FEX-Emu/FEX
1
0
Fork 0
mirror of https://github.com/FEX-Emu/FEX.git synced 2025-01-05 21:09:56 +00:00
Code Issues Actions 6 Packages Projects Releases Wiki Activity

IR: add UMaxV

Browse Source 
will be used to accelerate ptest

Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
This commit is contained in:
Alyssa Rosenzweig 2024-01-22 18:28:31 -04:00
parent 8e3d4a3e02
commit 8e7906a665
2 changed files with 33 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
FEXCore/Source/Interface/Core/JIT/Arm64/VectorOps.cpp
View File

@ -1372,6 +1372,32 @@ DEF_OP(VUMinV) {
} }
} }
DEF_OP(VUMaxV) {
const auto Op = IROp->C<IR::IROp_VUMaxV>();
const auto OpSize = IROp->Size;
const auto Is256Bit = OpSize == Core::CPUState::XMM_AVX_REG_SIZE;
const auto ElementSize = Op->Header.ElementSize;
const auto Dst = GetVReg(Node);
const auto Vector = GetVReg(Op->Vector.ID());
LOGMAN_THROW_AA_FMT(ElementSize == 1 || ElementSize == 2 || ElementSize == 4 || ElementSize == 8, "Invalid size");
const auto SubRegSize =
ElementSize == 1 ? ARMEmitter::SubRegSize::i8Bit :
ElementSize == 2 ? ARMEmitter::SubRegSize::i16Bit :
ElementSize == 4 ? ARMEmitter::SubRegSize::i32Bit :
ElementSize == 8 ? ARMEmitter::SubRegSize::i64Bit : ARMEmitter::SubRegSize::i8Bit;
if (HostSupportsSVE256 && Is256Bit) {
const auto Pred = PRED_TMP_32B;
umaxv(SubRegSize, Dst, Pred, Vector.Z());
} else {
// Vector
umaxv(SubRegSize, Dst.Q(), Vector.Q());
}
}
DEF_OP(VURAvg) { DEF_OP(VURAvg) {
const auto Op = IROp->C<IR::IROp_VURAvg>(); const auto Op = IROp->C<IR::IROp_VURAvg>();
const auto OpSize = IROp->Size; const auto OpSize = IROp->Size;

8
FEXCore/Source/Interface/IR/IR.json
View File

@ -1612,7 +1612,13 @@
"DestSize": "RegisterSize", "DestSize": "RegisterSize",
"NumElements": "RegisterSize / ElementSize" "NumElements": "RegisterSize / ElementSize"
}, },
"FPR = VUMaxV u8:#RegisterSize, u8:#ElementSize, FPR:$Vector": {
"Desc": ["Does a horizontal vector unsigned maximum of elements across the source vector",
"Result is a zero extended scalar"
],
"DestSize": "RegisterSize",
"NumElements": "RegisterSize / ElementSize"
},
"FPR = VFAbs u8:#RegisterSize, u8:#ElementSize, FPR:$Vector": { "FPR = VFAbs u8:#RegisterSize, u8:#ElementSize, FPR:$Vector": {
"DestSize": "RegisterSize", "DestSize": "RegisterSize",
"NumElements": "RegisterSize / ElementSize" "NumElements": "RegisterSize / ElementSize"