Upgrade calls to CreateShuffleVector to use the preferred form of passing an array of ints
No functionality change intended.
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 6e3a3df..87f5210 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -5711,7 +5711,7 @@
case NEON::BI__builtin_neon_vext_v:
case NEON::BI__builtin_neon_vextq_v: {
int CV = cast<ConstantInt>(Ops[2])->getSExtValue();
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i)
Indices.push_back(i+CV);
@@ -5983,7 +5983,7 @@
Value *SV = nullptr;
for (unsigned vi = 0; vi != 2; ++vi) {
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
Indices.push_back(i+vi);
Indices.push_back(i+e+vi);
@@ -6011,7 +6011,7 @@
Value *SV = nullptr;
for (unsigned vi = 0; vi != 2; ++vi) {
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i)
Indices.push_back(2*i+vi);
@@ -6029,7 +6029,7 @@
Value *SV = nullptr;
for (unsigned vi = 0; vi != 2; ++vi) {
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
Indices.push_back((i + vi*e) >> 1);
Indices.push_back(((i + vi*e) >> 1)+e);
@@ -6120,7 +6120,7 @@
TblOps.push_back(ExtOp);
// Build a vector containing sequential number like (0, 1, 2, ..., 15)
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
llvm::VectorType *TblTy = cast<llvm::VectorType>(Ops[0]->getType());
for (unsigned i = 0, e = TblTy->getNumElements(); i != e; ++i) {
Indices.push_back(2*i);
@@ -6957,7 +6957,7 @@
if (VTy->getElementType()->isIntegerTy(64)) {
// Extract the other lane.
Ops[1] = Builder.CreateBitCast(Ops[1], Ty);
- uint32_t Lane = cast<ConstantInt>(Ops[2])->getZExtValue();
+ int Lane = cast<ConstantInt>(Ops[2])->getZExtValue();
Value *SV = llvm::ConstantVector::get(ConstantInt::get(Int32Ty, 1-Lane));
Ops[1] = Builder.CreateShuffleVector(Ops[1], Ops[1], SV);
// Load the value as a one-element vector.
@@ -6967,9 +6967,8 @@
Value *Align = getAlignmentValue32(PtrOp0);
Value *Ld = Builder.CreateCall(F, {Ops[0], Align});
// Combine them.
- uint32_t Indices[] = {1 - Lane, Lane};
- SV = llvm::ConstantDataVector::get(getLLVMContext(), Indices);
- return Builder.CreateShuffleVector(Ops[1], Ld, SV, "vld1q_lane");
+ int Indices[] = {1 - Lane, Lane};
+ return Builder.CreateShuffleVector(Ops[1], Ld, Indices, "vld1q_lane");
}
LLVM_FALLTHROUGH;
case NEON::BI__builtin_neon_vld1_lane_v: {
@@ -7144,7 +7143,7 @@
static llvm::Value *VectorUnzip(CGBuilderTy &Builder, llvm::Value *V, bool Odd) {
// Make a shufflevector that extracts every other element of a vector (evens
// or odds, as desired).
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
unsigned InputElements =
cast<llvm::VectorType>(V->getType())->getNumElements();
for (unsigned i = 0; i < InputElements; i += 2)
@@ -7157,7 +7156,7 @@
llvm::Value *V1) {
// Make a shufflevector that interleaves two vectors element by element.
assert(V0->getType() == V1->getType() && "Can't zip different vector types");
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
unsigned InputElements =
cast<llvm::VectorType>(V0->getType())->getNumElements();
for (unsigned i = 0; i < InputElements; i++) {
@@ -7185,7 +7184,7 @@
unsigned ReverseWidth) {
// MVE-specific helper function which reverses the elements of a
// vector within every (ReverseWidth)-bit collection of lanes.
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
unsigned LaneSize = V->getType()->getScalarSizeInBits();
unsigned Elements = 128 / LaneSize;
unsigned Mask = ReverseWidth / LaneSize - 1;
@@ -9971,7 +9970,7 @@
Value *SV = nullptr;
for (unsigned vi = 0; vi != 2; ++vi) {
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
Indices.push_back(i+vi);
Indices.push_back(i+e+vi);
@@ -9990,7 +9989,7 @@
Value *SV = nullptr;
for (unsigned vi = 0; vi != 2; ++vi) {
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; ++i)
Indices.push_back(2*i+vi);
@@ -10008,7 +10007,7 @@
Value *SV = nullptr;
for (unsigned vi = 0; vi != 2; ++vi) {
- SmallVector<uint32_t, 16> Indices;
+ SmallVector<int, 16> Indices;
for (unsigned i = 0, e = VTy->getNumElements(); i != e; i += 2) {
Indices.push_back((i + vi*e) >> 1);
Indices.push_back(((i + vi*e) >> 1)+e);
@@ -10132,7 +10131,7 @@
// If we have less than 8 elements, then the starting mask was an i8 and
// we need to extract down to the right number of elements.
if (NumElts < 8) {
- uint32_t Indices[4];
+ int Indices[4];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i;
MaskVec = CGF.Builder.CreateShuffleVector(MaskVec, MaskVec,
@@ -10321,7 +10320,7 @@
}
if (NumElts < 8) {
- uint32_t Indices[8];
+ int Indices[8];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i;
for (unsigned i = NumElts; i != 8; ++i)
@@ -10661,9 +10660,8 @@
// Extract the subvector.
if (NumDstElts != cast<llvm::VectorType>(Src->getType())->getNumElements()) {
assert(NumDstElts == 4 && "Unexpected vector size");
- uint32_t ShuffleMask[4] = {0, 1, 2, 3};
Src = CGF.Builder.CreateShuffleVector(Src, UndefValue::get(Src->getType()),
- ShuffleMask);
+ ArrayRef<int>{0, 1, 2, 3});
}
// Bitcast from vXi16 to vXf16.
@@ -11545,7 +11543,7 @@
Index &= SubVectors - 1; // Remove any extra bits.
Index *= NumElts;
- uint32_t Indices[16];
+ int Indices[16];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i + Index;
@@ -11585,7 +11583,7 @@
Index &= SubVectors - 1; // Remove any extra bits.
Index *= SrcNumElts;
- uint32_t Indices[16];
+ int Indices[16];
for (unsigned i = 0; i != DstNumElts; ++i)
Indices[i] = (i >= SrcNumElts) ? SrcNumElts + (i % SrcNumElts) : i;
@@ -11646,7 +11644,7 @@
cast<llvm::VectorType>(Ops[0]->getType())->getNumElements();
unsigned Imm = cast<llvm::ConstantInt>(Ops[2])->getZExtValue();
- uint32_t Indices[16];
+ int Indices[16];
// If there are more than 8 elements, the immediate is used twice so make
// sure we handle that.
for (unsigned i = 0; i != NumElts; ++i)
@@ -11666,7 +11664,7 @@
// Splat the 8-bits of immediate 4 times to help the loop wrap around.
Imm = (Imm & 0xff) * 0x01010101;
- uint32_t Indices[32];
+ int Indices[32];
for (unsigned l = 0; l != NumElts; l += 8) {
for (unsigned i = 0; i != 4; ++i) {
Indices[l + i] = l + (Imm & 3);
@@ -11690,7 +11688,7 @@
// Splat the 8-bits of immediate 4 times to help the loop wrap around.
Imm = (Imm & 0xff) * 0x01010101;
- uint32_t Indices[32];
+ int Indices[32];
for (unsigned l = 0; l != NumElts; l += 8) {
for (unsigned i = 0; i != 4; ++i)
Indices[l + i] = l + i;
@@ -11722,7 +11720,7 @@
// Splat the 8-bits of immediate 4 times to help the loop wrap around.
Imm = (Imm & 0xff) * 0x01010101;
- uint32_t Indices[16];
+ int Indices[16];
for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
for (unsigned i = 0; i != NumLaneElts; ++i) {
Indices[i + l] = (Imm % NumLaneElts) + l;
@@ -11773,7 +11771,7 @@
unsigned NumElts = Ty->getNumElements();
// These intrinsics operate on 256-bit lanes of four 64-bit elements.
- uint32_t Indices[8];
+ int Indices[8];
for (unsigned l = 0; l != NumElts; l += 4)
for (unsigned i = 0; i != 4; ++i)
Indices[l + i] = l + ((Imm >> (2 * i)) & 0x3);
@@ -11804,7 +11802,7 @@
Ops[0] = llvm::Constant::getNullValue(Ops[0]->getType());
}
- uint32_t Indices[64];
+ int Indices[64];
// 256-bit palignr operates on 128-bit lanes so we need to handle that
for (unsigned l = 0; l != NumElts; l += 16) {
for (unsigned i = 0; i != 16; ++i) {
@@ -11832,7 +11830,7 @@
// Mask the shift amount to width of two vectors.
ShiftVal &= (2 * NumElts) - 1;
- uint32_t Indices[16];
+ int Indices[16];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i + ShiftVal;
@@ -11854,7 +11852,7 @@
unsigned NumLanes = Ty->getPrimitiveSizeInBits() == 512 ? 4 : 2;
unsigned NumLaneElts = NumElts / NumLanes;
- uint32_t Indices[16];
+ int Indices[16];
for (unsigned l = 0; l != NumElts; l += NumLaneElts) {
unsigned Index = (Imm % NumLanes) * NumLaneElts;
Imm /= NumLanes; // Discard the bits we just used.
@@ -11884,7 +11882,7 @@
// duplicate sources, but this can be dealt with in the backend.
Value *OutOps[2];
- uint32_t Indices[8];
+ int Indices[8];
for (unsigned l = 0; l != 2; ++l) {
// Determine the source for this lane.
if (Imm & (1 << ((l * 4) + 3)))
@@ -11922,7 +11920,7 @@
if (ShiftVal >= 16)
return llvm::Constant::getNullValue(ResultType);
- uint32_t Indices[64];
+ int Indices[64];
// 256/512-bit pslldq operates on 128-bit lanes so we need to handle that
for (unsigned l = 0; l != NumElts; l += 16) {
for (unsigned i = 0; i != 16; ++i) {
@@ -11952,7 +11950,7 @@
if (ShiftVal >= 16)
return llvm::Constant::getNullValue(ResultType);
- uint32_t Indices[64];
+ int Indices[64];
// 256/512-bit psrldq operates on 128-bit lanes so we need to handle that
for (unsigned l = 0; l != NumElts; l += 16) {
for (unsigned i = 0; i != 16; ++i) {
@@ -11982,7 +11980,7 @@
Value *In = getMaskVecValue(*this, Ops[0], NumElts);
- uint32_t Indices[64];
+ int Indices[64];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = NumElts + i - ShiftVal;
@@ -12004,7 +12002,7 @@
Value *In = getMaskVecValue(*this, Ops[0], NumElts);
- uint32_t Indices[64];
+ int Indices[64];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i + ShiftVal;
@@ -12284,7 +12282,7 @@
unsigned NumElts = Ops[0]->getType()->getIntegerBitWidth();
Value *LHS = getMaskVecValue(*this, Ops[0], NumElts);
Value *RHS = getMaskVecValue(*this, Ops[1], NumElts);
- uint32_t Indices[64];
+ int Indices[64];
for (unsigned i = 0; i != NumElts; ++i)
Indices[i] = i;
@@ -13443,15 +13441,9 @@
Ops[1] = Builder.CreateBitCast(Ops[1], llvm::VectorType::get(Int64Ty, 2));
if (getTarget().isLittleEndian()) {
- // Create a shuffle mask of (1, 0)
- Constant *ShuffleElts[2] = { ConstantInt::get(Int32Ty, 1),
- ConstantInt::get(Int32Ty, 0)
- };
- Constant *ShuffleMask = llvm::ConstantVector::get(ShuffleElts);
-
// Reverse the double words in the vector we will extract from.
Ops[0] = Builder.CreateBitCast(Ops[0], llvm::VectorType::get(Int64Ty, 2));
- Ops[0] = Builder.CreateShuffleVector(Ops[0], Ops[0], ShuffleMask);
+ Ops[0] = Builder.CreateShuffleVector(Ops[0], Ops[0], ArrayRef<int>{1, 0});
// Reverse the index.
Index = MaxIndex - Index;
@@ -13485,13 +13477,8 @@
// Emit the call, then reverse the double words of the results vector.
Value *Call = Builder.CreateCall(F, Ops);
- // Create a shuffle mask of (1, 0)
- Constant *ShuffleElts[2] = { ConstantInt::get(Int32Ty, 1),
- ConstantInt::get(Int32Ty, 0)
- };
- Constant *ShuffleMask = llvm::ConstantVector::get(ShuffleElts);
-
- Value *ShuffleCall = Builder.CreateShuffleVector(Call, Call, ShuffleMask);
+ Value *ShuffleCall =
+ Builder.CreateShuffleVector(Call, Call, ArrayRef<int>{1, 0});
return ShuffleCall;
} else {
Ops[1] = ConstantInt::getSigned(Int32Ty, Index);
@@ -13510,15 +13497,12 @@
// Account for endianness by treating this as just a shuffle. So we use the
// same indices for both LE and BE in order to produce expected results in
// both cases.
- unsigned ElemIdx0 = (Index & 2) >> 1;
- unsigned ElemIdx1 = 2 + (Index & 1);
+ int ElemIdx0 = (Index & 2) >> 1;
+ int ElemIdx1 = 2 + (Index & 1);
- Constant *ShuffleElts[2] = {ConstantInt::get(Int32Ty, ElemIdx0),
- ConstantInt::get(Int32Ty, ElemIdx1)};
- Constant *ShuffleMask = llvm::ConstantVector::get(ShuffleElts);
-
+ int ShuffleElts[2] = {ElemIdx0, ElemIdx1};
Value *ShuffleCall =
- Builder.CreateShuffleVector(Ops[0], Ops[1], ShuffleMask);
+ Builder.CreateShuffleVector(Ops[0], Ops[1], ShuffleElts);
QualType BIRetType = E->getType();
auto RetTy = ConvertType(BIRetType);
return Builder.CreateBitCast(ShuffleCall, RetTy);
@@ -13532,10 +13516,10 @@
Ops[1] = Builder.CreateBitCast(Ops[1], llvm::VectorType::get(Int32Ty, 4));
// Create a shuffle mask
- unsigned ElemIdx0;
- unsigned ElemIdx1;
- unsigned ElemIdx2;
- unsigned ElemIdx3;
+ int ElemIdx0;
+ int ElemIdx1;
+ int ElemIdx2;
+ int ElemIdx3;
if (getTarget().isLittleEndian()) {
// Little endian element N comes from element 8+N-Index of the
// concatenated wide vector (of course, using modulo arithmetic on
@@ -13552,14 +13536,9 @@
ElemIdx3 = Index + 3;
}
- Constant *ShuffleElts[4] = {ConstantInt::get(Int32Ty, ElemIdx0),
- ConstantInt::get(Int32Ty, ElemIdx1),
- ConstantInt::get(Int32Ty, ElemIdx2),
- ConstantInt::get(Int32Ty, ElemIdx3)};
-
- Constant *ShuffleMask = llvm::ConstantVector::get(ShuffleElts);
+ int ShuffleElts[4] = {ElemIdx0, ElemIdx1, ElemIdx2, ElemIdx3};
Value *ShuffleCall =
- Builder.CreateShuffleVector(Ops[0], Ops[1], ShuffleMask);
+ Builder.CreateShuffleVector(Ops[0], Ops[1], ShuffleElts);
QualType BIRetType = E->getType();
auto RetTy = ConvertType(BIRetType);
return Builder.CreateBitCast(ShuffleCall, RetTy);
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index cd95e78..bafe68d2 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -1745,12 +1745,9 @@
// Handle vec3 special.
if (VecTy && VecTy->getNumElements() == 3) {
// Our source is a vec3, do a shuffle vector to make it a vec4.
- llvm::Constant *Mask[] = {Builder.getInt32(0), Builder.getInt32(1),
- Builder.getInt32(2),
- llvm::UndefValue::get(Builder.getInt32Ty())};
- llvm::Value *MaskV = llvm::ConstantVector::get(Mask);
Value = Builder.CreateShuffleVector(Value, llvm::UndefValue::get(VecTy),
- MaskV, "extractVec");
+ ArrayRef<int>{0, 1, 2, -1},
+ "extractVec");
SrcTy = llvm::VectorType::get(VecTy->getElementType(), 4);
}
if (Addr.getElementType() != SrcTy) {
@@ -1886,13 +1883,12 @@
// Always use shuffle vector to try to retain the original program structure
unsigned NumResultElts = ExprVT->getNumElements();
- SmallVector<llvm::Constant*, 4> Mask;
+ SmallVector<int, 4> Mask;
for (unsigned i = 0; i != NumResultElts; ++i)
- Mask.push_back(Builder.getInt32(getAccessedFieldNo(i, Elts)));
+ Mask.push_back(getAccessedFieldNo(i, Elts));
- llvm::Value *MaskV = llvm::ConstantVector::get(Mask);
Vec = Builder.CreateShuffleVector(Vec, llvm::UndefValue::get(Vec->getType()),
- MaskV);
+ Mask);
return RValue::get(Vec);
}
@@ -2133,32 +2129,27 @@
// Use shuffle vector is the src and destination are the same number of
// elements and restore the vector mask since it is on the side it will be
// stored.
- SmallVector<llvm::Constant*, 4> Mask(NumDstElts);
+ SmallVector<int, 4> Mask(NumDstElts);
for (unsigned i = 0; i != NumSrcElts; ++i)
- Mask[getAccessedFieldNo(i, Elts)] = Builder.getInt32(i);
+ Mask[getAccessedFieldNo(i, Elts)] = i;
- llvm::Value *MaskV = llvm::ConstantVector::get(Mask);
- Vec = Builder.CreateShuffleVector(SrcVal,
- llvm::UndefValue::get(Vec->getType()),
- MaskV);
+ Vec = Builder.CreateShuffleVector(
+ SrcVal, llvm::UndefValue::get(Vec->getType()), Mask);
} else if (NumDstElts > NumSrcElts) {
// Extended the source vector to the same length and then shuffle it
// into the destination.
// FIXME: since we're shuffling with undef, can we just use the indices
// into that? This could be simpler.
- SmallVector<llvm::Constant*, 4> ExtMask;
+ SmallVector<int, 4> ExtMask;
for (unsigned i = 0; i != NumSrcElts; ++i)
- ExtMask.push_back(Builder.getInt32(i));
- ExtMask.resize(NumDstElts, llvm::UndefValue::get(Int32Ty));
- llvm::Value *ExtMaskV = llvm::ConstantVector::get(ExtMask);
- llvm::Value *ExtSrcVal =
- Builder.CreateShuffleVector(SrcVal,
- llvm::UndefValue::get(SrcVal->getType()),
- ExtMaskV);
+ ExtMask.push_back(i);
+ ExtMask.resize(NumDstElts, -1);
+ llvm::Value *ExtSrcVal = Builder.CreateShuffleVector(
+ SrcVal, llvm::UndefValue::get(SrcVal->getType()), ExtMask);
// build identity
- SmallVector<llvm::Constant*, 4> Mask;
+ SmallVector<int, 4> Mask;
for (unsigned i = 0; i != NumDstElts; ++i)
- Mask.push_back(Builder.getInt32(i));
+ Mask.push_back(i);
// When the vector size is odd and .odd or .hi is used, the last element
// of the Elts constant array will be one past the size of the vector.
@@ -2168,9 +2159,8 @@
// modify when what gets shuffled in
for (unsigned i = 0; i != NumSrcElts; ++i)
- Mask[getAccessedFieldNo(i, Elts)] = Builder.getInt32(i+NumDstElts);
- llvm::Value *MaskV = llvm::ConstantVector::get(Mask);
- Vec = Builder.CreateShuffleVector(Vec, ExtSrcVal, MaskV);
+ Mask[getAccessedFieldNo(i, Elts)] = i + NumDstElts;
+ Vec = Builder.CreateShuffleVector(Vec, ExtSrcVal, Mask);
} else {
// We should never shorten the vector
llvm_unreachable("unexpected shorten vector length");
diff --git a/clang/lib/CodeGen/CGExprScalar.cpp b/clang/lib/CodeGen/CGExprScalar.cpp
index f3f826e..c4f1faa 100644
--- a/clang/lib/CodeGen/CGExprScalar.cpp
+++ b/clang/lib/CodeGen/CGExprScalar.cpp
@@ -1650,18 +1650,17 @@
Value* V1 = CGF.EmitScalarExpr(E->getExpr(0));
Value* V2 = CGF.EmitScalarExpr(E->getExpr(1));
- SmallVector<llvm::Constant*, 32> indices;
+ SmallVector<int, 32> Indices;
for (unsigned i = 2; i < E->getNumSubExprs(); ++i) {
llvm::APSInt Idx = E->getShuffleMaskIdx(CGF.getContext(), i-2);
// Check for -1 and output it as undef in the IR.
if (Idx.isSigned() && Idx.isAllOnesValue())
- indices.push_back(llvm::UndefValue::get(CGF.Int32Ty));
+ Indices.push_back(-1);
else
- indices.push_back(Builder.getInt32(Idx.getZExtValue()));
+ Indices.push_back(Idx.getZExtValue());
}
- Value *SV = llvm::ConstantVector::get(indices);
- return Builder.CreateShuffleVector(V1, V2, SV, "shuffle");
+ return Builder.CreateShuffleVector(V1, V2, Indices, "shuffle");
}
Value *ScalarExprEmitter::VisitConvertVectorExpr(ConvertVectorExpr *E) {
@@ -4532,14 +4531,9 @@
static Value *ConvertVec3AndVec4(CGBuilderTy &Builder, CodeGenFunction &CGF,
Value *Src, unsigned NumElementsDst) {
llvm::Value *UnV = llvm::UndefValue::get(Src->getType());
- SmallVector<llvm::Constant*, 4> Args;
- Args.push_back(Builder.getInt32(0));
- Args.push_back(Builder.getInt32(1));
- Args.push_back(Builder.getInt32(2));
- if (NumElementsDst == 4)
- Args.push_back(llvm::UndefValue::get(CGF.Int32Ty));
- llvm::Constant *Mask = llvm::ConstantVector::get(Args);
- return Builder.CreateShuffleVector(Src, UnV, Mask);
+ static constexpr int Mask[] = {0, 1, 2, -1};
+ return Builder.CreateShuffleVector(Src, UnV,
+ llvm::makeArrayRef(Mask, NumElementsDst));
}
// Create cast instructions for converting LLVM value \p Src to LLVM type \p