summaryrefslogtreecommitdiffstats
path: root/contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
diff options
context:
space:
mode:
authordim <dim@FreeBSD.org>2014-03-21 17:53:59 +0000
committerdim <dim@FreeBSD.org>2014-03-21 17:53:59 +0000
commit9cedb8bb69b89b0f0c529937247a6a80cabdbaec (patch)
treec978f0e9ec1ab92dc8123783f30b08a7fd1e2a39 /contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
parent03fdc2934eb61c44c049a02b02aa974cfdd8a0eb (diff)
downloadFreeBSD-src-9cedb8bb69b89b0f0c529937247a6a80cabdbaec.zip
FreeBSD-src-9cedb8bb69b89b0f0c529937247a6a80cabdbaec.tar.gz
MFC 261991:
Upgrade our copy of llvm/clang to 3.4 release. This version supports all of the features in the current working draft of the upcoming C++ standard, provisionally named C++1y. The code generator's performance is greatly increased, and the loop auto-vectorizer is now enabled at -Os and -O2 in addition to -O3. The PowerPC backend has made several major improvements to code generation quality and compile time, and the X86, SPARC, ARM32, Aarch64 and SystemZ backends have all seen major feature work. Release notes for llvm and clang can be found here: <http://llvm.org/releases/3.4/docs/ReleaseNotes.html> <http://llvm.org/releases/3.4/tools/clang/docs/ReleaseNotes.html> MFC 262121 (by emaste): Update lldb for clang/llvm 3.4 import This commit largely restores the lldb source to the upstream r196259 snapshot with the addition of threaded inferior support and a few bug fixes. Specific upstream lldb revisions restored include: SVN git 181387 779e6ac 181703 7bef4e2 182099 b31044e 182650 f2dcf35 182683 0d91b80 183862 15c1774 183929 99447a6 184177 0b2934b 184948 4dc3761 184954 007e7bc 186990 eebd175 Sponsored by: DARPA, AFRL MFC 262186 (by emaste): Fix mismerge in r262121 A break statement was lost in the merge. The error had no functional impact, but restore it to reduce the diff against upstream. MFC 262303: Pull in r197521 from upstream clang trunk (by rdivacky): Use the integrated assembler by default on FreeBSD/ppc and ppc64. Requested by: jhibbits MFC 262611: Pull in r196874 from upstream llvm trunk: Fix a crash that occurs when PWD is invalid. MCJIT needs to be able to run in hostile environments, even when PWD is invalid. There's no need to crash MCJIT in this case. The obvious fix is to simply leave MCContext's CompilationDir empty when PWD can't be determined. This way, MCJIT clients, and other clients that link with LLVM don't need a valid working directory. If we do want to guarantee valid CompilationDir, that should be done only for clients of getCompilationDir(). This is as simple as checking for an empty string. The only current use of getCompilationDir is EmitGenDwarfInfo, which won't conceivably run with an invalid working dir. However, in the purely hypothetically and untestable case that this happens, the AT_comp_dir will be omitted from the compilation_unit DIE. This should help fix assertions occurring with ports-mgmt/tinderbox, when it is using jails, and sometimes invalidates clang's current working directory. Reported by: decke MFC 262809: Pull in r203007 from upstream clang trunk: Don't produce an alias between destructors with different calling conventions. Fixes pr19007. (Please note that is an LLVM PR identifier, not a FreeBSD one.) This should fix Firefox and/or libxul crashes (due to problems with regparm/stdcall calling conventions) on i386. Reported by: multiple users on freebsd-current PR: bin/187103 MFC 263048: Repair recognition of "CC" as an alias for the C++ compiler, since it was silently broken by upstream for a Windows-specific use-case. Apparently some versions of CMake still rely on this archaic feature... Reported by: rakuco MFC 263049: Garbage collect the old way of adding the libstdc++ include directories in clang's InitHeaderSearch.cpp. This has been superseded by David Chisnall's commit in r255321. Moreover, if libc++ is used, the libstdc++ include directories should not be in the search path at all. These directories are now only used if you pass -stdlib=libstdc++.
Diffstat (limited to 'contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp')
-rw-r--r--contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp90
1 files changed, 45 insertions, 45 deletions
diff --git a/contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp b/contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
index 17900da..c5e1dcb 100644
--- a/contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
+++ b/contrib/llvm/lib/Transforms/Vectorize/BBVectorize.cpp
@@ -356,7 +356,7 @@ namespace {
Instruction *J, unsigned o, bool IBeforeJ);
void getReplacementInputsForPair(LLVMContext& Context, Instruction *I,
- Instruction *J, SmallVector<Value *, 3> &ReplacedOperands,
+ Instruction *J, SmallVectorImpl<Value *> &ReplacedOperands,
bool IBeforeJ);
void replaceOutputsOfPair(LLVMContext& Context, Instruction *I,
@@ -533,7 +533,7 @@ namespace {
default: break;
case Instruction::GetElementPtr:
// We mark this instruction as zero-cost because scalar GEPs are usually
- // lowered to the intruction addressing mode. At the moment we don't
+ // lowered to the instruction addressing mode. At the moment we don't
// generate vector GEPs.
return 0;
case Instruction::Br:
@@ -625,10 +625,10 @@ namespace {
ConstantInt *IntOff = ConstOffSCEV->getValue();
int64_t Offset = IntOff->getSExtValue();
- Type *VTy = cast<PointerType>(IPtr->getType())->getElementType();
+ Type *VTy = IPtr->getType()->getPointerElementType();
int64_t VTyTSS = (int64_t) TD->getTypeStoreSize(VTy);
- Type *VTy2 = cast<PointerType>(JPtr->getType())->getElementType();
+ Type *VTy2 = JPtr->getType()->getPointerElementType();
if (VTy != VTy2 && Offset < 0) {
int64_t VTy2TSS = (int64_t) TD->getTypeStoreSize(VTy2);
OffsetInElmts = Offset/VTy2TSS;
@@ -1182,6 +1182,8 @@ namespace {
// Look for an instruction with which to pair instruction *I...
DenseSet<Value *> Users;
AliasSetTracker WriteSet(*AA);
+ if (I->mayWriteToMemory()) WriteSet.add(I);
+
bool JAfterStart = IAfterStart;
BasicBlock::iterator J = llvm::next(I);
for (unsigned ss = 0; J != E && ss <= Config.SearchLimit; ++J, ++ss) {
@@ -1403,6 +1405,8 @@ namespace {
DenseSet<Value *> Users;
AliasSetTracker WriteSet(*AA);
+ if (I->mayWriteToMemory()) WriteSet.add(I);
+
for (BasicBlock::iterator J = llvm::next(I); J != E; ++J) {
(void) trackUsesOfI(Users, WriteSet, I, J);
@@ -1602,7 +1606,7 @@ namespace {
DenseSet<ValuePair> CurrentPairs;
bool CanAdd = true;
- for (SmallVector<ValuePairWithDepth, 8>::iterator C2
+ for (SmallVectorImpl<ValuePairWithDepth>::iterator C2
= BestChildren.begin(), E2 = BestChildren.end();
C2 != E2; ++C2) {
if (C2->first.first == C->first.first ||
@@ -1642,7 +1646,7 @@ namespace {
if (!CanAdd) continue;
// And check the queue too...
- for (SmallVector<ValuePairWithDepth, 32>::iterator C2 = Q.begin(),
+ for (SmallVectorImpl<ValuePairWithDepth>::iterator C2 = Q.begin(),
E2 = Q.end(); C2 != E2; ++C2) {
if (C2->first.first == C->first.first ||
C2->first.first == C->first.second ||
@@ -1691,7 +1695,7 @@ namespace {
// to an already-selected child. Check for this here, and if a
// conflict is found, then remove the previously-selected child
// before adding this one in its place.
- for (SmallVector<ValuePairWithDepth, 8>::iterator C2
+ for (SmallVectorImpl<ValuePairWithDepth>::iterator C2
= BestChildren.begin(); C2 != BestChildren.end();) {
if (C2->first.first == C->first.first ||
C2->first.first == C->first.second ||
@@ -1706,7 +1710,7 @@ namespace {
BestChildren.push_back(ValuePairWithDepth(C->first, C->second));
}
- for (SmallVector<ValuePairWithDepth, 8>::iterator C
+ for (SmallVectorImpl<ValuePairWithDepth>::iterator C
= BestChildren.begin(), E2 = BestChildren.end();
C != E2; ++C) {
size_t DepthF = getDepthFactor(C->first.first);
@@ -2227,11 +2231,12 @@ namespace {
// The pointer value is taken to be the one with the lowest offset.
Value *VPtr = IPtr;
- Type *ArgTypeI = cast<PointerType>(IPtr->getType())->getElementType();
- Type *ArgTypeJ = cast<PointerType>(JPtr->getType())->getElementType();
+ Type *ArgTypeI = IPtr->getType()->getPointerElementType();
+ Type *ArgTypeJ = JPtr->getType()->getPointerElementType();
Type *VArgType = getVecTypeForPair(ArgTypeI, ArgTypeJ);
- Type *VArgPtrType = PointerType::get(VArgType,
- cast<PointerType>(IPtr->getType())->getAddressSpace());
+ Type *VArgPtrType
+ = PointerType::get(VArgType,
+ IPtr->getType()->getPointerAddressSpace());
return new BitCastInst(VPtr, VArgPtrType, getReplacementName(I, true, o),
/* insert before */ I);
}
@@ -2240,7 +2245,7 @@ namespace {
unsigned MaskOffset, unsigned NumInElem,
unsigned NumInElem1, unsigned IdxOffset,
std::vector<Constant*> &Mask) {
- unsigned NumElem1 = cast<VectorType>(J->getType())->getNumElements();
+ unsigned NumElem1 = J->getType()->getVectorNumElements();
for (unsigned v = 0; v < NumElem1; ++v) {
int m = cast<ShuffleVectorInst>(J)->getMaskValue(v);
if (m < 0) {
@@ -2267,18 +2272,18 @@ namespace {
Type *ArgTypeJ = J->getType();
Type *VArgType = getVecTypeForPair(ArgTypeI, ArgTypeJ);
- unsigned NumElemI = cast<VectorType>(ArgTypeI)->getNumElements();
+ unsigned NumElemI = ArgTypeI->getVectorNumElements();
// Get the total number of elements in the fused vector type.
// By definition, this must equal the number of elements in
// the final mask.
- unsigned NumElem = cast<VectorType>(VArgType)->getNumElements();
+ unsigned NumElem = VArgType->getVectorNumElements();
std::vector<Constant*> Mask(NumElem);
Type *OpTypeI = I->getOperand(0)->getType();
- unsigned NumInElemI = cast<VectorType>(OpTypeI)->getNumElements();
+ unsigned NumInElemI = OpTypeI->getVectorNumElements();
Type *OpTypeJ = J->getOperand(0)->getType();
- unsigned NumInElemJ = cast<VectorType>(OpTypeJ)->getNumElements();
+ unsigned NumInElemJ = OpTypeJ->getVectorNumElements();
// The fused vector will be:
// -----------------------------------------------------
@@ -2340,6 +2345,12 @@ namespace {
return ExpandedIEChain;
}
+ static unsigned getNumScalarElements(Type *Ty) {
+ if (VectorType *VecTy = dyn_cast<VectorType>(Ty))
+ return VecTy->getNumElements();
+ return 1;
+ }
+
// Returns the value to be used as the specified operand of the vector
// instruction that fuses I with J.
Value *BBVectorize::getReplacementInput(LLVMContext& Context, Instruction *I,
@@ -2355,17 +2366,8 @@ namespace {
Instruction *L = I, *H = J;
Type *ArgTypeL = ArgTypeI, *ArgTypeH = ArgTypeJ;
- unsigned numElemL;
- if (ArgTypeL->isVectorTy())
- numElemL = cast<VectorType>(ArgTypeL)->getNumElements();
- else
- numElemL = 1;
-
- unsigned numElemH;
- if (ArgTypeH->isVectorTy())
- numElemH = cast<VectorType>(ArgTypeH)->getNumElements();
- else
- numElemH = 1;
+ unsigned numElemL = getNumScalarElements(ArgTypeL);
+ unsigned numElemH = getNumScalarElements(ArgTypeH);
Value *LOp = L->getOperand(o);
Value *HOp = H->getOperand(o);
@@ -2426,11 +2428,12 @@ namespace {
if (CanUseInputs) {
unsigned LOpElem =
- cast<VectorType>(cast<Instruction>(LOp)->getOperand(0)->getType())
- ->getNumElements();
+ cast<Instruction>(LOp)->getOperand(0)->getType()
+ ->getVectorNumElements();
+
unsigned HOpElem =
- cast<VectorType>(cast<Instruction>(HOp)->getOperand(0)->getType())
- ->getNumElements();
+ cast<Instruction>(HOp)->getOperand(0)->getType()
+ ->getVectorNumElements();
// We have one or two input vectors. We need to map each index of the
// operands to the index of the original vector.
@@ -2646,14 +2649,14 @@ namespace {
getReplacementName(IBeforeJ ? I : J,
true, o, 1));
}
-
+
NHOp->insertBefore(IBeforeJ ? J : I);
HOp = NHOp;
}
}
if (ArgType->isVectorTy()) {
- unsigned numElem = cast<VectorType>(VArgType)->getNumElements();
+ unsigned numElem = VArgType->getVectorNumElements();
std::vector<Constant*> Mask(numElem);
for (unsigned v = 0; v < numElem; ++v) {
unsigned Idx = v;
@@ -2687,7 +2690,7 @@ namespace {
// to the vector instruction that fuses I with J.
void BBVectorize::getReplacementInputsForPair(LLVMContext& Context,
Instruction *I, Instruction *J,
- SmallVector<Value *, 3> &ReplacedOperands,
+ SmallVectorImpl<Value *> &ReplacedOperands,
bool IBeforeJ) {
unsigned NumOperands = I->getNumOperands();
@@ -2746,16 +2749,8 @@ namespace {
VectorType *VType = getVecTypeForPair(IType, JType);
unsigned numElem = VType->getNumElements();
- unsigned numElemI, numElemJ;
- if (IType->isVectorTy())
- numElemI = cast<VectorType>(IType)->getNumElements();
- else
- numElemI = 1;
-
- if (JType->isVectorTy())
- numElemJ = cast<VectorType>(JType)->getNumElements();
- else
- numElemJ = 1;
+ unsigned numElemI = getNumScalarElements(IType);
+ unsigned numElemJ = getNumScalarElements(JType);
if (IType->isVectorTy()) {
std::vector<Constant*> Mask1(numElemI), Mask2(numElemI);
@@ -2804,6 +2799,8 @@ namespace {
DenseSet<Value *> Users;
AliasSetTracker WriteSet(*AA);
+ if (I->mayWriteToMemory()) WriteSet.add(I);
+
for (; cast<Instruction>(L) != J; ++L)
(void) trackUsesOfI(Users, WriteSet, I, L, true, &LoadMoveSetPairs);
@@ -2824,6 +2821,8 @@ namespace {
DenseSet<Value *> Users;
AliasSetTracker WriteSet(*AA);
+ if (I->mayWriteToMemory()) WriteSet.add(I);
+
for (; cast<Instruction>(L) != J;) {
if (trackUsesOfI(Users, WriteSet, I, L, true, &LoadMoveSetPairs)) {
// Move this instruction
@@ -2853,6 +2852,7 @@ namespace {
DenseSet<Value *> Users;
AliasSetTracker WriteSet(*AA);
+ if (I->mayWriteToMemory()) WriteSet.add(I);
// Note: We cannot end the loop when we reach J because J could be moved
// farther down the use chain by another instruction pairing. Also, J
OpenPOWER on IntegriCloud