RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2024-12-04 10:04:33 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Implement four-wide shuffle with 2 shufps if no more than two elements come

Browse Source 
from each vector. e.g.
        shuffle(G1, G2, 7, 1, 5, 2)
==>
        movaps _G2, %xmm0
        shufps $151, _G1, %xmm0
        shufps $216, %xmm0, %xmm0


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28011 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Evan Cheng 2006-04-28 07:03:38 +00:00
parent a0de843535
commit 43f3bd310b
1 changed files with 47 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

49
lib/Target/X86/X86ISelLowering.cpp
View File

@ -2676,10 +2676,55 @@ X86TargetLowering::LowerVECTOR_SHUFFLE(SDOperand Op, SelectionDAG &DAG) {
} }
if (NumElems == 4) { if (NumElems == 4) {
// Break it into (shuffle shuffle_hi, shuffle_lo).
MVT::ValueType MaskVT = PermMask.getValueType(); MVT::ValueType MaskVT = PermMask.getValueType();
MVT::ValueType MaskEVT = MVT::getVectorBaseType(MaskVT); MVT::ValueType MaskEVT = MVT::getVectorBaseType(MaskVT);
std::map<unsigned, std::pair<int, int> > Locs; std::vector<std::pair<int, int> > Locs;
Locs.reserve(NumElems);
std::vector<SDOperand> Mask1(NumElems, DAG.getNode(ISD::UNDEF, MaskEVT));
std::vector<SDOperand> Mask2(NumElems, DAG.getNode(ISD::UNDEF, MaskEVT));
unsigned NumHi = 0;
unsigned NumLo = 0;
// If no more than two elements come from either vector. This can be
// implemented with two shuffles. First shuffle gather the elements.
// The second shuffle, which takes the first shuffle as both of its
// vector operands, put the elements into the right order.
for (unsigned i = 0; i != NumElems; ++i) {
SDOperand Elt = PermMask.getOperand(i);
if (Elt.getOpcode() == ISD::UNDEF) {
Locs[i] = std::make_pair(-1, -1);
} else {
unsigned Val = cast<ConstantSDNode>(Elt)->getValue();
if (Val < NumElems) {
Locs[i] = std::make_pair(0, NumLo);
Mask1[NumLo] = Elt;
NumLo++;
} else {
Locs[i] = std::make_pair(1, NumHi);
if (2+NumHi < NumElems)
Mask1[2+NumHi] = Elt;
NumHi++;
}
}
}
if (NumLo <= 2 && NumHi <= 2) {
V1 = DAG.getNode(ISD::VECTOR_SHUFFLE, VT, V1, V2,
DAG.getNode(ISD::BUILD_VECTOR, MaskVT, Mask1));
for (unsigned i = 0; i != NumElems; ++i) {
if (Locs[i].first == -1)
continue;
else {
unsigned Idx = (i < NumElems/2) ? 0 : NumElems;
Idx += Locs[i].first * (NumElems/2) + Locs[i].second;
Mask2[i] = DAG.getConstant(Idx, MaskEVT);
}
}
return DAG.getNode(ISD::VECTOR_SHUFFLE, VT, V1, V1,
DAG.getNode(ISD::BUILD_VECTOR, MaskVT, Mask2));
}
// Break it into (shuffle shuffle_hi, shuffle_lo).
Locs.clear();
std::vector<SDOperand> LoMask(NumElems, DAG.getNode(ISD::UNDEF, MaskEVT)); std::vector<SDOperand> LoMask(NumElems, DAG.getNode(ISD::UNDEF, MaskEVT));
std::vector<SDOperand> HiMask(NumElems, DAG.getNode(ISD::UNDEF, MaskEVT)); std::vector<SDOperand> HiMask(NumElems, DAG.getNode(ISD::UNDEF, MaskEVT));
std::vector<SDOperand> *MaskPtr = &LoMask; std::vector<SDOperand> *MaskPtr = &LoMask;