (all zeros payload) (-float.NaN) (DefaultNaN)
|
|
|
|
yield return 0x7FBFFFFF7FBFFFFFul; // +SNaN (all ones payload)
|
|
|
|
}
|
|
|
|
|
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
|
|
|
{
|
2019-04-12 12:14:16 -04:00
|
|
|
ulong rnd1 = (uint)BitConverter.SingleToInt32Bits(
|
|
|
|
(float)((int)TestContext.CurrentContext.Random.NextUInt()));
|
|
|
|
ulong rnd2 = (uint)BitConverter.SingleToInt32Bits(
|
|
|
|
(float)((uint)TestContext.CurrentContext.Random.NextUInt()));
|
|
|
|
|
|
|
|
ulong rnd3 = GenNormalS();
|
|
|
|
ulong rnd4 = GenSubnormalS();
|
2019-04-03 08:21:22 -04:00
|
|
|
|
|
|
|
yield return (rnd1 << 32) | rnd1;
|
|
|
|
yield return (rnd2 << 32) | rnd2;
|
2019-04-12 12:14:16 -04:00
|
|
|
|
|
|
|
yield return (rnd3 << 32) | rnd3;
|
|
|
|
yield return (rnd4 << 32) | rnd4;
|
2019-04-03 08:21:22 -04:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
private static IEnumerable<ulong> _1D_F_()
|
|
|
|
{
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0xFFEFFFFFFFFFFFFFul; // -Max Normal (double.MinValue)
|
2018-09-17 00:54:05 -04:00
|
|
|
yield return 0x8010000000000000ul; // -Min Normal
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x800FFFFFFFFFFFFFul; // -Max Subnormal
|
|
|
|
yield return 0x8000000000000001ul; // -Min Subnormal (-double.Epsilon)
|
|
|
|
yield return 0x7FEFFFFFFFFFFFFFul; // +Max Normal (double.MaxValue)
|
2018-09-17 00:54:05 -04:00
|
|
|
yield return 0x0010000000000000ul; // +Min Normal
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0x000FFFFFFFFFFFFFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000000000001ul; // +Min Subnormal (double.Epsilon)
|
2018-09-17 00:54:05 -04:00
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x8000000000000000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0xFFF0000000000000ul; // -Infinity
|
|
|
|
yield return 0x7FF0000000000000ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
2018-10-05 21:45:59 -04:00
|
|
|
yield return 0xFFF8000000000000ul; // -QNaN (all zeros payload) (double.NaN)
|
|
|
|
yield return 0xFFF7FFFFFFFFFFFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x7FF8000000000000ul; // +QNaN (all zeros payload) (-double.NaN) (DefaultNaN)
|
|
|
|
yield return 0x7FF7FFFFFFFFFFFFul; // +SNaN (all ones payload)
|
2018-09-17 00:54:05 -04:00
|
|
|
}
|
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong rnd1 = GenNormalD();
|
|
|
|
ulong rnd2 = GenSubnormalD();
|
2018-09-17 00:54:05 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
yield return rnd1;
|
|
|
|
yield return rnd2;
|
2018-09-17 00:54:05 -04:00
|
|
|
}
|
|
|
|
}
|
2019-04-03 08:21:22 -04:00
|
|
|
|
2019-04-12 12:14:16 -04:00
|
|
|
private static IEnumerable<ulong> _1D_F_X_()
|
2019-04-03 08:21:22 -04:00
|
|
|
{
|
|
|
|
// long
|
|
|
|
yield return 0xC3E0000000000001ul; // -9.2233720368547780E18d (-9223372036854778000)
|
|
|
|
yield return 0xC3E0000000000000ul; // -9.2233720368547760E18d (-9223372036854776000)
|
|
|
|
yield return 0xC3DFFFFFFFFFFFFFul; // -9.2233720368547750E18d (-9223372036854775000)
|
|
|
|
yield return 0x43E0000000000001ul; // 9.2233720368547780E18d (9223372036854778000)
|
|
|
|
yield return 0x43E0000000000000ul; // 9.2233720368547760E18d (9223372036854776000)
|
|
|
|
yield return 0x43DFFFFFFFFFFFFFul; // 9.2233720368547750E18d (9223372036854775000)
|
|
|
|
|
2019-04-12 12:14:16 -04:00
|
|
|
// ulong
|
|
|
|
yield return 0x43F0000000000001ul; // 1.8446744073709556e19d (18446744073709556000)
|
|
|
|
yield return 0x43F0000000000000ul; // 1.8446744073709552E19d (18446744073709552000)
|
|
|
|
yield return 0x43EFFFFFFFFFFFFFul; // 1.8446744073709550e19d (18446744073709550000)
|
|
|
|
|
2019-04-03 08:21:22 -04:00
|
|
|
yield return 0xFFEFFFFFFFFFFFFFul; // -Max Normal (double.MinValue)
|
|
|
|
yield return 0x8010000000000000ul; // -Min Normal
|
|
|
|
yield return 0x800FFFFFFFFFFFFFul; // -Max Subnormal
|
|
|
|
yield return 0x8000000000000001ul; // -Min Subnormal (-double.Epsilon)
|
|
|
|
yield return 0x7FEFFFFFFFFFFFFFul; // +Max Normal (double.MaxValue)
|
|
|
|
yield return 0x0010000000000000ul; // +Min Normal
|
|
|
|
yield return 0x000FFFFFFFFFFFFFul; // +Max Subnormal
|
|
|
|
yield return 0x0000000000000001ul; // +Min Subnormal (double.Epsilon)
|
|
|
|
|
|
|
|
if (!NoZeros)
|
|
|
|
{
|
|
|
|
yield return 0x8000000000000000ul; // -Zero
|
|
|
|
yield return 0x0000000000000000ul; // +Zero
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoInfs)
|
|
|
|
{
|
|
|
|
yield return 0xFFF0000000000000ul; // -Infinity
|
|
|
|
yield return 0x7FF0000000000000ul; // +Infinity
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!NoNaNs)
|
|
|
|
{
|
|
|
|
yield return 0xFFF8000000000000ul; // -QNaN (all zeros payload) (double.NaN)
|
|
|
|
yield return 0xFFF7FFFFFFFFFFFFul; // -SNaN (all ones payload)
|
|
|
|
yield return 0x7FF8000000000000ul; // +QNaN (all zeros payload) (-double.NaN) (DefaultNaN)
|
|
|
|
yield return 0x7FF7FFFFFFFFFFFFul; // +SNaN (all ones payload)
|
|
|
|
}
|
|
|
|
|
|
|
|
for (int cnt = 1; cnt <= RndCnt; cnt++)
|
|
|
|
{
|
2019-04-12 12:14:16 -04:00
|
|
|
ulong rnd1 = (ulong)BitConverter.DoubleToInt64Bits(
|
|
|
|
(double)((long)TestContext.CurrentContext.Random.NextULong()));
|
|
|
|
ulong rnd2 = (ulong)BitConverter.DoubleToInt64Bits(
|
|
|
|
(double)((ulong)TestContext.CurrentContext.Random.NextULong()));
|
|
|
|
|
|
|
|
ulong rnd3 = GenNormalD();
|
|
|
|
ulong rnd4 = GenSubnormalD();
|
2019-04-03 08:21:22 -04:00
|
|
|
|
|
|
|
yield return rnd1;
|
|
|
|
yield return rnd2;
|
2019-04-12 12:14:16 -04:00
|
|
|
|
|
|
|
yield return rnd3;
|
|
|
|
yield return rnd4;
|
2019-04-03 08:21:22 -04:00
|
|
|
}
|
|
|
|
}
|
2019-06-29 19:02:48 -04:00
|
|
|
|
2019-07-08 10:55:37 -04:00
|
|
|
private static IEnumerable<ulong> _GenLeadingSigns8B_()
|
2019-06-29 19:02:48 -04:00
|
|
|
{
|
2019-07-08 10:55:37 -04:00
|
|
|
for (int cnt = 0; cnt <= 7; cnt++)
|
|
|
|
{
|
|
|
|
ulong rnd1 = GenLeadingSignsMinus8(cnt);
|
|
|
|
ulong rnd2 = GenLeadingSignsPlus8(cnt);
|
|
|
|
|
|
|
|
yield return (rnd1 << 56) | (rnd1 << 48) | (rnd1 << 40) | (rnd1 << 32) |
|
|
|
|
(rnd1 << 24) | (rnd1 << 16) | (rnd1 << 08) | rnd1;
|
|
|
|
yield return (rnd2 << 56) | (rnd2 << 48) | (rnd2 << 40) | (rnd2 << 32) |
|
|
|
|
(rnd2 << 24) | (rnd2 << 16) | (rnd2 << 08) | rnd2;
|
|
|
|
}
|
2019-06-29 19:02:48 -04:00
|
|
|
}
|
|
|
|
|
2019-07-08 10:55:37 -04:00
|
|
|
private static IEnumerable<ulong> _GenLeadingSigns4H_()
|
2019-06-29 19:02:48 -04:00
|
|
|
{
|
2019-07-08 10:55:37 -04:00
|
|
|
for (int cnt = 0; cnt <= 15; cnt++)
|
|
|
|
{
|
|
|
|
ulong rnd1 = GenLeadingSignsMinus16(cnt);
|
|
|
|
ulong rnd2 = GenLeadingSignsPlus16(cnt);
|
|
|
|
|
|
|
|
yield return (rnd1 << 48) | (rnd1 << 32) | (rnd1 << 16) | rnd1;
|
|
|
|
yield return (rnd2 << 48) | (rnd2 << 32) | (rnd2 << 16) | rnd2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
private static IEnumerable<ulong> _GenLeadingSigns2S_()
|
|
|
|
{
|
|
|
|
for (int cnt = 0; cnt <= 31; cnt++)
|
|
|
|
{
|
|
|
|
ulong rnd1 = GenLeadingSignsMinus32(cnt);
|
|
|
|
ulong rnd2 = GenLeadingSignsPlus32(cnt);
|
|
|
|
|
|
|
|
yield return (rnd1 << 32) | rnd1;
|
|
|
|
yield return (rnd2 << 32) | rnd2;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
private static IEnumerable<ulong> _GenLeadingZeros8B_()
|
|
|
|
{
|
|
|
|
for (int cnt = 0; cnt <= 8; cnt++)
|
|
|
|
{
|
|
|
|
ulong rnd = GenLeadingZeros8(cnt);
|
|
|
|
|
|
|
|
yield return (rnd << 56) | (rnd << 48) | (rnd << 40) | (rnd << 32) |
|
|
|
|
(rnd << 24) | (rnd << 16) | (rnd << 08) | rnd;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
private static IEnumerable<ulong> _GenLeadingZeros4H_()
|
|
|
|
{
|
|
|
|
for (int cnt = 0; cnt <= 16; cnt++)
|
|
|
|
{
|
|
|
|
ulong rnd = GenLeadingZeros16(cnt);
|
|
|
|
|
|
|
|
yield return (rnd << 48) | (rnd << 32) | (rnd << 16) | rnd;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
private static IEnumerable<ulong> _GenLeadingZeros2S_()
|
|
|
|
{
|
|
|
|
for (int cnt = 0; cnt <= 32; cnt++)
|
|
|
|
{
|
|
|
|
ulong rnd = GenLeadingZeros32(cnt);
|
|
|
|
|
|
|
|
yield return (rnd << 32) | rnd;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
private static IEnumerable<ulong> _GenPopCnt8B_()
|
|
|
|
{
|
|
|
|
for (ulong cnt = 0ul; cnt <= 255ul; cnt++)
|
|
|
|
{
|
|
|
|
yield return (cnt << 56) | (cnt << 48) | (cnt << 40) | (cnt << 32) |
|
|
|
|
(cnt << 24) | (cnt << 16) | (cnt << 08) | cnt;
|
|
|
|
}
|
2019-06-29 19:02:48 -04:00
|
|
|
}
|
2018-09-17 00:54:05 -04:00
|
|
|
#endregion
|
|
|
|
|
|
|
|
#region "ValueSource (Opcodes)"
|
2019-05-29 20:29:24 -04:00
|
|
|
private static uint[] _SU_Add_Max_Min_V_V_8BB_4HH_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0E31B800u, // ADDV B0, V0.8B
|
|
|
|
0x0E30A800u, // SMAXV B0, V0.8B
|
|
|
|
0x0E31A800u, // SMINV B0, V0.8B
|
|
|
|
0x2E30A800u, // UMAXV B0, V0.8B
|
|
|
|
0x2E31A800u // UMINV B0, V0.8B
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _SU_Add_Max_Min_V_V_16BB_8HH_4SS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x4E31B800u, // ADDV B0, V0.16B
|
|
|
|
0x4E30A800u, // SMAXV B0, V0.16B
|
|
|
|
0x4E31A800u, // SMINV B0, V0.16B
|
|
|
|
0x6E30A800u, // UMAXV B0, V0.16B
|
|
|
|
0x6E31A800u // UMINV B0, V0.16B
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
private static uint[] _F_Abs_Neg_Recpx_Sqrt_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E20C020u, // FABS S0, S1
|
|
|
|
0x1E214020u, // FNEG S0, S1
|
|
|
|
0x5EA1F820u, // FRECPX S0, S1
|
|
|
|
0x1E21C020u // FSQRT S0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Abs_Neg_Recpx_Sqrt_S_D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E60C020u, // FABS D0, D1
|
|
|
|
0x1E614020u, // FNEG D0, D1
|
|
|
|
0x5EE1F820u, // FRECPX D0, D1
|
|
|
|
0x1E61C020u // FSQRT D0, D1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Abs_Neg_Sqrt_V_2S_4S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0EA0F800u, // FABS V0.2S, V0.2S
|
|
|
|
0x2EA0F800u, // FNEG V0.2S, V0.2S
|
|
|
|
0x2EA1F800u // FSQRT V0.2S, V0.2S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Abs_Neg_Sqrt_V_2D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x4EE0F800u, // FABS V0.2D, V0.2D
|
|
|
|
0x6EE0F800u, // FNEG V0.2D, V0.2D
|
|
|
|
0x6EE1F800u // FSQRT V0.2D, V0.2D
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Add_P_S_2SS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x7E30D820u // FADDP S0, V1.2S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Add_P_S_2DD_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x7E70D820u // FADDP D0, V1.2D
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cm_EqGeGtLeLt_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x5EA0D820u, // FCMEQ S0, S1, #0.0
|
|
|
|
0x7EA0C820u, // FCMGE S0, S1, #0.0
|
|
|
|
0x5EA0C820u, // FCMGT S0, S1, #0.0
|
|
|
|
0x7EA0D820u, // FCMLE S0, S1, #0.0
|
|
|
|
0x5EA0E820u // FCMLT S0, S1, #0.0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cm_EqGeGtLeLt_S_D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x5EE0D820u, // FCMEQ D0, D1, #0.0
|
|
|
|
0x7EE0C820u, // FCMGE D0, D1, #0.0
|
|
|
|
0x5EE0C820u, // FCMGT D0, D1, #0.0
|
|
|
|
0x7EE0D820u, // FCMLE D0, D1, #0.0
|
|
|
|
0x5EE0E820u // FCMLT D0, D1, #0.0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cm_EqGeGtLeLt_V_2S_4S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0EA0D800u, // FCMEQ V0.2S, V0.2S, #0.0
|
|
|
|
0x2EA0C800u, // FCMGE V0.2S, V0.2S, #0.0
|
|
|
|
0x0EA0C800u, // FCMGT V0.2S, V0.2S, #0.0
|
|
|
|
0x2EA0D800u, // FCMLE V0.2S, V0.2S, #0.0
|
|
|
|
0x0EA0E800u // FCMLT V0.2S, V0.2S, #0.0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cm_EqGeGtLeLt_V_2D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x4EE0D800u, // FCMEQ V0.2D, V0.2D, #0.0
|
|
|
|
0x6EE0C800u, // FCMGE V0.2D, V0.2D, #0.0
|
|
|
|
0x4EE0C800u, // FCMGT V0.2D, V0.2D, #0.0
|
|
|
|
0x6EE0D800u, // FCMLE V0.2D, V0.2D, #0.0
|
|
|
|
0x4EE0E800u // FCMLT V0.2D, V0.2D, #0.0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-11-17 21:41:16 -05:00
|
|
|
private static uint[] _F_Cmp_Cmpe_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E202028u, // FCMP S1, #0.0
|
|
|
|
0x1E202038u // FCMPE S1, #0.0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cmp_Cmpe_S_D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E602028u, // FCMP D1, #0.0
|
|
|
|
0x1E602038u // FCMPE D1, #0.0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
private static uint[] _F_Cvt_S_SD_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E22C020u // FCVT D0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cvt_S_DS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E624020u // FCVT S0, D1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2019-05-30 18:51:39 -04:00
|
|
|
private static uint[] _F_Cvt_S_SH_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E23C020u // FCVT H0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cvt_S_HS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1EE24020u // FCVT S0, H1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-03-24 17:53:49 -04:00
|
|
|
private static uint[] _F_Cvt_ANZ_SU_S_S_()
|
2018-09-17 00:54:05 -04:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2020-03-24 17:53:49 -04:00
|
|
|
0x5E21C820u, // FCVTAS S0, S1
|
|
|
|
0x7E21C820u, // FCVTAU S0, S1
|
2018-09-17 00:54:05 -04:00
|
|
|
0x5E21A820u, // FCVTNS S0, S1
|
|
|
|
0x7E21A820u, // FCVTNU S0, S1
|
|
|
|
0x5EA1B820u, // FCVTZS S0, S1
|
|
|
|
0x7EA1B820u // FCVTZU S0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-03-24 17:53:49 -04:00
|
|
|
private static uint[] _F_Cvt_ANZ_SU_S_D_()
|
2018-09-17 00:54:05 -04:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2020-03-24 17:53:49 -04:00
|
|
|
0x5E61C820u, // FCVTAS D0, D1
|
|
|
|
0x7E61C820u, // FCVTAU D0, D1
|
2018-09-17 00:54:05 -04:00
|
|
|
0x5E61A820u, // FCVTNS D0, D1
|
|
|
|
0x7E61A820u, // FCVTNU D0, D1
|
|
|
|
0x5EE1B820u, // FCVTZS D0, D1
|
|
|
|
0x7EE1B820u // FCVTZU D0, D1
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-03-24 17:53:49 -04:00
|
|
|
private static uint[] _F_Cvt_ANZ_SU_V_2S_4S_()
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-09-17 00:54:05 -04:00
|
|
|
return new uint[]
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2020-03-24 17:53:49 -04:00
|
|
|
0x0E21C800u, // FCVTAS V0.2S, V0.2S
|
|
|
|
0x2E21C800u, // FCVTAU V0.2S, V0.2S
|
2018-09-17 00:54:05 -04:00
|
|
|
0x0E21A800u, // FCVTNS V0.2S, V0.2S
|
|
|
|
0x2E21A800u, // FCVTNU V0.2S, V0.2S
|
|
|
|
0x0EA1B800u, // FCVTZS V0.2S, V0.2S
|
|
|
|
0x2EA1B800u // FCVTZU V0.2S, V0.2S
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-03-24 17:53:49 -04:00
|
|
|
private static uint[] _F_Cvt_ANZ_SU_V_2D_()
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-09-17 00:54:05 -04:00
|
|
|
return new uint[]
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2020-03-24 17:53:49 -04:00
|
|
|
0x4E61C800u, // FCVTAS V0.2D, V0.2D
|
|
|
|
0x6E61C800u, // FCVTAU V0.2D, V0.2D
|
2018-09-17 00:54:05 -04:00
|
|
|
0x4E61A800u, // FCVTNS V0.2D, V0.2D
|
|
|
|
0x6E61A800u, // FCVTNU V0.2D, V0.2D
|
|
|
|
0x4EE1B800u, // FCVTZS V0.2D, V0.2D
|
|
|
|
0x6EE1B800u // FCVTZU V0.2D, V0.2D
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
};
|
|
|
|
}
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-10-23 10:12:45 -04:00
|
|
|
private static uint[] _F_Cvtl_V_4H4S_8H4S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0E217800u // FCVTL V0.4S, V0.4H
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cvtl_V_2S2D_4S2D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0E617800u // FCVTL V0.2D, V0.2S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cvtn_V_4S4H_4S8H_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0E216800u // FCVTN V0.4H, V0.4S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Cvtn_V_2D2S_2D4S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0E616800u // FCVTN V0.2S, V0.2D
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2020-05-27 12:51:59 -04:00
|
|
|
private static uint[] _F_Max_Min_Nm_V_V_4SS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x6E30C800u, // FMAXNMV S0, V0.4S
|
2020-07-13 07:08:47 -04:00
|
|
|
0x6E30F800u, // FMAXV S0, V0.4S
|
|
|
|
0x6EB0C800u, // FMINNMV S0, V0.4S
|
|
|
|
0x6EB0F800u // FMINV S0, V0.4S
|
2020-05-27 12:51:59 -04:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2019-06-29 19:02:48 -04:00
|
|
|
private static uint[] _F_Mov_Ftoi_SW_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E260000u // FMOV W0, S0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_Ftoi_DX_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x9E660000u // FMOV X0, D0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_Ftoi1_DX_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x9EAE0000u // FMOV X0, V0.D[1]
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_Itof_WS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E270000u // FMOV S0, W0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_Itof_XD_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x9E670000u // FMOV D0, X0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_Itof1_XD_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x9EAF0000u // FMOV V0.D[1], X0
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E204020u // FMOV S0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Mov_S_D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E604020u // FMOV D0, D1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
private static uint[] _F_Recpe_Rsqrte_S_S_()
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2018-12-26 12:11:36 -05:00
|
|
|
0x5EA1D820u, // FRECPE S0, S1
|
|
|
|
0x7EA1D820u // FRSQRTE S0, S1
|
2018-10-05 21:45:59 -04:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
private static uint[] _F_Recpe_Rsqrte_S_D_()
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2018-12-26 12:11:36 -05:00
|
|
|
0x5EE1D820u, // FRECPE D0, D1
|
|
|
|
0x7EE1D820u // FRSQRTE D0, D1
|
2018-10-05 21:45:59 -04:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
private static uint[] _F_Recpe_Rsqrte_V_2S_4S_()
|
2018-11-17 21:41:16 -05:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2018-12-26 12:11:36 -05:00
|
|
|
0x0EA1D800u, // FRECPE V0.2S, V0.2S
|
|
|
|
0x2EA1D800u // FRSQRTE V0.2S, V0.2S
|
2018-11-17 21:41:16 -05:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
private static uint[] _F_Recpe_Rsqrte_V_2D_()
|
2018-11-17 21:41:16 -05:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2018-12-26 12:11:36 -05:00
|
|
|
0x4EE1D800u, // FRECPE V0.2D, V0.2D
|
|
|
|
0x6EE1D800u // FRSQRTE V0.2D, V0.2D
|
2018-11-17 21:41:16 -05:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2019-04-25 18:58:29 -04:00
|
|
|
private static uint[] _F_Rint_AMNPZ_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E264020u, // FRINTA S0, S1
|
|
|
|
0x1E254020u, // FRINTM S0, S1
|
|
|
|
0x1E244020u, // FRINTN S0, S1
|
|
|
|
0x1E24C020u, // FRINTP S0, S1
|
|
|
|
0x1E25C020u // FRINTZ S0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_AMNPZ_S_D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E664020u, // FRINTA D0, D1
|
|
|
|
0x1E654020u, // FRINTM D0, D1
|
|
|
|
0x1E644020u, // FRINTN D0, D1
|
|
|
|
0x1E64C020u, // FRINTP D0, D1
|
|
|
|
0x1E65C020u // FRINTZ D0, D1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_AMNPZ_V_2S_4S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x2E218800u, // FRINTA V0.2S, V0.2S
|
|
|
|
0x0E219800u, // FRINTM V0.2S, V0.2S
|
|
|
|
0x0E218800u, // FRINTN V0.2S, V0.2S
|
|
|
|
0x0EA18800u, // FRINTP V0.2S, V0.2S
|
|
|
|
0x0EA19800u // FRINTZ V0.2S, V0.2S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_AMNPZ_V_2D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x6E618800u, // FRINTA V0.2D, V0.2D
|
|
|
|
0x4E619800u, // FRINTM V0.2D, V0.2D
|
|
|
|
0x4E618800u, // FRINTN V0.2D, V0.2D
|
|
|
|
0x4EE18800u, // FRINTP V0.2D, V0.2D
|
|
|
|
0x4EE19800u // FRINTZ V0.2D, V0.2D
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_IX_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E27C020u, // FRINTI S0, S1
|
|
|
|
0x1E274020u // FRINTX S0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_IX_S_D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x1E67C020u, // FRINTI D0, D1
|
|
|
|
0x1E674020u // FRINTX D0, D1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_IX_V_2S_4S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x2EA19800u, // FRINTI V0.2S, V0.2S
|
|
|
|
0x2E219800u // FRINTX V0.2S, V0.2S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _F_Rint_IX_V_2D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x6EE19800u, // FRINTI V0.2D, V0.2D
|
|
|
|
0x6E619800u // FRINTX V0.2D, V0.2D
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2019-06-29 19:02:48 -04:00
|
|
|
private static uint[] _SU_Addl_V_V_8BH_4HS_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x0E303800u, // SADDLV H0, V0.8B
|
|
|
|
0x2E303800u // UADDLV H0, V0.8B
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _SU_Addl_V_V_16BH_8HS_4SD_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x4E303800u, // SADDLV H0, V0.16B
|
|
|
|
0x6E303800u // UADDLV H0, V0.16B
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-11-17 21:41:16 -05:00
|
|
|
private static uint[] _SU_Cvt_F_S_S_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x5E21D820u, // SCVTF S0, S1
|
|
|
|
0x7E21D820u // UCVTF S0, S1
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _SU_Cvt_F_S_D_()
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2018-11-17 21:41:16 -05:00
|
|
|
0x5E61D820u, // SCVTF D0, D1
|
|
|
|
0x7E61D820u // UCVTF D0, D1
|
2018-10-05 21:45:59 -04:00
|
|
|
};
|
|
|
|
}
|
|
|
|
|
2018-11-17 21:41:16 -05:00
|
|
|
private static uint[] _SU_Cvt_F_V_2S_4S_()
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
2018-11-17 21:41:16 -05:00
|
|
|
0x0E21D800u, // SCVTF V0.2S, V0.2S
|
|
|
|
0x2E21D800u // UCVTF V0.2S, V0.2S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _SU_Cvt_F_V_2D_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x4E61D800u, // SCVTF V0.2D, V0.2D
|
|
|
|
0x6E61D800u // UCVTF V0.2D, V0.2D
|
2018-10-05 21:45:59 -04:00
|
|
|
};
|
|
|
|
}
|
2018-10-28 18:27:50 -04:00
|
|
|
|
|
|
|
private static uint[] _Sha1h_Sha1su1_V_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x5E280800u, // SHA1H S0, S0
|
|
|
|
0x5E281800u // SHA1SU1 V0.4S, V0.4S
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
private static uint[] _Sha256su0_V_()
|
|
|
|
{
|
|
|
|
return new uint[]
|
|
|
|
{
|
|
|
|
0x5E282800u // SHA256SU0 V0.4S, V0.4S
|
|
|
|
};
|
|
|
|
}
|
2018-04-20 11:40:15 -04:00
|
|
|
#endregion
|
|
|
|
|
2018-09-08 13:23:07 -04:00
|
|
|
private const int RndCnt = 2;
|
2018-07-14 23:53:26 -04:00
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
private static readonly bool NoZeros = false;
|
|
|
|
private static readonly bool NoInfs = false;
|
|
|
|
private static readonly bool NoNaNs = false;
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("ABS <V><d>, <V><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Abs_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-04-20 11:40:15 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5EE0B800; // ABS D0, D0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-20 11:40:15 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("ABS <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Abs_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-04-20 11:40:15 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E20B800; // ABS V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-20 11:40:15 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("ABS <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Abs_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-20 11:40:15 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E20B800; // ABS V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-20 11:40:15 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("ADDP <V><d>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Addp_S_2DD([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-04-21 15:15:04 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5EF1B800; // ADDP D0, V0.2D
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-21 15:15:04 -04:00
|
|
|
}
|
|
|
|
|
2019-05-29 20:29:24 -04:00
|
|
|
[Test, Pairwise]
|
|
|
|
public void SU_Add_Max_Min_V_V_8BB_4HH([ValueSource("_SU_Add_Max_Min_V_V_8BB_4HH_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b00u, 0b01u)] uint size) // <8BB, 4HH>
|
2018-04-21 15:15:04 -04:00
|
|
|
{
|
2019-05-29 20:29:24 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((size & 3) << 22);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
2019-05-29 20:29:24 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-21 15:15:04 -04:00
|
|
|
}
|
|
|
|
|
2019-05-29 20:29:24 -04:00
|
|
|
[Test, Pairwise]
|
|
|
|
public void SU_Add_Max_Min_V_V_16BB_8HH_4SS([ValueSource("_SU_Add_Max_Min_V_V_16BB_8HH_4SS_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16BB, 8HH, 4SS>
|
2018-04-21 15:15:04 -04:00
|
|
|
{
|
2019-05-29 20:29:24 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((size & 3) << 22);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
2019-05-29 20:29:24 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
2018-04-21 15:15:04 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-21 15:15:04 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CLS <Vd>.<T>, <Vn>.<T>")]
|
2019-06-12 08:03:31 -04:00
|
|
|
public void Cls_V_8B_16B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
2019-06-12 08:03:31 -04:00
|
|
|
[ValueSource("_GenLeadingSigns8B_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <8B, 16B>
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E204800; // CLS V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2019-06-12 08:03:31 -04:00
|
|
|
opcode |= ((q & 1) << 30);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CLS <Vd>.<T>, <Vn>.<T>")]
|
2019-06-12 08:03:31 -04:00
|
|
|
public void Cls_V_4H_8H([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_4H_")] [Random(RndCnt)] ulong z,
|
2019-06-12 08:03:31 -04:00
|
|
|
[ValueSource("_GenLeadingSigns4H_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <4H, 8H>
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
{
|
2019-06-12 08:03:31 -04:00
|
|
|
uint opcode = 0x0E604800; // CLS V0.4H, V0.4H
|
2018-11-01 00:22:09 -04:00
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2019-06-12 08:03:31 -04:00
|
|
|
opcode |= ((q & 1) << 30);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2019-06-12 08:03:31 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise, Description("CLS <Vd>.<T>, <Vn>.<T>")]
|
|
|
|
public void Cls_V_2S_4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_2S_")] [Random(RndCnt)] ulong z,
|
2019-06-12 08:03:31 -04:00
|
|
|
[ValueSource("_GenLeadingSigns2S_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
|
|
|
{
|
|
|
|
uint opcode = 0x0EA04800; // CLS V0.2S, V0.2S
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CLZ <Vd>.<T>, <Vn>.<T>")]
|
2019-06-12 08:03:31 -04:00
|
|
|
public void Clz_V_8B_16B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
2019-06-12 08:03:31 -04:00
|
|
|
[ValueSource("_GenLeadingZeros8B_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <8B, 16B>
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E204800; // CLZ V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2019-06-12 08:03:31 -04:00
|
|
|
opcode |= ((q & 1) << 30);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CLZ <Vd>.<T>, <Vn>.<T>")]
|
2019-06-12 08:03:31 -04:00
|
|
|
public void Clz_V_4H_8H([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_4H_")] [Random(RndCnt)] ulong z,
|
2019-06-12 08:03:31 -04:00
|
|
|
[ValueSource("_GenLeadingZeros4H_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <4H, 8H>
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
{
|
2019-06-12 08:03:31 -04:00
|
|
|
uint opcode = 0x2E604800; // CLZ V0.4H, V0.4H
|
2018-11-01 00:22:09 -04:00
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2019-06-12 08:03:31 -04:00
|
|
|
opcode |= ((q & 1) << 30);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2019-06-12 08:03:31 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise, Description("CLZ <Vd>.<T>, <Vn>.<T>")]
|
|
|
|
public void Clz_V_2S_4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_2S_")] [Random(RndCnt)] ulong z,
|
2019-06-12 08:03:31 -04:00
|
|
|
[ValueSource("_GenLeadingZeros2S_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
|
|
|
{
|
|
|
|
uint opcode = 0x2EA04800; // CLZ V0.2S, V0.2S
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Cls_V, Clz_V, Orn_V instructions. Add 18 Tests: And_V, Bic_V, Bif_V, Bit_V, Bsl_V, Cls_V, Clz_V, Orn_V, Orr_V. (#104)
* Update AOpCodeTable.cs
* Update AInstEmitSimdLogical.cs
* Update AInstEmitSimdArithmetic.cs
* Update ASoftFallback.cs
* Update AInstEmitAlu.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
2018-04-25 22:20:22 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMEQ <V><d>, <V><n>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmeq_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5EE09800; // CMEQ D0, D0, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMEQ <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmeq_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E209800; // CMEQ V0.8B, V0.8B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMEQ <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmeq_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E209800; // CMEQ V0.16B, V0.16B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMGE <V><d>, <V><n>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmge_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7EE08800; // CMGE D0, D0, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMGE <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmge_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E208800; // CMGE V0.8B, V0.8B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMGE <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmge_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E208800; // CMGE V0.16B, V0.16B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMGT <V><d>, <V><n>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmgt_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5EE08800; // CMGT D0, D0, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMGT <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmgt_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E208800; // CMGT V0.8B, V0.8B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMGT <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmgt_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E208800; // CMGT V0.16B, V0.16B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMLE <V><d>, <V><n>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmle_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7EE09800; // CMLE D0, D0, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMLE <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmle_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E209800; // CMLE V0.8B, V0.8B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMLE <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmle_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E209800; // CMLE V0.16B, V0.16B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMLT <V><d>, <V><n>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmlt_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5EE0A800; // CMLT D0, D0, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMLT <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmlt_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E20A800; // CMLT V0.8B, V0.8B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CMLT <Vd>.<T>, <Vn>.<T>, #0")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cmlt_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-18 13:55:26 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E20A800; // CMLT V0.16B, V0.16B, #0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CNT <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cnt_V_8B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_GenPopCnt8B_")] [Random(RndCnt)] ulong a)
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E205800; // CNT V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("CNT <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Cnt_V_16B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
2019-07-08 10:55:37 -04:00
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_GenPopCnt8B_")] [Random(RndCnt)] ulong a)
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E205800; // CNT V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
Add Sse Opt. for Cmeq_V_2D, Cmgt_V_2D (Reg). Add Sse Opt. for Crc32cb, Crc32ch, Crc32cw, Crc32cx. Add 10 simple tests for Fcmgt, Fcmge, Fcmeq, Fcmle, Fcmlt (S, V) (Reg, Zero). Add 2 Cnt_V tests. (#183)
* Add files via upload
* Add files via upload
* Add files via upload
* CPE
* Add EmitSse42Crc32()
* Update CpuTestSimdCmp.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update Instructions.cs
2018-06-25 21:32:29 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Abs_Neg_Recpx_Sqrt_S_S([ValueSource("_F_Abs_Neg_Recpx_Sqrt_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Abs_Neg_Recpx_Sqrt_S_D([ValueSource("_F_Abs_Neg_Recpx_Sqrt_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Abs_Neg_Sqrt_V_2S_4S([ValueSource("_F_Abs_Neg_Sqrt_V_2S_4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Abs_Neg_Sqrt_V_2D([ValueSource("_F_Abs_Neg_Sqrt_V_2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong z,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Add_P_S_2SS([ValueSource("_F_Add_P_S_2SS_")] uint opcodes,
|
|
|
|
[ValueSource("_2S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Add_P_S_2DD([ValueSource("_F_Add_P_S_2DD_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cm_EqGeGtLeLt_S_S([ValueSource("_F_Cm_EqGeGtLeLt_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cm_EqGeGtLeLt_S_D([ValueSource("_F_Cm_EqGeGtLeLt_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cm_EqGeGtLeLt_V_2S_4S([ValueSource("_F_Cm_EqGeGtLeLt_V_2S_4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cm_EqGeGtLeLt_V_2D([ValueSource("_F_Cm_EqGeGtLeLt_V_2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong z,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
2018-11-17 21:41:16 -05:00
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cmp_Cmpe_S_S([ValueSource("_F_Cmp_Cmpe_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
|
|
|
{
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-11-17 21:41:16 -05:00
|
|
|
|
|
|
|
bool v = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
bool c = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
bool z = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
bool n = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v1: v1, overflow: v, carry: c, zero: z, negative: n);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc);
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cmp_Cmpe_S_D([ValueSource("_F_Cmp_Cmpe_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-11-17 21:41:16 -05:00
|
|
|
|
|
|
|
bool v = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
bool c = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
bool z = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
bool n = TestContext.CurrentContext.Random.NextBool();
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v1: v1, overflow: v, carry: c, zero: z, negative: n);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc);
|
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void F_Cvt_S_SD([ValueSource("_F_Cvt_S_SD_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-10-05 21:45:59 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void F_Cvt_S_DS([ValueSource("_F_Cvt_S_DS_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-05-30 18:51:39 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
2020-11-18 13:35:54 -05:00
|
|
|
[Test, Pairwise] [Explicit] // Unicorn seems to default all rounding modes to RMode.Rn.
|
2019-05-30 18:51:39 -04:00
|
|
|
public void F_Cvt_S_SH([ValueSource("_F_Cvt_S_SH_")] uint opcodes,
|
2020-11-18 13:35:54 -05:00
|
|
|
[ValueSource("_1S_F_")] ulong a,
|
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2019-05-30 18:51:39 -04:00
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-05-30 18:51:39 -04:00
|
|
|
|
2020-11-18 13:35:54 -05:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
2019-05-30 18:51:39 -04:00
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Cvt_S_HS([ValueSource("_F_Cvt_S_HS_")] uint opcodes,
|
|
|
|
[ValueSource("_1H_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2020-03-24 17:53:49 -04:00
|
|
|
public void F_Cvt_ANZ_SU_S_S([ValueSource("_F_Cvt_ANZ_SU_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_W_")] ulong a)
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-10-05 21:45:59 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2020-03-24 17:53:49 -04:00
|
|
|
public void F_Cvt_ANZ_SU_S_D([ValueSource("_F_Cvt_ANZ_SU_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_X_")] ulong a)
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2020-03-24 17:53:49 -04:00
|
|
|
public void F_Cvt_ANZ_SU_V_2S_4S([ValueSource("_F_Cvt_ANZ_SU_V_2S_4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_W_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_W_")] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2020-03-24 17:53:49 -04:00
|
|
|
public void F_Cvt_ANZ_SU_V_2D([ValueSource("_F_Cvt_ANZ_SU_V_2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_X_")] ulong z,
|
|
|
|
[ValueSource("_1D_F_X_")] ulong a)
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
|
2018-09-17 00:54:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
}
|
|
|
|
|
2018-10-23 10:12:45 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void F_Cvtl_V_4H4S_8H4S([ValueSource("_F_Cvtl_V_4H4S_8H4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H_F_")] ulong z,
|
|
|
|
[ValueSource("_4H_F_")] ulong a,
|
2018-12-26 12:11:36 -05:00
|
|
|
[Values(0b0u, 0b1u)] uint q, // <4H4S, 8H4S>
|
2018-11-01 00:22:09 -04:00
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2018-10-23 10:12:45 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(q == 0u ? a : 0ul, q == 1u ? a : 0ul);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Ahp);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Ofc | Fpsr.Ufc | Fpsr.Ixc);
|
2018-10-23 10:12:45 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void F_Cvtl_V_2S2D_4S2D([ValueSource("_F_Cvtl_V_2S2D_4S2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
2018-12-26 12:11:36 -05:00
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S2D, 4S2D>
|
2018-10-23 10:12:45 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(q == 0u ? a : 0ul, q == 1u ? a : 0ul);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
[Test, Pairwise] [Explicit] // Unicorn seems to default all rounding modes to RMode.Rn.
|
|
|
|
public void F_Cvtn_V_4S4H_4S8H([ValueSource("_F_Cvtn_V_4S4H_4S8H_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
2018-12-26 12:11:36 -05:00
|
|
|
[Values(0b0u, 0b1u)] uint q, // <4S4H, 4S8H>
|
2018-11-01 00:22:09 -04:00
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2018-10-23 10:12:45 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Ahp);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Ofc | Fpsr.Ufc | Fpsr.Ixc | Fpsr.Idc);
|
2018-10-23 10:12:45 -04:00
|
|
|
}
|
|
|
|
|
2020-11-18 13:35:54 -05:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void F_Cvtn_V_2D2S_2D4S([ValueSource("_F_Cvtn_V_2D2S_2D4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong z,
|
|
|
|
[ValueSource("_1D_F_")] ulong a,
|
2018-12-26 12:11:36 -05:00
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2D2S, 2D4S>
|
2018-10-23 10:12:45 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
2018-10-23 10:12:45 -04:00
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
2020-05-27 12:51:59 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Max_Min_Nm_V_V_4SS([ValueSource("_F_Max_Min_Nm_V_V_4SS_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
|
|
|
|
int fpcr = rnd & (1 << (int)Fpcr.Fz);
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Idc);
|
|
|
|
}
|
|
|
|
|
2019-06-29 19:02:48 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_Ftoi_SW([ValueSource("_F_Mov_Ftoi_SW_")] uint opcodes,
|
|
|
|
[Values(0u, 31u)] uint rd,
|
|
|
|
[Values(1u)] uint rn,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
ulong x0 = (ulong)TestContext.CurrentContext.Random.NextUInt() << 32;
|
|
|
|
uint w31 = TestContext.CurrentContext.Random.NextUInt();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, x0: x0, x31: w31, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_Ftoi_DX([ValueSource("_F_Mov_Ftoi_DX_")] uint opcodes,
|
|
|
|
[Values(0u, 31u)] uint rd,
|
|
|
|
[Values(1u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
ulong x31 = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, x31: x31, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_Ftoi1_DX([ValueSource("_F_Mov_Ftoi1_DX_")] uint opcodes,
|
|
|
|
[Values(0u, 31u)] uint rd,
|
|
|
|
[Values(1u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
ulong x31 = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v1 = MakeVectorE1(a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, x31: x31, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_Itof_WS([ValueSource("_F_Mov_Itof_WS_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 31u)] uint rn,
|
|
|
|
[ValueSource("_W_")] [Random(RndCnt)] uint wn)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
uint w31 = TestContext.CurrentContext.Random.NextUInt();
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, x1: wn, x31: w31, v0: v0);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_Itof_XD([ValueSource("_F_Mov_Itof_XD_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 31u)] uint rn,
|
|
|
|
[ValueSource("_X_")] [Random(RndCnt)] ulong xn)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
ulong x31 = TestContext.CurrentContext.Random.NextULong();
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, x1: xn, x31: x31, v0: v0);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_Itof1_XD([ValueSource("_F_Mov_Itof1_XD_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 31u)] uint rn,
|
|
|
|
[ValueSource("_X_")] [Random(RndCnt)] ulong xn)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
|
|
|
ulong x31 = TestContext.CurrentContext.Random.NextULong();
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0(z);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, x1: xn, x31: x31, v0: v0);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_S_S([ValueSource("_F_Mov_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Mov_S_D([ValueSource("_F_Mov_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-12-26 12:11:36 -05:00
|
|
|
public void F_Recpe_Rsqrte_S_S([ValueSource("_F_Recpe_Rsqrte_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a,
|
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-11-01 00:22:09 -04:00
|
|
|
|
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Fz);
|
2018-11-01 00:22:09 -04:00
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Dzc | Fpsr.Ofc | Fpsr.Ufc | Fpsr.Ixc | Fpsr.Idc);
|
2018-10-05 21:45:59 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-12-26 12:11:36 -05:00
|
|
|
public void F_Recpe_Rsqrte_S_D([ValueSource("_F_Recpe_Rsqrte_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a,
|
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Fz);
|
2018-11-01 00:22:09 -04:00
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Dzc | Fpsr.Ofc | Fpsr.Ufc | Fpsr.Ixc | Fpsr.Idc);
|
2018-10-05 21:45:59 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-12-26 12:11:36 -05:00
|
|
|
public void F_Recpe_Rsqrte_V_2S_4S([ValueSource("_F_Recpe_Rsqrte_V_2S_4S_")] uint opcodes,
|
2018-11-17 21:41:16 -05:00
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
2018-12-26 12:11:36 -05:00
|
|
|
[Values(0b0u, 0b1u)] uint q, // <2S, 4S>
|
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Fz);
|
2018-11-01 00:22:09 -04:00
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Dzc | Fpsr.Ofc | Fpsr.Ufc | Fpsr.Ixc | Fpsr.Idc);
|
2018-10-05 21:45:59 -04:00
|
|
|
}
|
|
|
|
|
2018-10-13 22:35:16 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
2018-12-26 12:11:36 -05:00
|
|
|
public void F_Recpe_Rsqrte_V_2D([ValueSource("_F_Recpe_Rsqrte_V_2D_")] uint opcodes,
|
2018-11-17 21:41:16 -05:00
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong z,
|
2018-12-26 12:11:36 -05:00
|
|
|
[ValueSource("_1D_F_")] ulong a,
|
|
|
|
[Values(RMode.Rn)] RMode rMode)
|
2018-10-05 21:45:59 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
int rnd = (int)TestContext.CurrentContext.Random.NextUInt();
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Fz);
|
2018-11-01 00:22:09 -04:00
|
|
|
fpcr |= rnd & (1 << (int)Fpcr.Dn);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
2018-10-05 21:45:59 -04:00
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Ioc | Fpsr.Dzc | Fpsr.Ofc | Fpsr.Ufc | Fpsr.Ixc | Fpsr.Idc);
|
2018-10-05 21:45:59 -04:00
|
|
|
}
|
|
|
|
|
2019-04-25 18:58:29 -04:00
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_AMNPZ_S_S([ValueSource("_F_Rint_AMNPZ_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_AMNPZ_S_D([ValueSource("_F_Rint_AMNPZ_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_AMNPZ_V_2S_4S([ValueSource("_F_Rint_AMNPZ_V_2S_4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_AMNPZ_V_2D([ValueSource("_F_Rint_AMNPZ_V_2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong z,
|
|
|
|
[ValueSource("_1D_F_")] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_IX_S_S([ValueSource("_F_Rint_IX_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_F_")] ulong a,
|
|
|
|
[Values] RMode rMode)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_IX_S_D([ValueSource("_F_Rint_IX_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_F_")] ulong a,
|
|
|
|
[Values] RMode rMode)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_IX_V_2S_4S([ValueSource("_F_Rint_IX_V_2S_4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_F_")] ulong z,
|
|
|
|
[ValueSource("_2S_F_")] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q, // <2S, 4S>
|
|
|
|
[Values] RMode rMode)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void F_Rint_IX_V_2D([ValueSource("_F_Rint_IX_V_2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_F_")] ulong z,
|
|
|
|
[ValueSource("_1D_F_")] ulong a,
|
|
|
|
[Values] RMode rMode)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2019-04-25 18:58:29 -04:00
|
|
|
|
|
|
|
int fpcr = (int)rMode << (int)Fpcr.RMode;
|
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1, fpcr: fpcr);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("NEG <V><d>, <V><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Neg_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
2018-04-20 11:40:15 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7EE0B800; // NEG D0, D0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-20 11:40:15 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("NEG <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Neg_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-04-20 11:40:15 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E20B800; // NEG V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-20 11:40:15 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("NEG <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Neg_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-20 11:40:15 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E20B800; // NEG V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-20 11:40:15 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-04-20 11:40:15 -04:00
|
|
|
}
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("NOT <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Not_V_8B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E205800; // NOT V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("NOT <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Not_V_16B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong a)
|
2018-06-18 13:55:26 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E205800; // NOT V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-18 13:55:26 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-06-18 13:55:26 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("RBIT <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rbit_V_8B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong a)
|
2018-07-03 02:31:16 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E605800; // RBIT V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("RBIT <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rbit_V_16B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong a)
|
2018-07-03 02:31:16 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E605800; // RBIT V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("REV16 <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rev16_V_8B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong a)
|
2018-07-03 02:31:16 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E201800; // REV16 V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("REV16 <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rev16_V_16B([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B_")] [Random(RndCnt)] ulong a)
|
2018-07-03 02:31:16 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E201800; // REV16 V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("REV32 <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rev32_V_8B_4H([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong a,
|
2018-07-03 02:31:16 -04:00
|
|
|
[Values(0b00u, 0b01u)] uint size) // <8B, 4H>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E200800; // REV32 V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("REV32 <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rev32_V_16B_8H([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong a,
|
2018-07-03 02:31:16 -04:00
|
|
|
[Values(0b00u, 0b01u)] uint size) // <16B, 8H>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E200800; // REV32 V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("REV64 <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rev64_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-07-03 02:31:16 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E200800; // REV64 V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("REV64 <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Rev64_V_16B_8H_4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-07-03 02:31:16 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16B, 8H, 4S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E200800; // REV64 V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-03 02:31:16 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-03 02:31:16 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SADALP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sadalp_V_8B4H_4H2S_2S1D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B4H, 4H2S, 2S1D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E206800; // SADALP V0.4H, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SADALP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sadalp_V_16B8H_8H4S_4S2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H, 8H4S, 4S2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E206800; // SADALP V0.8H, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SADDLP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Saddlp_V_8B4H_4H2S_2S1D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B4H, 4H2S, 2S1D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E202800; // SADDLP V0.4H, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SADDLP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Saddlp_V_16B8H_8H4S_4S2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H, 8H4S, 4S2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E202800; // SADDLP V0.8H, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
2019-06-29 19:02:48 -04:00
|
|
|
[Test, Pairwise]
|
|
|
|
public void SU_Addl_V_V_8BH_4HS([ValueSource("_SU_Addl_V_V_8BH_4HS_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b00u, 0b01u)] uint size) // <8BH, 4HS>
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((size & 3) << 22);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise]
|
|
|
|
public void SU_Addl_V_V_16BH_8HS_4SD([ValueSource("_SU_Addl_V_V_16BH_8HS_4SD_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16BH, 8HS, 4SD>
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((size & 3) << 22);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2019-06-29 19:02:48 -04:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
2018-11-17 21:41:16 -05:00
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void SU_Cvt_F_S_S([ValueSource("_SU_Cvt_F_S_S_")] uint opcodes,
|
|
|
|
[ValueSource("_1S_")] [Random(RndCnt)] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-11-17 21:41:16 -05:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void SU_Cvt_F_S_D([ValueSource("_SU_Cvt_F_S_D_")] uint opcodes,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
|
|
|
{
|
|
|
|
ulong z = TestContext.CurrentContext.Random.NextULong();
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE1(z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-11-17 21:41:16 -05:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
2019-12-29 20:22:47 -05:00
|
|
|
CompareAgainstUnicorn();
|
2018-11-17 21:41:16 -05:00
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void SU_Cvt_F_V_2S_4S([ValueSource("_SU_Cvt_F_V_2S_4S_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_2S_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b0u, 0b1u)] uint q) // <2S, 4S>
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcodes |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a * q);
|
2018-11-17 21:41:16 -05:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise] [Explicit]
|
|
|
|
public void SU_Cvt_F_V_2D([ValueSource("_SU_Cvt_F_V_2D_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1D_")] [Random(RndCnt)] ulong a)
|
|
|
|
{
|
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-11-17 21:41:16 -05:00
|
|
|
|
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
|
|
|
|
2019-12-29 20:22:47 -05:00
|
|
|
CompareAgainstUnicorn();
|
2018-11-17 21:41:16 -05:00
|
|
|
}
|
|
|
|
|
2018-10-28 18:27:50 -04:00
|
|
|
[Test, Pairwise]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sha1h_Sha1su1_V([ValueSource("_Sha1h_Sha1su1_V_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[Random(RndCnt / 2)] ulong z0, [Random(RndCnt / 2)] ulong z1,
|
|
|
|
[Random(RndCnt / 2)] ulong a0, [Random(RndCnt / 2)] ulong a1)
|
2018-10-28 18:27:50 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-10-28 18:27:50 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z0, z1);
|
|
|
|
V128 v1 = MakeVectorE0E1(a0, a1);
|
2018-10-28 18:27:50 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
2018-10-28 18:27:50 -04:00
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
|
|
|
[Test, Pairwise]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sha256su0_V([ValueSource("_Sha256su0_V_")] uint opcodes,
|
|
|
|
[Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[Random(RndCnt / 2)] ulong z0, [Random(RndCnt / 2)] ulong z1,
|
|
|
|
[Random(RndCnt / 2)] ulong a0, [Random(RndCnt / 2)] ulong a1)
|
2018-08-16 20:44:44 -04:00
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
opcodes |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
2018-08-16 20:44:44 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z0, z1);
|
|
|
|
V128 v1 = MakeVectorE0E1(a0, a1);
|
2018-08-16 20:44:44 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcodes, v0: v0, v1: v1);
|
2018-08-16 20:44:44 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-16 20:44:44 -04:00
|
|
|
}
|
|
|
|
|
2018-12-26 12:11:36 -05:00
|
|
|
[Test, Pairwise, Description("SHLL{2} <Vd>.<Ta>, <Vn>.<Tb>, #<shift>")]
|
|
|
|
public void Shll_V([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size, // <shift: 8, 16, 32>
|
|
|
|
[Values(0b0u, 0b1u)] uint q)
|
|
|
|
{
|
|
|
|
uint opcode = 0x2E213800; // SHLL V0.8H, V0.8B, #8
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
|
|
|
opcode |= ((q & 1) << 30);
|
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(q == 0u ? a : 0ul, q == 1u ? a : 0ul);
|
2018-12-26 12:11:36 -05:00
|
|
|
|
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
|
|
|
|
|
|
|
CompareAgainstUnicorn();
|
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQABS <V><d>, <V><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqabs_S_B_H_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <B, H, S, D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5E207800; // SQABS B0, B0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQABS <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqabs_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E207800; // SQABS V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQABS <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqabs_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E207800; // SQABS V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQNEG <V><d>, <V><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqneg_S_B_H_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <B, H, S, D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7E207800; // SQNEG B0, B0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQNEG <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqneg_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E207800; // SQNEG V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQNEG <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqneg_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E207800; // SQNEG V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQXTN <Vb><d>, <Va><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqxtn_S_HB_SH_DS([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-29 19:39:58 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <HB, SH, DS>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5E214800; // SQXTN B0, H0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-04-29 19:39:58 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQXTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqxtn_V_8H8B_4S4H_2D2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-29 19:39:58 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H8B, 4S4H, 2D2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E214800; // SQXTN V0.8B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-04-29 19:39:58 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQXTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqxtn_V_8H16B_4S8H_2D4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-29 19:39:58 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H16B, 4S8H, 2D4S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E214800; // SQXTN2 V0.16B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-06-25 22:36:20 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQXTUN <Vb><d>, <Va><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqxtun_S_HB_SH_DS([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-25 22:36:20 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <HB, SH, DS>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7E212800; // SQXTUN B0, H0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-06-25 22:36:20 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQXTUN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqxtun_V_8H8B_4S4H_2D2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-25 22:36:20 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H8B, 4S4H, 2D2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E212800; // SQXTUN V0.8B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-06-25 22:36:20 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SQXTUN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Sqxtun_V_8H16B_4S8H_2D4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-06-25 22:36:20 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H16B, 4S8H, 2D4S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E212800; // SQXTUN2 V0.16B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-06-25 22:36:20 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SUQADD <V><d>, <V><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Suqadd_S_B_H_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <B, H, S, D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x5E203800; // SUQADD B0, B0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SUQADD <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Suqadd_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E203800; // SUQADD V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("SUQADD <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Suqadd_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E203800; // SUQADD V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-04-29 19:39:58 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UADALP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uadalp_V_8B4H_4H2S_2S1D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B4H, 4H2S, 2S1D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E206800; // UADALP V0.4H, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UADALP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uadalp_V_16B8H_8H4S_4S2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H, 8H4S, 4S2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E206800; // UADALP V0.8H, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UADDLP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uaddlp_V_8B4H_4H2S_2S1D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B4H, 4H2S, 2S1D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E202800; // UADDLP V0.4H, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UADDLP <Vd>.<Ta>, <Vn>.<Tb>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uaddlp_V_16B8H_8H4S_4S2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-13 17:10:02 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H, 8H4S, 4S2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E202800; // UADDLP V0.8H, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-13 17:10:02 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-08-13 17:10:02 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UQXTN <Vb><d>, <Va><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uqxtn_S_HB_SH_DS([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-29 19:39:58 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <HB, SH, DS>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7E214800; // UQXTN B0, H0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-04-29 19:39:58 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UQXTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uqxtn_V_8H8B_4S4H_2D2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-29 19:39:58 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H8B, 4S4H, 2D2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E214800; // UQXTN V0.8B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-04-29 19:39:58 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("UQXTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Uqxtn_V_8H16B_4S8H_2D4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-04-29 19:39:58 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H16B, 4S8H, 2D4S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E214800; // UQXTN2 V0.16B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-04-29 19:39:58 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("USQADD <V><d>, <V><n>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Usqadd_S_B_H_S_D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_1B1H1S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <B, H, S, D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x7E203800; // USQADD B0, B0
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("USQADD <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Usqadd_V_8B_4H_2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8B, 4H, 2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x2E203800; // USQADD V0.8B, V0.8B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0(a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-08-04 15:58:54 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("USQADD <Vd>.<T>, <Vn>.<T>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Usqadd_V_16B_8H_4S_2D([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-08-04 15:58:54 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u, 0b11u)] uint size) // <16B, 8H, 4S, 2D>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x6E203800; // USQADD V0.16B, V0.16B
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-08-04 15:58:54 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
CompareAgainstUnicorn(fpsrMask: Fpsr.Qc);
|
2018-04-29 19:39:58 -04:00
|
|
|
}
|
2018-07-18 20:06:28 -04:00
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("XTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Xtn_V_8H8B_4S4H_2D2S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-07-18 20:06:28 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H8B, 4S4H, 2D2S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x0E212800; // XTN V0.8B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-07-18 20:06:28 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-07-18 20:06:28 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-18 20:06:28 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-18 20:06:28 -04:00
|
|
|
}
|
|
|
|
|
Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)
* Update AOpCodeTable.cs
* Update AInstEmitSimdCvt.cs
* Update Pseudocode.cs
* Update Instructions.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Update CpuTestSimd.cs
* Update AOpCodeTable.cs
* Update AInstEmitSimdArithmetic.cs
* Update Instructions.cs
* Update CpuTestSimdReg.cs
* Add QCFlagBit.
* Add QCFlagBit.
2018-09-01 10:52:51 -04:00
|
|
|
[Test, Pairwise, Description("XTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
|
2018-11-01 00:22:09 -04:00
|
|
|
public void Xtn_V_8H16B_4S8H_2D4S([Values(0u)] uint rd,
|
|
|
|
[Values(1u, 0u)] uint rn,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong z,
|
|
|
|
[ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong a,
|
2018-07-18 20:06:28 -04:00
|
|
|
[Values(0b00u, 0b01u, 0b10u)] uint size) // <8H16B, 4S8H, 2D4S>
|
|
|
|
{
|
2018-11-01 00:22:09 -04:00
|
|
|
uint opcode = 0x4E212800; // XTN2 V0.16B, V0.8H
|
|
|
|
opcode |= ((rn & 31) << 5) | ((rd & 31) << 0);
|
|
|
|
opcode |= ((size & 3) << 22);
|
2018-07-18 20:06:28 -04:00
|
|
|
|
Add a new JIT compiler for CPU code (#693)
* Start of the ARMeilleure project
* Refactoring around the old IRAdapter, now renamed to PreAllocator
* Optimize the LowestBitSet method
* Add CLZ support and fix CLS implementation
* Add missing Equals and GetHashCode overrides on some structs, misc small tweaks
* Implement the ByteSwap IR instruction, and some refactoring on the assembler
* Implement the DivideUI IR instruction and fix 64-bits IDIV
* Correct constant operand type on CSINC
* Move division instructions implementation to InstEmitDiv
* Fix destination type for the ConditionalSelect IR instruction
* Implement UMULH and SMULH, with new IR instructions
* Fix some issues with shift instructions
* Fix constant types for BFM instructions
* Fix up new tests using the new V128 struct
* Update tests
* Move DIV tests to a separate file
* Add support for calls, and some instructions that depends on them
* Start adding support for SIMD & FP types, along with some of the related ARM instructions
* Fix some typos and the divide instruction with FP operands
* Fix wrong method call on Clz_V
* Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes
* Implement SIMD logical instructions and more misc. fixes
* Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations
* Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes
* Implement SIMD shift instruction and fix Dup_V
* Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table
* Fix check with tolerance on tester
* Implement FP & SIMD comparison instructions, and some fixes
* Update FCVT (Scalar) encoding on the table to support the Half-float variants
* Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes
* Use old memory access methods, made a start on SIMD memory insts support, some fixes
* Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes
* Fix arguments count with struct return values, other fixes
* More instructions
* Misc. fixes and integrate LDj3SNuD fixes
* Update tests
* Add a faster linear scan allocator, unwinding support on windows, and other changes
* Update Ryujinx.HLE
* Update Ryujinx.Graphics
* Fix V128 return pointer passing, RCX is clobbered
* Update Ryujinx.Tests
* Update ITimeZoneService
* Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks
* Use generic GetFunctionPointerForDelegate method and other tweaks
* Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics
* Remove some unused code on the assembler
* Fix REX.W prefix regression on float conversion instructions, add some sort of profiler
* Add hardware capability detection
* Fix regression on Sha1h and revert Fcm** changes
* Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator
* Fix silly mistake introduced on last commit on CpuId
* Generate inline stack probes when the stack allocation is too large
* Initial support for the System-V ABI
* Support multiple destination operands
* Fix SSE2 VectorInsert8 path, and other fixes
* Change placement of XMM callee save and restore code to match other compilers
* Rename Dest to Destination and Inst to Instruction
* Fix a regression related to calls and the V128 type
* Add an extra space on comments to match code style
* Some refactoring
* Fix vector insert FP32 SSE2 path
* Port over the ARM32 instructions
* Avoid memory protection races on JIT Cache
* Another fix on VectorInsert FP32 (thanks to LDj3SNuD
* Float operands don't need to use the same register when VEX is supported
* Add a new register allocator, higher quality code for hot code (tier up), and other tweaks
* Some nits, small improvements on the pre allocator
* CpuThreadState is gone
* Allow changing CPU emulators with a config entry
* Add runtime identifiers on the ARMeilleure project
* Allow switching between CPUs through a config entry (pt. 2)
* Change win10-x64 to win-x64 on projects
* Update the Ryujinx project to use ARMeilleure
* Ensure that the selected register is valid on the hybrid allocator
* Allow exiting on returns to 0 (should fix test regression)
* Remove register assignments for most used variables on the hybrid allocator
* Do not use fixed registers as spill temp
* Add missing namespace and remove unneeded using
* Address PR feedback
* Fix types, etc
* Enable AssumeStrictAbiCompliance by default
* Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 14:56:22 -04:00
|
|
|
V128 v0 = MakeVectorE0E1(z, z);
|
|
|
|
V128 v1 = MakeVectorE0E1(a, a);
|
2018-07-18 20:06:28 -04:00
|
|
|
|
2018-11-01 00:22:09 -04:00
|
|
|
SingleOpcode(opcode, v0: v0, v1: v1);
|
2018-07-18 20:06:28 -04:00
|
|
|
|
2018-09-01 10:24:05 -04:00
|
|
|
CompareAgainstUnicorn();
|
2018-07-18 20:06:28 -04:00
|
|
|
}
|
2018-04-20 11:40:15 -04:00
|
|
|
#endif
|
|
|
|
}
|
|
|
|
}
|