FEX-Emu/FEX
1
0
Fork 0
mirror of https://github.com/FEX-Emu/FEX.git synced 2024-12-13 17:15:41 +00:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

OpcodeDispatcher: extract VPERMDIndices

Browse Source 
and rename things accordingly.

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
This commit is contained in:
Alyssa Rosenzweig 2024-06-24 18:15:06 -04:00
parent 2580efda0d
commit 6009f36403
2 changed files with 17 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

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

@ -663,6 +663,8 @@ public:
void VPCMPISTRIOp(OpcodeArgs);
void VPCMPISTRMOp(OpcodeArgs);
Ref VPERMDIndices(OpSize DstSize, Ref Indices, Ref IndexMask, Ref Repeating3210);
void VPERM2Op(OpcodeArgs);
void VPERMDOp(OpcodeArgs);
void VPERMQOp(OpcodeArgs);

24
FEXCore/Source/Interface/Core/OpcodeDispatcher/Vector.cpp
View File

@ -4504,14 +4504,8 @@ void OpDispatchBuilder::VPERM2Op(OpcodeArgs) {
StoreResult(FPRClass, Op, Result, -1);
}
void OpDispatchBuilder::VPERMDOp(OpcodeArgs) {
const auto DstSize = GetDstSize(Op);
Ref Indices = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags);
Ref Src = LoadSource(FPRClass, Op, Op->Src[1], Op->Flags);
Ref OpDispatchBuilder::VPERMDIndices(OpSize DstSize, Ref Indices, Ref IndexMask, Ref Repeating3210) {
// Get rid of any junk unrelated to the relevant selector index bits (bits [2:0])
Ref IndexMask = _VectorImm(DstSize, 4, 0b111);
Ref SanitizedIndices = _VAnd(DstSize, 1, Indices, IndexMask);
// Build up the broadcasted index mask. e.g. On x86-64, the selector index
@ -4569,9 +4563,21 @@ void OpDispatchBuilder::VPERMDOp(OpcodeArgs) {
// ╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝╚════╝
//
// Which finally lets us permute the source vector and be done with everything.
return _VAdd(DstSize, 1, ShiftedIndices, Repeating3210);
}
void OpDispatchBuilder::VPERMDOp(OpcodeArgs) {
const auto DstSize = GetDstSize(Op);
Ref Indices = LoadSource(FPRClass, Op, Op->Src[0], Op->Flags);
Ref Src = LoadSource(FPRClass, Op, Op->Src[1], Op->Flags);
// Get rid of any junk unrelated to the relevant selector index bits (bits [2:0])
Ref IndexMask = _VectorImm(DstSize, 4, 0b111);
Ref AddConst = _Constant(0x03020100);
Ref AddVector = _VDupFromGPR(DstSize, 4, AddConst);
Ref FinalIndices = _VAdd(DstSize, 1, ShiftedIndices, AddVector);
Ref Repeating3210 = _VDupFromGPR(DstSize, 4, AddConst);
Ref FinalIndices = VPERMDIndices(OpSizeFromDst(Op), Indices, IndexMask, Repeating3210);
// Now lets finally shuffle this bad boy around.
Ref Result = _VTBL1(DstSize, Src, FinalIndices);