RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2025-04-04 01:01:37 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

X86: Properly decode shuffle masks when the constant pool type is weird

Browse Source 
It's possible for the constant pool entry for the shuffle mask to come
from a completely different operation.  This occurs when Constants have
the same bit pattern but have different types.

Make DecodePSHUFBMask tolerant of types which, after a bitcast, are
appropriately sized vector types.

This fixes PR22188.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@225597 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
David Majnemer 2015-01-11 05:08:57 +00:00
parent 776673ea09
commit d2f4460ee7
5 changed files with 90 additions and 66 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

102
lib/Target/X86/Utils/X86ShuffleDecode.cpp
View File

@ -13,7 +13,10 @@
//===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===//
#include "X86ShuffleDecode.h" #include "X86ShuffleDecode.h"
#include "llvm/Analysis/ConstantFolding.h"
#include "llvm/IR/Constants.h" #include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/CodeGen/MachineValueType.h" #include "llvm/CodeGen/MachineValueType.h"
//===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===//
@ -253,57 +256,64 @@ void DecodeVPERM2X128Mask(MVT VT, unsigned Imm,
} }
} }
void DecodePSHUFBMask(const Constant *C, SmallVectorImpl<int> &ShuffleMask) { void DecodePSHUFBMask(const Constant *C, const DataLayout *TD,
SmallVectorImpl<int> &ShuffleMask) {
Type *MaskTy = C->getType(); Type *MaskTy = C->getType();
assert(MaskTy->isVectorTy() && "Expected a vector constant mask!"); // It is not an error for the PSHUFB mask to not be a vector of i8 because the
assert(MaskTy->getVectorElementType()->isIntegerTy(8) && // constant pool uniques constants by their bit representation.
"Expected i8 constant mask elements!"); // e.g. the following take up the same space in the constant pool:
int NumElements = MaskTy->getVectorNumElements(); // i128 -170141183420855150465331762880109871104
//
// <2 x i64> <i64 -9223372034707292160, i64 -9223372034707292160>
//
// <4 x i32> <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32
// -2147483648>
unsigned MaskTySize = MaskTy->getPrimitiveSizeInBits();
VectorType *DestTy = nullptr;
if (MaskTySize == 128)
DestTy = VectorType::get(Type::getInt8Ty(MaskTy->getContext()), 16);
else if (MaskTySize == 256)
DestTy = VectorType::get(Type::getInt8Ty(MaskTy->getContext()), 32);
// FIXME: Add support for AVX-512. // FIXME: Add support for AVX-512.
assert((NumElements == 16 || NumElements == 32) && if (!DestTy)
"Only 128-bit and 256-bit vectors supported!"); return;
if (DestTy != MaskTy) {
if (!CastInst::castIsValid(Instruction::BitCast, const_cast<Constant *>(C),
DestTy))
return;
C = ConstantFoldInstOperands(Instruction::BitCast, DestTy,
const_cast<Constant *>(C), TD);
MaskTy = DestTy;
}
int NumElements = MaskTy->getVectorNumElements();
ShuffleMask.reserve(NumElements); ShuffleMask.reserve(NumElements);
if (auto *CDS = dyn_cast<ConstantDataSequential>(C)) { for (int i = 0; i < NumElements; ++i) {
assert((unsigned)NumElements == CDS->getNumElements() && // For AVX vectors with 32 bytes the base of the shuffle is the 16-byte
"Constant mask has a different number of elements!"); // lane of the vector we're inside.
int Base = i < 16 ? 0 : 16;
for (int i = 0; i < NumElements; ++i) { Constant *COp = C->getAggregateElement(i);
// For AVX vectors with 32 bytes the base of the shuffle is the 16-byte if (!COp) {
// lane of the vector we're inside. ShuffleMask.clear();
int Base = i < 16 ? 0 : 16; return;
uint64_t Element = CDS->getElementAsInteger(i); } else if (isa<UndefValue>(COp)) {
// If the high bit (7) of the byte is set, the element is zeroed. ShuffleMask.push_back(SM_SentinelUndef);
if (Element & (1 << 7)) continue;
ShuffleMask.push_back(SM_SentinelZero);
else {
// Only the least significant 4 bits of the byte are used.
int Index = Base + (Element & 0xf);
ShuffleMask.push_back(Index);
}
} }
} else if (auto *CV = dyn_cast<ConstantVector>(C)) { uint64_t Element = cast<ConstantInt>(COp)->getZExtValue();
assert((unsigned)NumElements == CV->getNumOperands() && // If the high bit (7) of the byte is set, the element is zeroed.
"Constant mask has a different number of elements!"); if (Element & (1 << 7))
ShuffleMask.push_back(SM_SentinelZero);
for (int i = 0; i < NumElements; ++i) { else {
// For AVX vectors with 32 bytes the base of the shuffle is the 16-byte // Only the least significant 4 bits of the byte are used.
// lane of the vector we're inside. int Index = Base + (Element & 0xf);
int Base = i < 16 ? 0 : 16; ShuffleMask.push_back(Index);
Constant *COp = CV->getOperand(i);
if (isa<UndefValue>(COp)) {
ShuffleMask.push_back(SM_SentinelUndef);
continue;
}
uint64_t Element = cast<ConstantInt>(COp)->getZExtValue();
// If the high bit (7) of the byte is set, the element is zeroed.
if (Element & (1 << 7))
ShuffleMask.push_back(SM_SentinelZero);
else {
// Only the least significant 4 bits of the byte are used.
int Index = Base + (Element & 0xf);
ShuffleMask.push_back(Index);
}
} }
} }
} }

4
lib/Target/X86/Utils/X86ShuffleDecode.h
View File

@ -24,6 +24,7 @@
namespace llvm { namespace llvm {
class Constant; class Constant;
class DataLayout;
class MVT; class MVT;
enum { SM_SentinelUndef = -1, SM_SentinelZero = -2 }; enum { SM_SentinelUndef = -1, SM_SentinelZero = -2 };
@ -68,7 +69,8 @@ void DecodeUNPCKHMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask); void DecodeUNPCKLMask(MVT VT, SmallVectorImpl<int> &ShuffleMask);
/// \brief Decode a PSHUFB mask from an IR-level vector constant. /// \brief Decode a PSHUFB mask from an IR-level vector constant.
void DecodePSHUFBMask(const Constant *C, SmallVectorImpl<int> &ShuffleMask); void DecodePSHUFBMask(const Constant *C, const DataLayout *TD,
SmallVectorImpl<int> &ShuffleMask);
/// \brief Decode a PSHUFB mask from a raw array of constants such as from /// \brief Decode a PSHUFB mask from a raw array of constants such as from
/// BUILD_VECTOR. /// BUILD_VECTOR.

38
lib/Target/X86/X86ISelLowering.cpp
View File

@ -5365,7 +5365,7 @@ static SDValue getShuffleVectorZeroOrUndef(SDValue V2, unsigned Idx,
/// IsUnary to true if only uses one source. Note that this will set IsUnary for /// IsUnary to true if only uses one source. Note that this will set IsUnary for
/// shuffles which use a single input multiple times, and in those cases it will /// shuffles which use a single input multiple times, and in those cases it will
/// adjust the mask to only have indices within that single input. /// adjust the mask to only have indices within that single input.
static bool getTargetShuffleMask(SDNode *N, MVT VT, static bool getTargetShuffleMask(SDNode *N, MVT VT, const DataLayout *TD,
SmallVectorImpl<int> &Mask, bool &IsUnary) { SmallVectorImpl<int> &Mask, bool &IsUnary) {
unsigned NumElems = VT.getVectorNumElements(); unsigned NumElems = VT.getVectorNumElements();
SDValue ImmN; SDValue ImmN;
@ -5472,13 +5472,7 @@ static bool getTargetShuffleMask(SDNode *N, MVT VT,
return false; return false;
if (auto *C = dyn_cast<Constant>(MaskCP->getConstVal())) { if (auto *C = dyn_cast<Constant>(MaskCP->getConstVal())) {
// FIXME: Support AVX-512 here. DecodePSHUFBMask(C, TD, Mask);
Type *Ty = C->getType();
if (!Ty->isVectorTy() || (Ty->getVectorNumElements() != 16 &&
Ty->getVectorNumElements() != 32))
return false;
DecodePSHUFBMask(C, Mask);
break; break;
} }
@ -5541,6 +5535,7 @@ static SDValue getShuffleScalarElt(SDNode *N, unsigned Index, SelectionDAG &DAG,
SDValue V = SDValue(N, 0); SDValue V = SDValue(N, 0);
EVT VT = V.getValueType(); EVT VT = V.getValueType();
unsigned Opcode = V.getOpcode(); unsigned Opcode = V.getOpcode();
const DataLayout *TD = DAG.getSubtarget().getDataLayout();
// Recurse into ISD::VECTOR_SHUFFLE node to find scalars. // Recurse into ISD::VECTOR_SHUFFLE node to find scalars.
if (const ShuffleVectorSDNode *SV = dyn_cast<ShuffleVectorSDNode>(N)) { if (const ShuffleVectorSDNode *SV = dyn_cast<ShuffleVectorSDNode>(N)) {
@ -5562,7 +5557,7 @@ static SDValue getShuffleScalarElt(SDNode *N, unsigned Index, SelectionDAG &DAG,
SmallVector<int, 16> ShuffleMask; SmallVector<int, 16> ShuffleMask;
bool IsUnary; bool IsUnary;
if (!getTargetShuffleMask(N, ShufVT, ShuffleMask, IsUnary)) if (!getTargetShuffleMask(N, ShufVT, TD, ShuffleMask, IsUnary))
return SDValue(); return SDValue();
int Elt = ShuffleMask[Index]; int Elt = ShuffleMask[Index];
@ -22117,7 +22112,8 @@ static bool combineX86ShufflesRecursively(SDValue Op, SDValue Root,
return false; return false;
SmallVector<int, 16> OpMask; SmallVector<int, 16> OpMask;
bool IsUnary; bool IsUnary;
bool HaveMask = getTargetShuffleMask(Op.getNode(), VT, OpMask, IsUnary); bool HaveMask = getTargetShuffleMask(
Op.getNode(), VT, Subtarget->getDataLayout(), OpMask, IsUnary);
// We only can combine unary shuffles which we can decode the mask for. // We only can combine unary shuffles which we can decode the mask for.
if (!HaveMask || !IsUnary) if (!HaveMask || !IsUnary)
return false; return false;
@ -22208,10 +22204,12 @@ static bool combineX86ShufflesRecursively(SDValue Op, SDValue Root,
/// ///
/// This is a very minor wrapper around getTargetShuffleMask to easy forming v4 /// This is a very minor wrapper around getTargetShuffleMask to easy forming v4
/// PSHUF-style masks that can be reused with such instructions. /// PSHUF-style masks that can be reused with such instructions.
static SmallVector<int, 4> getPSHUFShuffleMask(SDValue N) { static SmallVector<int, 4> getPSHUFShuffleMask(SDValue N,
const DataLayout *TD) {
SmallVector<int, 4> Mask; SmallVector<int, 4> Mask;
bool IsUnary; bool IsUnary;
bool HaveMask = getTargetShuffleMask(N.getNode(), N.getSimpleValueType(), Mask, IsUnary); bool HaveMask = getTargetShuffleMask(N.getNode(), N.getSimpleValueType(), TD,
Mask, IsUnary);
(void)HaveMask; (void)HaveMask;
assert(HaveMask); assert(HaveMask);
@ -22243,6 +22241,7 @@ combineRedundantDWordShuffle(SDValue N, MutableArrayRef<int> Mask,
assert(N.getOpcode() == X86ISD::PSHUFD && assert(N.getOpcode() == X86ISD::PSHUFD &&
"Called with something other than an x86 128-bit half shuffle!"); "Called with something other than an x86 128-bit half shuffle!");
SDLoc DL(N); SDLoc DL(N);
const DataLayout *TD = DAG.getSubtarget().getDataLayout();
// Walk up a single-use chain looking for a combinable shuffle. Keep a stack // Walk up a single-use chain looking for a combinable shuffle. Keep a stack
// of the shuffles in the chain so that we can form a fresh chain to replace // of the shuffles in the chain so that we can form a fresh chain to replace
@ -22328,7 +22327,7 @@ combineRedundantDWordShuffle(SDValue N, MutableArrayRef<int> Mask,
return SDValue(); return SDValue();
// Merge this node's mask and our incoming mask. // Merge this node's mask and our incoming mask.
SmallVector<int, 4> VMask = getPSHUFShuffleMask(V); SmallVector<int, 4> VMask = getPSHUFShuffleMask(V, TD);
for (int &M : Mask) for (int &M : Mask)
M = VMask[M]; M = VMask[M];
V = DAG.getNode(V.getOpcode(), DL, V.getValueType(), V.getOperand(0), V = DAG.getNode(V.getOpcode(), DL, V.getValueType(), V.getOperand(0),
@ -22377,6 +22376,7 @@ static bool combineRedundantHalfShuffle(SDValue N, MutableArrayRef<int> Mask,
"Called with something other than an x86 128-bit half shuffle!"); "Called with something other than an x86 128-bit half shuffle!");
SDLoc DL(N); SDLoc DL(N);
unsigned CombineOpcode = N.getOpcode(); unsigned CombineOpcode = N.getOpcode();
const DataLayout *TD = DAG.getSubtarget().getDataLayout();
// Walk up a single-use chain looking for a combinable shuffle. // Walk up a single-use chain looking for a combinable shuffle.
SDValue V = N.getOperand(0); SDValue V = N.getOperand(0);
@ -22415,7 +22415,7 @@ static bool combineRedundantHalfShuffle(SDValue N, MutableArrayRef<int> Mask,
// Merge this node's mask and our incoming mask (adjusted to account for all // Merge this node's mask and our incoming mask (adjusted to account for all
// the pshufd instructions encountered). // the pshufd instructions encountered).
SmallVector<int, 4> VMask = getPSHUFShuffleMask(V); SmallVector<int, 4> VMask = getPSHUFShuffleMask(V, TD);
for (int &M : Mask) for (int &M : Mask)
M = VMask[M]; M = VMask[M];
V = DAG.getNode(V.getOpcode(), DL, MVT::v8i16, V.getOperand(0), V = DAG.getNode(V.getOpcode(), DL, MVT::v8i16, V.getOperand(0),
@ -22437,13 +22437,14 @@ static SDValue PerformTargetShuffleCombine(SDValue N, SelectionDAG &DAG,
const X86Subtarget *Subtarget) { const X86Subtarget *Subtarget) {
SDLoc DL(N); SDLoc DL(N);
MVT VT = N.getSimpleValueType(); MVT VT = N.getSimpleValueType();
const DataLayout *TD = Subtarget->getDataLayout();
SmallVector<int, 4> Mask; SmallVector<int, 4> Mask;
switch (N.getOpcode()) { switch (N.getOpcode()) {
case X86ISD::PSHUFD: case X86ISD::PSHUFD:
case X86ISD::PSHUFLW: case X86ISD::PSHUFLW:
case X86ISD::PSHUFHW: case X86ISD::PSHUFHW:
Mask = getPSHUFShuffleMask(N); Mask = getPSHUFShuffleMask(N, TD);
assert(Mask.size() == 4); assert(Mask.size() == 4);
break; break;
default: default:
@ -22495,8 +22496,8 @@ static SDValue PerformTargetShuffleCombine(SDValue N, SelectionDAG &DAG,
while (D.getOpcode() == ISD::BITCAST && D.hasOneUse()) while (D.getOpcode() == ISD::BITCAST && D.hasOneUse())
D = D.getOperand(0); D = D.getOperand(0);
if (D.getOpcode() == X86ISD::PSHUFD && D.hasOneUse()) { if (D.getOpcode() == X86ISD::PSHUFD && D.hasOneUse()) {
SmallVector<int, 4> VMask = getPSHUFShuffleMask(V); SmallVector<int, 4> VMask = getPSHUFShuffleMask(V, TD);
SmallVector<int, 4> DMask = getPSHUFShuffleMask(D); SmallVector<int, 4> DMask = getPSHUFShuffleMask(D, TD);
int NOffset = N.getOpcode() == X86ISD::PSHUFLW ? 0 : 4; int NOffset = N.getOpcode() == X86ISD::PSHUFLW ? 0 : 4;
int VOffset = V.getOpcode() == X86ISD::PSHUFLW ? 0 : 4; int VOffset = V.getOpcode() == X86ISD::PSHUFLW ? 0 : 4;
int WordMask[8]; int WordMask[8];
@ -22749,9 +22750,10 @@ static SDValue XFormVExtractWithShuffleIntoLoad(SDNode *N, SelectionDAG &DAG,
if (!InVec.hasOneUse()) if (!InVec.hasOneUse())
return SDValue(); return SDValue();
const DataLayout *TD = DAG.getSubtarget().getDataLayout();
SmallVector<int, 16> ShuffleMask; SmallVector<int, 16> ShuffleMask;
bool UnaryShuffle; bool UnaryShuffle;
if (!getTargetShuffleMask(InVec.getNode(), CurrentVT.getSimpleVT(), if (!getTargetShuffleMask(InVec.getNode(), CurrentVT.getSimpleVT(), TD,
ShuffleMask, UnaryShuffle)) ShuffleMask, UnaryShuffle))
return SDValue(); return SDValue();

2
lib/Target/X86/X86MCInstLower.cpp
View File

@ -1161,7 +1161,7 @@ void X86AsmPrinter::EmitInstruction(const MachineInstr *MI) {
if (auto *C = getConstantFromPool(*MI, MaskOp)) { if (auto *C = getConstantFromPool(*MI, MaskOp)) {
SmallVector<int, 16> Mask; SmallVector<int, 16> Mask;
DecodePSHUFBMask(C, Mask); DecodePSHUFBMask(C, TM.getSubtargetImpl()->getDataLayout(), Mask);
if (!Mask.empty()) if (!Mask.empty())
OutStreamer.AddComment(getShuffleComment(DstOp, SrcOp, Mask)); OutStreamer.AddComment(getShuffleComment(DstOp, SrcOp, Mask));
} }

10
test/CodeGen/X86/pshufb-mask-comments.ll
View File

@ -27,4 +27,14 @@ define <16 x i8> @test3(<16 x i8> %V) {
ret <16 x i8> %1 ret <16 x i8> %1
} }
; Test that we won't crash when the constant was reused for another instruction.
define <16 x i8> @test4(<2 x i64>* %V) {
; CHECK-LABEL: test4
; CHECK: pshufb {{.*}}# xmm0 = xmm0[8,9,10,11,12,13,14,15,0,1,2,3,4,5,6,7]
store <2 x i64> <i64 1084818905618843912, i64 506097522914230528>, <2 x i64>* %V, align 16
%1 = tail call <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8> undef, <16 x i8> <i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>)
ret <16 x i8> %1
}
declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>) nounwind readnone declare <16 x i8> @llvm.x86.ssse3.pshuf.b.128(<16 x i8>, <16 x i8>) nounwind readnone