FEX-Emu/FEX
1
0
Fork 0
mirror of https://github.com/FEX-Emu/FEX.git synced 2024-12-14 09:28:34 +00:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

AVX128: Implement support for round{ss,sd}

Browse Source
This commit is contained in:
Ryan Houdek 2024-06-19 03:06:26 -07:00 committed by Alyssa Rosenzweig
parent 0595f1e044
commit 702ecf7637
2 changed files with 26 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
FEXCore/Source/Interface/Core/OpcodeDispatcher.h
View File

@ -1139,6 +1139,8 @@ public:
template<size_t ElementSize>
void AVX128_VectorRound(OpcodeArgs);
template<size_t ElementSize>
void AVX128_InsertScalarRound(OpcodeArgs);
// End of AVX 128-bit implementation

26
FEXCore/Source/Interface/Core/OpcodeDispatcher/AVX_128.cpp
View File

@ -341,8 +341,8 @@ void OpDispatchBuilder::InstallAVX128Handlers() {
// TODO: {OPD(3, 0b01, 0x06), 1, &OpDispatchBuilder::VPERM2Op},
{OPD(3, 0b01, 0x08), 1, &OpDispatchBuilder::AVX128_VectorRound<4>},
{OPD(3, 0b01, 0x09), 1, &OpDispatchBuilder::AVX128_VectorRound<8>},
// TODO: {OPD(3, 0b01, 0x0A), 1, &OpDispatchBuilder::AVXInsertScalarRound<4>},
// TODO: {OPD(3, 0b01, 0x0B), 1, &OpDispatchBuilder::AVXInsertScalarRound<8>},
{OPD(3, 0b01, 0x0A), 1, &OpDispatchBuilder::AVX128_InsertScalarRound<4>},
{OPD(3, 0b01, 0x0B), 1, &OpDispatchBuilder::AVX128_InsertScalarRound<8>},
// TODO: {OPD(3, 0b01, 0x0C), 1, &OpDispatchBuilder::VPBLENDDOp},
// TODO: {OPD(3, 0b01, 0x0D), 1, &OpDispatchBuilder::VBLENDPDOp},
// TODO: {OPD(3, 0b01, 0x0E), 1, &OpDispatchBuilder::VPBLENDWOp},
@ -1696,4 +1696,26 @@ void OpDispatchBuilder::AVX128_VectorRound(OpcodeArgs) {
[this, Mode](size_t, Ref Src) { return VectorRoundImpl(OpSize::i128Bit, ElementSize, Src, Mode); });
}
template<size_t ElementSize>
void OpDispatchBuilder::AVX128_InsertScalarRound(OpcodeArgs) {
// We load the full vector width when dealing with a source vector,
// so that we don't do any unnecessary zero extension to the scalar
// element that we're going to operate on.
const auto SrcSize = GetSrcSize(Op);
auto Src1 = AVX128_LoadSource_WithOpSize(Op, Op->Src[0], Op->Flags, false);
RefPair Src2 {};
if (Op->Src[1].IsGPR()) {
Src2 = AVX128_LoadSource_WithOpSize(Op, Op->Src[1], Op->Flags, false);
} else {
Src2.Low = LoadSource_WithOpSize(FPRClass, Op, Op->Src[1], SrcSize, Op->Flags);
}
// If OpSize == ElementSize then it only does the lower scalar op
const auto SourceMode = TranslateRoundType(Op->Src[2].Literal());
Ref Result = _VFToIScalarInsert(OpSize::i128Bit, ElementSize, Src1.Low, Src2.Low, SourceMode, false);
AVX128_StoreResult_WithOpSize(Op, Op->Dest, AVX128_Zext(Result));
}
} // namespace FEXCore::IR