d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge llvm 3.6.0rc2 from ^/vendor/llvm/dist, merge clang 3.6.0rc2 from

Browse Source 
^/vendor/clang/dist, resolve conflicts, and cleanup patches.
This commit is contained in:
Dimitry Andric 2015-01-31 21:57:38 +00:00
commit 3de688eb16
101 changed files with 1231 additions and 1076 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
contrib/llvm/include/llvm/IR/LegacyPassManagers.h
View File

@ -195,6 +195,9 @@ public:
/// then return NULL.
Pass *findAnalysisPass(AnalysisID AID);
/// Retrieve the PassInfo for an analysis.
const PassInfo *findAnalysisPassInfo(AnalysisID AID) const;
/// Find analysis usage information for the pass P.
AnalysisUsage *findAnalysisUsage(Pass *P);
@ -251,6 +254,12 @@ private:
SmallVector<ImmutablePass *, 16> ImmutablePasses;
DenseMap<Pass *, AnalysisUsage *> AnUsageMap;
/// Collection of PassInfo objects found via analysis IDs and in this top
/// level manager. This is used to memoize queries to the pass registry.
/// FIXME: This is an egregious hack because querying the pass registry is
/// either slow or racy.
mutable DenseMap<AnalysisID, const PassInfo *> AnalysisPassInfos;
};

6
contrib/llvm/include/llvm/MC/MCAsmBackend.h
View File

@ -61,6 +61,12 @@ public:
/// markers. If not, data region directives will be ignored.
bool hasDataInCodeSupport() const { return HasDataInCodeSupport; }
/// doesSectionRequireSymbols - Check whether the given section requires that
/// all symbols (even temporaries) have symbol table entries.
virtual bool doesSectionRequireSymbols(const MCSection &Section) const {
return false;
}
/// @name Target Fixup Interfaces
/// @{

6
contrib/llvm/include/llvm/MC/MCAssembler.h
View File

@ -11,7 +11,6 @@
#define LLVM_MC_MCASSEMBLER_H
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/PointerIntPair.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallString.h"
@ -882,8 +881,6 @@ private:
iplist<MCSymbolData> Symbols;
DenseSet<const MCSymbol *> LocalsUsedInReloc;
/// The map of sections to their associated assembler backend data.
//
// FIXME: Avoid this indirection?
@ -983,9 +980,6 @@ private:
MCFragment &F, const MCFixup &Fixup);
public:
void addLocalUsedInReloc(const MCSymbol &Sym);
bool isLocalUsedInReloc(const MCSymbol &Sym) const;
/// Compute the effective fragment size assuming it is laid out at the given
/// \p SectionAddress and \p FragmentOffset.
uint64_t computeFragmentSize(const MCAsmLayout &Layout,

28
contrib/llvm/include/llvm/MC/MCMachObjectWriter.h
View File

@ -68,10 +68,12 @@ public:
/// @name API
/// @{
virtual void RecordRelocation(MachObjectWriter *Writer, MCAssembler &Asm,
virtual void RecordRelocation(MachObjectWriter *Writer,
const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
const MCFixup &Fixup,
MCValue Target,
uint64_t &FixedValue) = 0;
/// @}
@ -95,14 +97,8 @@ class MachObjectWriter : public MCObjectWriter {
/// @name Relocation Data
/// @{
struct RelAndSymbol {
const MCSymbolData *Sym;
MachO::any_relocation_info MRE;
RelAndSymbol(const MCSymbolData *Sym, const MachO::any_relocation_info &MRE)
: Sym(Sym), MRE(MRE) {}
};
llvm::DenseMap<const MCSectionData *, std::vector<RelAndSymbol>> Relocations;
llvm::DenseMap<const MCSectionData*,
std::vector<MachO::any_relocation_info> > Relocations;
llvm::DenseMap<const MCSectionData*, unsigned> IndirectSymBase;
/// @}
@ -217,15 +213,9 @@ public:
// - Input errors, where something cannot be correctly encoded. 'as' allows
// these through in many cases.
// Add a relocation to be output in the object file. At the time this is
// called, the symbol indexes are not know, so if the relocation refers
// to a symbol it should be passed as \p RelSymbol so that it can be updated
// afterwards. If the relocation doesn't refer to a symbol, nullptr should be
// used.
void addRelocation(const MCSymbolData *RelSymbol, const MCSectionData *SD,
void addRelocation(const MCSectionData *SD,
MachO::any_relocation_info &MRE) {
RelAndSymbol P(RelSymbol, MRE);
Relocations[SD].push_back(P);
Relocations[SD].push_back(MRE);
}
void RecordScatteredRelocation(const MCAssembler &Asm,
@ -241,7 +231,7 @@ public:
const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue);
void RecordRelocation(MCAssembler &Asm, const MCAsmLayout &Layout,
void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, bool &IsPCRel,
uint64_t &FixedValue) override;

6
contrib/llvm/include/llvm/MC/MCObjectWriter.h
View File

@ -76,10 +76,12 @@ public:
/// post layout binding. The implementation is responsible for storing
/// information about the relocation so that it can be emitted during
/// WriteObject().
virtual void RecordRelocation(MCAssembler &Asm, const MCAsmLayout &Layout,
virtual void RecordRelocation(const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
bool &IsPCRel, uint64_t &FixedValue) = 0;
bool &IsPCRel,
uint64_t &FixedValue) = 0;
/// \brief Check whether the difference (A - B) between two symbol
/// references is fully resolved.

7
contrib/llvm/include/llvm/Transforms/Utils/SimplifyLibCalls.h
View File

@ -17,6 +17,7 @@
#include "llvm/ADT/StringRef.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/Target/TargetLibraryInfo.h"
namespace llvm {
class Value;
@ -53,8 +54,10 @@ private:
Value *optimizeMemCpyChk(CallInst *CI, IRBuilder<> &B);
Value *optimizeMemMoveChk(CallInst *CI, IRBuilder<> &B);
Value *optimizeMemSetChk(CallInst *CI, IRBuilder<> &B);
Value *optimizeStrCpyChk(CallInst *CI, IRBuilder<> &B);
Value *optimizeStrNCpyChk(CallInst *CI, IRBuilder<> &B);
// Str/Stp cpy are similar enough to be handled in the same functions.
Value *optimizeStrpCpyChk(CallInst *CI, IRBuilder<> &B, LibFunc::Func Func);
Value *optimizeStrpNCpyChk(CallInst *CI, IRBuilder<> &B, LibFunc::Func Func);
/// \brief Checks whether the call \p CI to a fortified libcall is foldable
/// to the non-fortified version.

15
contrib/llvm/lib/Analysis/ScalarEvolution.cpp
View File

@ -3154,8 +3154,9 @@ const SCEV *ScalarEvolution::getMinusSCEV(const SCEV *LHS, const SCEV *RHS,
if (LHS == RHS)
return getConstant(LHS->getType(), 0);
// X - Y --> X + -Y
return getAddExpr(LHS, getNegativeSCEV(RHS), Flags);
// X - Y --> X + -Y.
// X -(nsw || nuw) Y --> X + -Y.
return getAddExpr(LHS, getNegativeSCEV(RHS));
}
/// getTruncateOrZeroExtend - Return a SCEV corresponding to a conversion of the
@ -3461,12 +3462,10 @@ const SCEV *ScalarEvolution::createNodeForPHI(PHINode *PN) {
if (isKnownPositive(getMinusSCEV(getSCEV(GEP), Ptr)))
Flags = setFlags(Flags, SCEV::FlagNUW);
}
} else if (const SubOperator *OBO =
dyn_cast<SubOperator>(BEValueV)) {
if (OBO->hasNoUnsignedWrap())
Flags = setFlags(Flags, SCEV::FlagNUW);
if (OBO->hasNoSignedWrap())
Flags = setFlags(Flags, SCEV::FlagNSW);
// We cannot transfer nuw and nsw flags from subtraction
// operations -- sub nuw X, Y is not the same as add nuw X, -Y
// for instance.
}
const SCEV *StartVal = getSCEV(StartValueV);

7
contrib/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
View File

@ -626,10 +626,7 @@ static uint64_t getBaseTypeSize(DwarfDebug *DD, DIDerivedType Ty) {
DIType BaseType = DD->resolve(Ty.getTypeDerivedFrom());
// If this type is not derived from any type or the type is a declaration then
// take conservative approach.
if (!BaseType.isValid() || BaseType.isForwardDecl())
return Ty.getSizeInBits();
assert(BaseType.isValid());
// If this is a derived type, go ahead and get the base type, unless it's a
// reference then it's just the size of the field. Pointer types have no need
@ -1473,7 +1470,7 @@ void DwarfUnit::constructMemberDIE(DIE &Buffer, DIDerivedType DT) {
uint64_t FieldSize = getBaseTypeSize(DD, DT);
uint64_t OffsetInBytes;
if (Size != FieldSize) {
if (FieldSize && Size != FieldSize) {
// Handle bitfield, assume bytes are 8 bits.
addUInt(MemberDie, dwarf::DW_AT_byte_size, None, FieldSize/8);
addUInt(MemberDie, dwarf::DW_AT_bit_size, None, Size);

30
contrib/llvm/lib/CodeGen/MachineFunctionPass.cpp
View File

@ -11,11 +11,19 @@
//
//===----------------------------------------------------------------------===//
#include "llvm/IR/Function.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/CodeGen/MachineFunctionAnalysis.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/DominanceFrontier.h"
#include "llvm/Analysis/IVUsers.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/MemoryDependenceAnalysis.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/CodeGen/MachineFunctionAnalysis.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/CodeGen/StackProtector.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
using namespace llvm;
Pass *MachineFunctionPass::createPrinterPass(raw_ostream &O,
@ -43,15 +51,13 @@ void MachineFunctionPass::getAnalysisUsage(AnalysisUsage &AU) const {
// because CodeGen overloads that to mean preserving the MachineBasicBlock
// CFG in addition to the LLVM IR CFG.
AU.addPreserved<AliasAnalysis>();
AU.addPreserved("scalar-evolution");
AU.addPreserved("iv-users");
AU.addPreserved("memdep");
AU.addPreserved("live-values");
AU.addPreserved("domtree");
AU.addPreserved("domfrontier");
AU.addPreserved("loops");
AU.addPreserved("lda");
AU.addPreserved("stack-protector");
AU.addPreserved<DominanceFrontier>();
AU.addPreserved<DominatorTreeWrapperPass>();
AU.addPreserved<IVUsers>();
AU.addPreserved<LoopInfo>();
AU.addPreserved<MemoryDependenceAnalysis>();
AU.addPreserved<ScalarEvolution>();
AU.addPreserved<StackProtector>();
FunctionPass::getAnalysisUsage(AU);
}

22
contrib/llvm/lib/IR/Core.cpp
View File

@ -563,9 +563,23 @@ LLVMValueRef LLVMGetMetadata(LLVMValueRef Inst, unsigned KindID) {
return nullptr;
}
void LLVMSetMetadata(LLVMValueRef Inst, unsigned KindID, LLVMValueRef MD) {
MDNode *N =
MD ? cast<MDNode>(unwrap<MetadataAsValue>(MD)->getMetadata()) : nullptr;
// MetadataAsValue uses a canonical format which strips the actual MDNode for
// MDNode with just a single constant value, storing just a ConstantAsMetadata
// This undoes this canonicalization, reconstructing the MDNode.
static MDNode *extractMDNode(MetadataAsValue *MAV) {
Metadata *MD = MAV->getMetadata();
assert((isa<MDNode>(MD) || isa<ConstantAsMetadata>(MD)) &&
"Expected a metadata node or a canonicalized constant");
if (MDNode *N = dyn_cast<MDNode>(MD))
return N;
return MDNode::get(MAV->getContext(), MD);
}
void LLVMSetMetadata(LLVMValueRef Inst, unsigned KindID, LLVMValueRef Val) {
MDNode *N = Val ? extractMDNode(unwrap<MetadataAsValue>(Val)) : nullptr;
unwrap<Instruction>(Inst)->setMetadata(KindID, N);
}
@ -795,7 +809,7 @@ void LLVMAddNamedMetadataOperand(LLVMModuleRef M, const char* name,
return;
if (!Val)
return;
N->addOperand(cast<MDNode>(unwrap<MetadataAsValue>(Val)->getMetadata()));
N->addOperand(extractMDNode(unwrap<MetadataAsValue>(Val)));
}
/*--.. Operations on scalar constants ......................................--*/

36
contrib/llvm/lib/IR/LegacyPassManager.cpp
View File

@ -600,8 +600,7 @@ void PMTopLevelManager::schedulePass(Pass *P) {
// If P is an analysis pass and it is available then do not
// generate the analysis again. Stale analysis info should not be
// available at this point.
const PassInfo *PI =
PassRegistry::getPassRegistry()->getPassInfo(P->getPassID());
const PassInfo *PI = findAnalysisPassInfo(P->getPassID());
if (PI && PI->isAnalysis() && findAnalysisPass(P->getPassID())) {
delete P;
return;
@ -619,7 +618,7 @@ void PMTopLevelManager::schedulePass(Pass *P) {
Pass *AnalysisPass = findAnalysisPass(*I);
if (!AnalysisPass) {
const PassInfo *PI = PassRegistry::getPassRegistry()->getPassInfo(*I);
const PassInfo *PI = findAnalysisPassInfo(*I);
if (!PI) {
// Pass P is not in the global PassRegistry
@ -716,8 +715,7 @@ Pass *PMTopLevelManager::findAnalysisPass(AnalysisID AID) {
return *I;
// If Pass not found then check the interfaces implemented by Immutable Pass
const PassInfo *PassInf =
PassRegistry::getPassRegistry()->getPassInfo(PI);
const PassInfo *PassInf = findAnalysisPassInfo(PI);
assert(PassInf && "Expected all immutable passes to be initialized");
const std::vector<const PassInfo*> &ImmPI =
PassInf->getInterfacesImplemented();
@ -731,6 +729,17 @@ Pass *PMTopLevelManager::findAnalysisPass(AnalysisID AID) {
return nullptr;
}
const PassInfo *PMTopLevelManager::findAnalysisPassInfo(AnalysisID AID) const {
const PassInfo *&PI = AnalysisPassInfos[AID];
if (!PI)
PI = PassRegistry::getPassRegistry()->getPassInfo(AID);
else
assert(PI == PassRegistry::getPassRegistry()->getPassInfo(AID) &&
"The pass info pointer changed for an analysis ID!");
return PI;
}
// Print passes managed by this top level manager.
void PMTopLevelManager::dumpPasses() const {
@ -759,8 +768,7 @@ void PMTopLevelManager::dumpArguments() const {
dbgs() << "Pass Arguments: ";
for (SmallVectorImpl<ImmutablePass *>::const_iterator I =
ImmutablePasses.begin(), E = ImmutablePasses.end(); I != E; ++I)
if (const PassInfo *PI =
PassRegistry::getPassRegistry()->getPassInfo((*I)->getPassID())) {
if (const PassInfo *PI = findAnalysisPassInfo((*I)->getPassID())) {
assert(PI && "Expected all immutable passes to be initialized");
if (!PI->isAnalysisGroup())
dbgs() << " -" << PI->getPassArgument();
@ -824,7 +832,7 @@ void PMDataManager::recordAvailableAnalysis(Pass *P) {
// This pass is the current implementation of all of the interfaces it
// implements as well.
const PassInfo *PInf = PassRegistry::getPassRegistry()->getPassInfo(PI);
const PassInfo *PInf = TPM->findAnalysisPassInfo(PI);
if (!PInf) return;
const std::vector<const PassInfo*> &II = PInf->getInterfacesImplemented();
for (unsigned i = 0, e = II.size(); i != e; ++i)
@ -957,7 +965,7 @@ void PMDataManager::freePass(Pass *P, StringRef Msg,
}
AnalysisID PI = P->getPassID();
if (const PassInfo *PInf = PassRegistry::getPassRegistry()->getPassInfo(PI)) {
if (const PassInfo *PInf = TPM->findAnalysisPassInfo(PI)) {
// Remove the pass itself (if it is not already removed).
AvailableAnalysis.erase(PI);
@ -1037,7 +1045,7 @@ void PMDataManager::add(Pass *P, bool ProcessAnalysis) {
for (SmallVectorImpl<AnalysisID>::iterator
I = ReqAnalysisNotAvailable.begin(),
E = ReqAnalysisNotAvailable.end() ;I != E; ++I) {
const PassInfo *PI = PassRegistry::getPassRegistry()->getPassInfo(*I);
const PassInfo *PI = TPM->findAnalysisPassInfo(*I);
Pass *AnalysisPass = PI->createPass();
this->addLowerLevelRequiredPass(P, AnalysisPass);
}
@ -1142,7 +1150,7 @@ void PMDataManager::dumpPassArguments() const {
PMD->dumpPassArguments();
else
if (const PassInfo *PI =
PassRegistry::getPassRegistry()->getPassInfo((*I)->getPassID()))
TPM->findAnalysisPassInfo((*I)->getPassID()))
if (!PI->isAnalysisGroup())
dbgs() << " -" << PI->getPassArgument();
}
@ -1218,7 +1226,7 @@ void PMDataManager::dumpAnalysisUsage(StringRef Msg, const Pass *P,
dbgs() << (const void*)P << std::string(getDepth()*2+3, ' ') << Msg << " Analyses:";
for (unsigned i = 0; i != Set.size(); ++i) {
if (i) dbgs() << ',';
const PassInfo *PInf = PassRegistry::getPassRegistry()->getPassInfo(Set[i]);
const PassInfo *PInf = TPM->findAnalysisPassInfo(Set[i]);
if (!PInf) {
// Some preserved passes, such as AliasAnalysis, may not be initialized by
// all drivers.
@ -1658,8 +1666,8 @@ void MPPassManager::addLowerLevelRequiredPass(Pass *P, Pass *RequiredPass) {
OnTheFlyManagers[P] = FPP;
}
const PassInfo * RequiredPassPI =
PassRegistry::getPassRegistry()->getPassInfo(RequiredPass->getPassID());
const PassInfo *RequiredPassPI =
TPM->findAnalysisPassInfo(RequiredPass->getPassID());
Pass *FoundPass = nullptr;
if (RequiredPassPI && RequiredPassPI->isAnalysis()) {

10
contrib/llvm/lib/MC/ELFObjectWriter.cpp
View File

@ -219,7 +219,7 @@ class ELFObjectWriter : public MCObjectWriter {
const MCSymbolData *SD, uint64_t C,
unsigned Type) const;
void RecordRelocation(MCAssembler &Asm, const MCAsmLayout &Layout,
void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, bool &IsPCRel,
uint64_t &FixedValue) override;
@ -789,11 +789,13 @@ static const MCSymbol *getWeakRef(const MCSymbolRefExpr &Ref) {
return nullptr;
}
void ELFObjectWriter::RecordRelocation(MCAssembler &Asm,
void ELFObjectWriter::RecordRelocation(const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
bool &IsPCRel, uint64_t &FixedValue) {
const MCFixup &Fixup,
MCValue Target,
bool &IsPCRel,
uint64_t &FixedValue) {
const MCSectionData *FixupSection = Fragment->getParent();
uint64_t C = Target.getConstant();
uint64_t FixupOffset = Layout.getFragmentOffset(Fragment) + Fixup.getOffset();

17
contrib/llvm/lib/MC/MCAsmInfoDarwin.cpp
View File

@ -27,7 +27,22 @@ bool MCAsmInfoDarwin::isSectionAtomizableBySymbols(
// contain.
// Sections holding 2 byte strings require symbols in order to be atomized.
// There is no dedicated section for 4 byte strings.
if (SMO.getType() == MachO::S_CSTRING_LITERALS)
if (SMO.getKind().isMergeable1ByteCString())
return false;
if (SMO.getSegmentName() == "__TEXT" &&
SMO.getSectionName() == "__objc_classname" &&
SMO.getType() == MachO::S_CSTRING_LITERALS)
return false;
if (SMO.getSegmentName() == "__TEXT" &&
SMO.getSectionName() == "__objc_methname" &&
SMO.getType() == MachO::S_CSTRING_LITERALS)
return false;
if (SMO.getSegmentName() == "__TEXT" &&
SMO.getSectionName() == "__objc_methtype" &&
SMO.getType() == MachO::S_CSTRING_LITERALS)
return false;
if (SMO.getSegmentName() == "__DATA" && SMO.getSectionName() == "__cfstring")

16
contrib/llvm/lib/MC/MCAssembler.cpp
View File

@ -425,16 +425,6 @@ bool MCAssembler::isThumbFunc(const MCSymbol *Symbol) const {
return true;
}
void MCAssembler::addLocalUsedInReloc(const MCSymbol &Sym) {
assert(Sym.isTemporary());
LocalsUsedInReloc.insert(&Sym);
}
bool MCAssembler::isLocalUsedInReloc(const MCSymbol &Sym) const {
assert(Sym.isTemporary());
return LocalsUsedInReloc.count(&Sym);
}
bool MCAssembler::isSymbolLinkerVisible(const MCSymbol &Symbol) const {
// Non-temporary labels should always be visible to the linker.
if (!Symbol.isTemporary())
@ -444,10 +434,8 @@ bool MCAssembler::isSymbolLinkerVisible(const MCSymbol &Symbol) const {
if (!Symbol.isInSection())
return false;
if (isLocalUsedInReloc(Symbol))
return true;
return false;
// Otherwise, check if the section requires symbols even for temporary labels.
return getBackend().doesSectionRequireSymbols(Symbol.getSection());
}
const MCSymbolData *MCAssembler::getAtom(const MCSymbolData *SD) const {

41
contrib/llvm/lib/MC/MachObjectWriter.cpp
View File

@ -448,11 +448,14 @@ void MachObjectWriter::WriteLinkerOptionsLoadCommand(
assert(OS.tell() - Start == Size);
}
void MachObjectWriter::RecordRelocation(MCAssembler &Asm,
void MachObjectWriter::RecordRelocation(const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
bool &IsPCRel, uint64_t &FixedValue) {
const MCFixup &Fixup,
MCValue Target,
bool &IsPCRel,
uint64_t &FixedValue) {
TargetObjectWriter->RecordRelocation(this, Asm, Layout, Fragment, Fixup,
Target, FixedValue);
}
@ -613,22 +616,6 @@ void MachObjectWriter::ComputeSymbolTable(
ExternalSymbolData[i].SymbolData->setIndex(Index++);
for (unsigned i = 0, e = UndefinedSymbolData.size(); i != e; ++i)
UndefinedSymbolData[i].SymbolData->setIndex(Index++);
for (const MCSectionData &SD : Asm) {
std::vector<RelAndSymbol> &Relocs = Relocations[&SD];
for (RelAndSymbol &Rel : Relocs) {
if (!Rel.Sym)
continue;
// Set the Index and the IsExtern bit.
unsigned Index = Rel.Sym->getIndex();
assert(isInt<24>(Index));
if (IsLittleEndian)
Rel.MRE.r_word1 = (Rel.MRE.r_word1 & (-1 << 24)) | Index | (1 << 27);
else
Rel.MRE.r_word1 = (Rel.MRE.r_word1 & 0xff) | Index << 8 | (1 << 4);
}
}
}
void MachObjectWriter::computeSectionAddresses(const MCAssembler &Asm,
@ -675,6 +662,10 @@ void MachObjectWriter::ExecutePostLayoutBinding(MCAssembler &Asm,
// Mark symbol difference expressions in variables (from .set or = directives)
// as absolute.
markAbsoluteVariableSymbols(Asm, Layout);
// Compute symbol table information and bind symbol indices.
ComputeSymbolTable(Asm, LocalSymbolData, ExternalSymbolData,
UndefinedSymbolData);
}
bool MachObjectWriter::
@ -758,10 +749,6 @@ IsSymbolRefDifferenceFullyResolvedImpl(const MCAssembler &Asm,
void MachObjectWriter::WriteObject(MCAssembler &Asm,
const MCAsmLayout &Layout) {
// Compute symbol table information and bind symbol indices.
ComputeSymbolTable(Asm, LocalSymbolData, ExternalSymbolData,
UndefinedSymbolData);
unsigned NumSections = Asm.size();
const MCAssembler::VersionMinInfoType &VersionInfo =
Layout.getAssembler().getVersionMinInfo();
@ -852,7 +839,7 @@ void MachObjectWriter::WriteObject(MCAssembler &Asm,
uint64_t RelocTableEnd = SectionDataStart + SectionDataFileSize;
for (MCAssembler::const_iterator it = Asm.begin(),
ie = Asm.end(); it != ie; ++it) {
std::vector<RelAndSymbol> &Relocs = Relocations[it];
std::vector<MachO::any_relocation_info> &Relocs = Relocations[it];
unsigned NumRelocs = Relocs.size();
uint64_t SectionStart = SectionDataStart + getSectionAddress(it);
WriteSection(Asm, Layout, *it, SectionStart, RelocTableEnd, NumRelocs);
@ -946,10 +933,10 @@ void MachObjectWriter::WriteObject(MCAssembler &Asm,
ie = Asm.end(); it != ie; ++it) {
// Write the section relocation entries, in reverse order to match 'as'
// (approximately, the exact algorithm is more complicated than this).
std::vector<RelAndSymbol> &Relocs = Relocations[it];
std::vector<MachO::any_relocation_info> &Relocs = Relocations[it];
for (unsigned i = 0, e = Relocs.size(); i != e; ++i) {
Write32(Relocs[e - i - 1].MRE.r_word0);
Write32(Relocs[e - i - 1].MRE.r_word1);
Write32(Relocs[e - i - 1].r_word0);
Write32(Relocs[e - i - 1].r_word1);
}
}

12
contrib/llvm/lib/MC/WinCOFFObjectWriter.cpp
View File

@ -175,7 +175,7 @@ public:
const MCFragment &FB, bool InSet,
bool IsPCRel) const override;
void RecordRelocation(MCAssembler &Asm, const MCAsmLayout &Layout,
void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, bool &IsPCRel,
uint64_t &FixedValue) override;
@ -661,9 +661,13 @@ bool WinCOFFObjectWriter::IsSymbolRefDifferenceFullyResolvedImpl(
InSet, IsPCRel);
}
void WinCOFFObjectWriter::RecordRelocation(
MCAssembler &Asm, const MCAsmLayout &Layout, const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target, bool &IsPCRel, uint64_t &FixedValue) {
void WinCOFFObjectWriter::RecordRelocation(const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup,
MCValue Target,
bool &IsPCRel,
uint64_t &FixedValue) {
assert(Target.getSymA() && "Relocation must reference a symbol!");
const MCSymbol &Symbol = Target.getSymA()->getSymbol();

26
contrib/llvm/lib/Support/Triple.cpp
View File

@ -246,13 +246,21 @@ static Triple::ArchType parseARMArch(StringRef ArchName) {
if (ArchName.startswith("armv")) {
offset = 3;
arch = Triple::arm;
if (ArchName.endswith("eb")) {
arch = Triple::armeb;
ArchName = ArchName.substr(0, ArchName.size() - 2);
} else
arch = Triple::arm;
} else if (ArchName.startswith("armebv")) {
offset = 5;
arch = Triple::armeb;
} else if (ArchName.startswith("thumbv")) {
offset = 5;
arch = Triple::thumb;
if (ArchName.endswith("eb")) {
arch = Triple::thumbeb;
ArchName = ArchName.substr(0, ArchName.size() - 2);
} else
arch = Triple::thumb;
} else if (ArchName.startswith("thumbebv")) {
offset = 7;
arch = Triple::thumbeb;
@ -271,6 +279,8 @@ static Triple::ArchType parseARMArch(StringRef ArchName) {
}
static Triple::ArchType parseArch(StringRef ArchName) {
Triple::ArchType ARMArch(parseARMArch(ArchName));
return StringSwitch<Triple::ArchType>(ArchName)
.Cases("i386", "i486", "i586", "i686", Triple::x86)
// FIXME: Do we need to support these?
@ -280,9 +290,10 @@ static Triple::ArchType parseArch(StringRef ArchName) {
.Cases("powerpc64", "ppu", Triple::ppc64)
.Case("powerpc64le", Triple::ppc64le)
.Case("xscale", Triple::arm)
.StartsWith("arm", parseARMArch(ArchName))
.StartsWith("thumb", parseARMArch(ArchName))
.StartsWith("aarch64", parseARMArch(ArchName))
.Case("xscaleeb", Triple::armeb)
.StartsWith("arm", ARMArch)
.StartsWith("thumb", ARMArch)
.StartsWith("aarch64", ARMArch)
.Case("msp430", Triple::msp430)
.Cases("mips", "mipseb", "mipsallegrex", Triple::mips)
.Cases("mipsel", "mipsallegrexel", Triple::mipsel)
@ -379,6 +390,9 @@ static Triple::ObjectFormatType parseFormat(StringRef EnvironmentName) {
}
static Triple::SubArchType parseSubArch(StringRef SubArchName) {
if (SubArchName.endswith("eb"))
SubArchName = SubArchName.substr(0, SubArchName.size() - 2);
return StringSwitch<Triple::SubArchType>(SubArchName)
.EndsWith("v8", Triple::ARMSubArch_v8)
.EndsWith("v8a", Triple::ARMSubArch_v8)
@ -1022,6 +1036,8 @@ const char *Triple::getARMCPUForArch(StringRef MArch) const {
offset = 5;
if (offset != StringRef::npos && MArch.substr(offset, 2) == "eb")
offset += 2;
if (MArch.endswith("eb"))
MArch = MArch.substr(0, MArch.size() - 2);
if (offset != StringRef::npos)
result = llvm::StringSwitch<const char *>(MArch.substr(offset))
.Cases("v2", "v2a", "arm2")

39
contrib/llvm/lib/Target/AArch64/AArch64CallingConvention.td
View File

@ -204,6 +204,44 @@ def RetCC_AArch64_WebKit_JS : CallingConv<[
[Q0, Q1, Q2, Q3, Q4, Q5, Q6, Q7]>>
]>;
//===----------------------------------------------------------------------===//
// ARM64 Calling Convention for GHC
//===----------------------------------------------------------------------===//
// This calling convention is specific to the Glasgow Haskell Compiler.
// The only documentation is the GHC source code, specifically the C header
// file:
//
// https://github.com/ghc/ghc/blob/master/includes/stg/MachRegs.h
//
// which defines the registers for the Spineless Tagless G-Machine (STG) that
// GHC uses to implement lazy evaluation. The generic STG machine has a set of
// registers which are mapped to appropriate set of architecture specific
// registers for each CPU architecture.
//
// The STG Machine is documented here:
//
// https://ghc.haskell.org/trac/ghc/wiki/Commentary/Compiler/GeneratedCode
//
// The AArch64 register mapping is under the heading "The ARMv8/AArch64 ABI
// register mapping".
def CC_AArch64_GHC : CallingConv<[
// Handle all vector types as either f64 or v2f64.
CCIfType<[v1i64, v2i32, v4i16, v8i8, v2f32], CCBitConvertToType<f64>>,
CCIfType<[v2i64, v4i32, v8i16, v16i8, v4f32, f128], CCBitConvertToType<v2f64>>,
CCIfType<[v2f64], CCAssignToReg<[Q4, Q5]>>,
CCIfType<[f32], CCAssignToReg<[S8, S9, S10, S11]>>,
CCIfType<[f64], CCAssignToReg<[D12, D13, D14, D15]>>,
// Promote i8/i16/i32 arguments to i64.
CCIfType<[i8, i16, i32], CCPromoteToType<i64>>,
// Pass in STG registers: Base, Sp, Hp, R1, R2, R3, R4, R5, R6, SpLim
CCIfType<[i64], CCAssignToReg<[X19, X20, X21, X22, X23, X24, X25, X26, X27, X28]>>
]>;
// FIXME: LR is only callee-saved in the sense that *we* preserve it and are
// presumably a callee to someone. External functions may not do so, but this
// is currently safe since BL has LR as an implicit-def and what happens after a
@ -249,3 +287,4 @@ def CSR_AArch64_AllRegs
(sequence "S%u", 0, 31), (sequence "D%u", 0, 31),
(sequence "Q%u", 0, 31))>;
def CSR_AArch64_NoRegs : CalleeSavedRegs<(add)>;

2
contrib/llvm/lib/Target/AArch64/AArch64FastISel.cpp
View File

@ -302,6 +302,8 @@ static unsigned getImplicitScaleFactor(MVT VT) {
CCAssignFn *AArch64FastISel::CCAssignFnForCall(CallingConv::ID CC) const {
if (CC == CallingConv::WebKit_JS)
return CC_AArch64_WebKit_JS;
if (CC == CallingConv::GHC)
return CC_AArch64_GHC;
return Subtarget->isTargetDarwin() ? CC_AArch64_DarwinPCS : CC_AArch64_AAPCS;
}

10
contrib/llvm/lib/Target/AArch64/AArch64FrameLowering.cpp
View File

@ -215,6 +215,11 @@ void AArch64FrameLowering::emitPrologue(MachineFunction &MF) const {
bool HasFP = hasFP(MF);
DebugLoc DL = MBB.findDebugLoc(MBBI);
// All calls are tail calls in GHC calling conv, and functions have no
// prologue/epilogue.
if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
return;
int NumBytes = (int)MFI->getStackSize();
if (!AFI->hasStackFrame()) {
assert(!HasFP && "unexpected function without stack frame but with FP");
@ -451,6 +456,11 @@ void AArch64FrameLowering::emitEpilogue(MachineFunction &MF,
int NumBytes = MFI->getStackSize();
const AArch64FunctionInfo *AFI = MF.getInfo<AArch64FunctionInfo>();
// All calls are tail calls in GHC calling conv, and functions have no
// prologue/epilogue.
if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
return;
// Initial and residual are named for consitency with the prologue. Note that
// in the epilogue, the residual adjustment is executed first.
uint64_t ArgumentPopSize = 0;

2
contrib/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
View File

@ -1990,6 +1990,8 @@ CCAssignFn *AArch64TargetLowering::CCAssignFnForCall(CallingConv::ID CC,
llvm_unreachable("Unsupported calling convention.");
case CallingConv::WebKit_JS:
return CC_AArch64_WebKit_JS;
case CallingConv::GHC:
return CC_AArch64_GHC;
case CallingConv::C:
case CallingConv::Fast:
if (!Subtarget->isTargetDarwin())

10
contrib/llvm/lib/Target/AArch64/AArch64RegisterInfo.cpp
View File

@ -44,6 +44,10 @@ AArch64RegisterInfo::AArch64RegisterInfo(const AArch64InstrInfo *tii,
const MCPhysReg *
AArch64RegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
assert(MF && "Invalid MachineFunction pointer.");
if (MF->getFunction()->getCallingConv() == CallingConv::GHC)
// GHC set of callee saved regs is empty as all those regs are
// used for passing STG regs around
return CSR_AArch64_NoRegs_SaveList;
if (MF->getFunction()->getCallingConv() == CallingConv::AnyReg)
return CSR_AArch64_AllRegs_SaveList;
else
@ -52,6 +56,9 @@ AArch64RegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
const uint32_t *
AArch64RegisterInfo::getCallPreservedMask(CallingConv::ID CC) const {
if (CC == CallingConv::GHC)
// This is academic becase all GHC calls are (supposed to be) tail calls
return CSR_AArch64_NoRegs_RegMask;
if (CC == CallingConv::AnyReg)
return CSR_AArch64_AllRegs_RegMask;
else
@ -67,7 +74,7 @@ const uint32_t *AArch64RegisterInfo::getTLSCallPreservedMask() const {
}
const uint32_t *
AArch64RegisterInfo::getThisReturnPreservedMask(CallingConv::ID) const {
AArch64RegisterInfo::getThisReturnPreservedMask(CallingConv::ID CC) const {
// This should return a register mask that is the same as that returned by
// getCallPreservedMask but that additionally preserves the register used for
// the first i64 argument (which must also be the register used to return a
@ -75,6 +82,7 @@ AArch64RegisterInfo::getThisReturnPreservedMask(CallingConv::ID) const {
//
// In case that the calling convention does not use the same register for
// both, the function should return NULL (does not currently apply)
assert(CC != CallingConv::GHC && "should not be GHC calling convention.");
return CSR_AArch64_AAPCS_ThisReturn_RegMask;
}

36
contrib/llvm/lib/Target/AArch64/MCTargetDesc/AArch64AsmBackend.cpp
View File

@ -317,6 +317,42 @@ public:
MachO::CPU_SUBTYPE_ARM64_ALL);
}
bool doesSectionRequireSymbols(const MCSection &Section) const override {
// Any section for which the linker breaks things into atoms needs to
// preserve symbols, including assembler local symbols, to identify
// those atoms. These sections are:
// Sections of type:
//
// S_CSTRING_LITERALS (e.g. __cstring)
// S_LITERAL_POINTERS (e.g. objc selector pointers)
// S_16BYTE_LITERALS, S_8BYTE_LITERALS, S_4BYTE_LITERALS
//
// Sections named:
//
// __TEXT,__eh_frame
// __TEXT,__ustring
// __DATA,__cfstring
// __DATA,__objc_classrefs
// __DATA,__objc_catlist
//
// FIXME: It would be better if the compiler used actual linker local
// symbols for each of these sections rather than preserving what
// are ostensibly assembler local symbols.
const MCSectionMachO &SMO = static_cast<const MCSectionMachO &>(Section);
return (SMO.getType() == MachO::S_CSTRING_LITERALS ||
SMO.getType() == MachO::S_4BYTE_LITERALS ||
SMO.getType() == MachO::S_8BYTE_LITERALS ||
SMO.getType() == MachO::S_16BYTE_LITERALS ||
SMO.getType() == MachO::S_LITERAL_POINTERS ||
(SMO.getSegmentName() == "__TEXT" &&
(SMO.getSectionName() == "__eh_frame" ||
SMO.getSectionName() == "__ustring")) ||
(SMO.getSegmentName() == "__DATA" &&
(SMO.getSectionName() == "__cfstring" ||
SMO.getSectionName() == "__objc_classrefs" ||
SMO.getSectionName() == "__objc_catlist")));
}
/// \brief Generate the compact unwind encoding from the CFI directives.
uint32_t generateCompactUnwindEncoding(
ArrayRef<MCCFIInstruction> Instrs) const override {

78
contrib/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
View File

@ -10,7 +10,6 @@
#include "MCTargetDesc/AArch64FixupKinds.h"
#include "MCTargetDesc/AArch64MCTargetDesc.h"
#include "llvm/ADT/Twine.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCAsmLayout.h"
#include "llvm/MC/MCAssembler.h"
#include "llvm/MC/MCContext.h"
@ -34,7 +33,7 @@ public:
: MCMachObjectTargetWriter(true /* is64Bit */, CPUType, CPUSubtype,
/*UseAggressiveSymbolFolding=*/true) {}
void RecordRelocation(MachObjectWriter *Writer, MCAssembler &Asm,
void RecordRelocation(MachObjectWriter *Writer, const MCAssembler &Asm,
const MCAsmLayout &Layout, const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) override;
@ -113,25 +112,8 @@ bool AArch64MachObjectWriter::getAArch64FixupKindMachOInfo(
}
}
static bool canUseLocalRelocation(const MCSectionMachO &Section,
const MCSymbol &Symbol, unsigned Log2Size) {
// Debug info sections can use local relocations.
if (Section.hasAttribute(MachO::S_ATTR_DEBUG))
return true;
// Otherwise, only pointer sized relocations are supported.
if (Log2Size != 3)
return false;
// But only if they don't point to a cstring.
if (!Symbol.isInSection())
return true;
const MCSectionMachO &RefSec = cast<MCSectionMachO>(Symbol.getSection());
return RefSec.getType() != MachO::S_CSTRING_LITERALS;
}
void AArch64MachObjectWriter::RecordRelocation(
MachObjectWriter *Writer, MCAssembler &Asm, const MCAsmLayout &Layout,
MachObjectWriter *Writer, const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) {
unsigned IsPCRel = Writer->isFixupKindPCRel(Asm, Fixup.getKind());
@ -141,9 +123,9 @@ void AArch64MachObjectWriter::RecordRelocation(
unsigned Log2Size = 0;
int64_t Value = 0;
unsigned Index = 0;
unsigned IsExtern = 0;
unsigned Type = 0;
unsigned Kind = Fixup.getKind();
const MCSymbolData *RelSymbol = nullptr;
FixupOffset += Fixup.getOffset();
@ -189,8 +171,10 @@ void AArch64MachObjectWriter::RecordRelocation(
// FIXME: Should this always be extern?
// SymbolNum of 0 indicates the absolute section.
Type = MachO::ARM64_RELOC_UNSIGNED;
Index = 0;
if (IsPCRel) {
IsExtern = 1;
Asm.getContext().FatalError(Fixup.getLoc(),
"PC relative absolute relocation!");
@ -214,12 +198,15 @@ void AArch64MachObjectWriter::RecordRelocation(
Layout.getSymbolOffset(&B_SD) ==
Layout.getFragmentOffset(Fragment) + Fixup.getOffset()) {
// SymB is the PC, so use a PC-rel pointer-to-GOT relocation.
Index = A_Base->getIndex();
IsExtern = 1;
Type = MachO::ARM64_RELOC_POINTER_TO_GOT;
IsPCRel = 1;
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 = (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
Writer->addRelocation(A_Base, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) |
(IsExtern << 27) | (Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
return;
} else if (Target.getSymA()->getKind() != MCSymbolRefExpr::VK_None ||
Target.getSymB()->getKind() != MCSymbolRefExpr::VK_None)
@ -265,30 +252,25 @@ void AArch64MachObjectWriter::RecordRelocation(
? 0
: Writer->getSymbolAddress(B_Base, Layout));
Index = A_Base->getIndex();
IsExtern = 1;
Type = MachO::ARM64_RELOC_UNSIGNED;
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 = (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
Writer->addRelocation(A_Base, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) |
(IsExtern << 27) | (Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
RelSymbol = B_Base;
Index = B_Base->getIndex();
IsExtern = 1;
Type = MachO::ARM64_RELOC_SUBTRACTOR;
} else { // A + constant
const MCSymbol *Symbol = &Target.getSymA()->getSymbol();
const MCSectionMachO &Section = static_cast<const MCSectionMachO &>(
Fragment->getParent()->getSection());
bool CanUseLocalRelocation =
canUseLocalRelocation(Section, *Symbol, Log2Size);
if (Symbol->isTemporary() && (Value || !CanUseLocalRelocation)) {
const MCSection &Sec = Symbol->getSection();
if (!Asm.getContext().getAsmInfo()->isSectionAtomizableBySymbols(Sec))
Asm.addLocalUsedInReloc(*Symbol);
}
const MCSymbolData &SD = Asm.getSymbolData(*Symbol);
const MCSymbolData *Base = Asm.getAtom(&SD);
const MCSectionMachO &Section = static_cast<const MCSectionMachO &>(
Fragment->getParent()->getSection());
// If the symbol is a variable and we weren't able to get a Base for it
// (i.e., it's not in the symbol table associated with a section) resolve
@ -328,13 +310,16 @@ void AArch64MachObjectWriter::RecordRelocation(
// sections, and for pointer-sized relocations (.quad), we allow section
// relocations. It's code sections that run into trouble.
if (Base) {
RelSymbol = Base;
Index = Base->getIndex();
IsExtern = 1;
// Add the local offset, if needed.
if (Base != &SD)
Value += Layout.getSymbolOffset(&SD) - Layout.getSymbolOffset(Base);
} else if (Symbol->isInSection()) {
if (!CanUseLocalRelocation)
// Pointer-sized relocations can use a local relocation. Otherwise,
// we have to be in a debug info section.
if (!Section.hasAttribute(MachO::S_ATTR_DEBUG) && Log2Size != 3)
Asm.getContext().FatalError(
Fixup.getLoc(),
"unsupported relocation of local symbol '" + Symbol->getName() +
@ -344,6 +329,7 @@ void AArch64MachObjectWriter::RecordRelocation(
const MCSectionData &SymSD =
Asm.getSectionData(SD.getSymbol().getSection());
Index = SymSD.getOrdinal() + 1;
IsExtern = 0;
Value += Writer->getSymbolAddress(&SD, Layout);
if (IsPCRel)
@ -376,16 +362,16 @@ void AArch64MachObjectWriter::RecordRelocation(
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 =
(Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
Writer->addRelocation(RelSymbol, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) |
(IsExtern << 27) | (Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
// Now set up the Addend relocation.
Type = MachO::ARM64_RELOC_ADDEND;
Index = Value;
RelSymbol = nullptr;
IsPCRel = 0;
Log2Size = 2;
IsExtern = 0;
// Put zero into the instruction itself. The addend is in the relocation.
Value = 0;
@ -397,9 +383,9 @@ void AArch64MachObjectWriter::RecordRelocation(
// struct relocation_info (8 bytes)
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 =
(Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
Writer->addRelocation(RelSymbol, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) | (IsPCRel << 24) | (Log2Size << 25) |
(IsExtern << 27) | (Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
}
MCObjectWriter *llvm::createAArch64MachObjectWriter(raw_ostream &OS,

17
contrib/llvm/lib/Target/ARM/ARMLoadStoreOptimizer.cpp
View File

@ -567,10 +567,21 @@ ARMLoadStoreOpt::MergeOps(MachineBasicBlock &MBB,
// MOV NewBase, Base
// ADDS NewBase, #imm8.
if (Base != NewBase && Offset >= 8) {
const ARMSubtarget &Subtarget = MBB.getParent()->getTarget()
.getSubtarget<ARMSubtarget>();
// Need to insert a MOV to the new base first.
BuildMI(MBB, MBBI, dl, TII->get(ARM::tMOVr), NewBase)
.addReg(Base, getKillRegState(BaseKill))
.addImm(Pred).addReg(PredReg);
if (isARMLowRegister(NewBase) && isARMLowRegister(Base) &&
!Subtarget.hasV6Ops()) {
// thumbv4t doesn't have lo->lo copies, and we can't predicate tMOVSr
if (Pred != ARMCC::AL)
return false;
BuildMI(MBB, MBBI, dl, TII->get(ARM::tMOVSr), NewBase)
.addReg(Base, getKillRegState(BaseKill));
} else
BuildMI(MBB, MBBI, dl, TII->get(ARM::tMOVr), NewBase)
.addReg(Base, getKillRegState(BaseKill))
.addImm(Pred).addReg(PredReg);
// Set up BaseKill and Base correctly to insert the ADDS/SUBS below.
Base = NewBase;
BaseKill = false;

58
contrib/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
View File

@ -9191,27 +9191,39 @@ bool ARMAsmParser::parseDirectiveCPU(SMLoc L) {
// FIXME: This is duplicated in getARMFPUFeatures() in
// tools/clang/lib/Driver/Tools.cpp
static const struct {
const unsigned Fpu;
const unsigned ID;
const uint64_t Enabled;
const uint64_t Disabled;
} Fpus[] = {
{ARM::VFP, ARM::FeatureVFP2, ARM::FeatureNEON},
{ARM::VFPV2, ARM::FeatureVFP2, ARM::FeatureNEON},
{ARM::VFPV3, ARM::FeatureVFP3, ARM::FeatureNEON},
{ARM::VFPV3_D16, ARM::FeatureVFP3 | ARM::FeatureD16, ARM::FeatureNEON},
{ARM::VFPV4, ARM::FeatureVFP4, ARM::FeatureNEON},
{ARM::VFPV4_D16, ARM::FeatureVFP4 | ARM::FeatureD16, ARM::FeatureNEON},
{ARM::FPV5_D16, ARM::FeatureFPARMv8 | ARM::FeatureD16,
ARM::FeatureNEON | ARM::FeatureCrypto},
{ARM::FP_ARMV8, ARM::FeatureFPARMv8,
ARM::FeatureNEON | ARM::FeatureCrypto},
{ARM::NEON, ARM::FeatureNEON, 0},
{ARM::NEON_VFPV4, ARM::FeatureVFP4 | ARM::FeatureNEON, 0},
{ARM::NEON_FP_ARMV8, ARM::FeatureFPARMv8 | ARM::FeatureNEON,
ARM::FeatureCrypto},
{ARM::CRYPTO_NEON_FP_ARMV8,
ARM::FeatureFPARMv8 | ARM::FeatureNEON | ARM::FeatureCrypto, 0},
{ARM::SOFTVFP, 0, 0},
} FPUs[] = {
{ARM::VFP, ARM::FeatureVFP2, ARM::FeatureNEON},
{ARM::VFPV2, ARM::FeatureVFP2, ARM::FeatureNEON},
{ARM::VFPV3, ARM::FeatureVFP2 | ARM::FeatureVFP3, ARM::FeatureNEON},
{ARM::VFPV3_D16, ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureD16,
ARM::FeatureNEON},
{ARM::VFPV4, ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4,
ARM::FeatureNEON},
{ARM::VFPV4_D16,
ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4 | ARM::FeatureD16,
ARM::FeatureNEON},
{ARM::FPV5_D16, ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4 |
ARM::FeatureFPARMv8 | ARM::FeatureD16,
ARM::FeatureNEON | ARM::FeatureCrypto},
{ARM::FP_ARMV8, ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4 |
ARM::FeatureFPARMv8,
ARM::FeatureNEON | ARM::FeatureCrypto},
{ARM::NEON, ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureNEON, 0},
{ARM::NEON_VFPV4,
ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4 | ARM::FeatureNEON,
0},
{ARM::NEON_FP_ARMV8,
ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4 |
ARM::FeatureFPARMv8 | ARM::FeatureNEON,
ARM::FeatureCrypto},
{ARM::CRYPTO_NEON_FP_ARMV8,
ARM::FeatureVFP2 | ARM::FeatureVFP3 | ARM::FeatureVFP4 |
ARM::FeatureFPARMv8 | ARM::FeatureNEON | ARM::FeatureCrypto,
0},
{ARM::SOFTVFP, 0, 0},
};
/// parseDirectiveFPU
@ -9229,14 +9241,14 @@ bool ARMAsmParser::parseDirectiveFPU(SMLoc L) {
return false;
}
for (const auto &Fpu : Fpus) {
if (Fpu.Fpu != ID)
for (const auto &Entry : FPUs) {
if (Entry.ID != ID)
continue;
// Need to toggle features that should be on but are off and that
// should off but are on.
uint64_t Toggle = (Fpu.Enabled & ~STI.getFeatureBits()) |
(Fpu.Disabled & STI.getFeatureBits());
uint64_t Toggle = (Entry.Enabled & ~STI.getFeatureBits()) |
(Entry.Disabled & STI.getFeatureBits());
setAvailableFeatures(ComputeAvailableFeatures(STI.ToggleFeature(Toggle)));
break;
}

37
contrib/llvm/lib/Target/ARM/MCTargetDesc/ARMMachObjectWriter.cpp
View File

@ -54,10 +54,10 @@ public:
: MCMachObjectTargetWriter(Is64Bit, CPUType, CPUSubtype,
/*UseAggressiveSymbolFolding=*/true) {}
void RecordRelocation(MachObjectWriter *Writer, MCAssembler &Asm,
const MCAsmLayout &Layout, const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) override;
void RecordRelocation(MachObjectWriter *Writer,
const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, uint64_t &FixedValue) override;
};
}
@ -232,7 +232,7 @@ RecordARMScatteredHalfRelocation(MachObjectWriter *Writer,
(IsPCRel << 30) |
MachO::R_SCATTERED);
MRE.r_word1 = Value2;
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
}
MachO::any_relocation_info MRE;
@ -243,7 +243,7 @@ RecordARMScatteredHalfRelocation(MachObjectWriter *Writer,
(IsPCRel << 30) |
MachO::R_SCATTERED);
MRE.r_word1 = Value;
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
}
void ARMMachObjectWriter::RecordARMScatteredRelocation(MachObjectWriter *Writer,
@ -297,7 +297,7 @@ void ARMMachObjectWriter::RecordARMScatteredRelocation(MachObjectWriter *Writer,
(IsPCRel << 30) |
MachO::R_SCATTERED);
MRE.r_word1 = Value2;
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
}
MachO::any_relocation_info MRE;
@ -307,7 +307,7 @@ void ARMMachObjectWriter::RecordARMScatteredRelocation(MachObjectWriter *Writer,
(IsPCRel << 30) |
MachO::R_SCATTERED);
MRE.r_word1 = Value;
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
}
bool ARMMachObjectWriter::requiresExternRelocation(MachObjectWriter *Writer,
@ -351,10 +351,11 @@ bool ARMMachObjectWriter::requiresExternRelocation(MachObjectWriter *Writer,
}
void ARMMachObjectWriter::RecordRelocation(MachObjectWriter *Writer,
MCAssembler &Asm,
const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
const MCFixup &Fixup,
MCValue Target,
uint64_t &FixedValue) {
unsigned IsPCRel = Writer->isFixupKindPCRel(Asm, Fixup.getKind());
unsigned Log2Size;
@ -400,8 +401,8 @@ void ARMMachObjectWriter::RecordRelocation(MachObjectWriter *Writer,
// See <reloc.h>.
uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
unsigned Index = 0;
unsigned IsExtern = 0;
unsigned Type = 0;
const MCSymbolData *RelSymbol = nullptr;
if (Target.isAbsolute()) { // constant
// FIXME!
@ -421,7 +422,8 @@ void ARMMachObjectWriter::RecordRelocation(MachObjectWriter *Writer,
// Check whether we need an external or internal relocation.
if (requiresExternRelocation(Writer, Asm, *Fragment, RelocType, SD,
FixedValue)) {
RelSymbol = SD;
IsExtern = 1;
Index = SD->getIndex();
// For external relocations, make sure to offset the fixup value to
// compensate for the addend of the symbol address, if it was
@ -445,8 +447,11 @@ void ARMMachObjectWriter::RecordRelocation(MachObjectWriter *Writer,
// struct relocation_info (8 bytes)
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 =
(Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
MRE.r_word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
(IsExtern << 27) |
(Type << 28));
// Even when it's not a scattered relocation, movw/movt always uses
// a PAIR relocation.
@ -471,10 +476,10 @@ void ARMMachObjectWriter::RecordRelocation(MachObjectWriter *Writer,
(Log2Size << 25) |
(MachO::ARM_RELOC_PAIR << 28));
Writer->addRelocation(nullptr, Fragment->getParent(), MREPair);
Writer->addRelocation(Fragment->getParent(), MREPair);
}
Writer->addRelocation(RelSymbol, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
}
MCObjectWriter *llvm::createARMMachObjectWriter(raw_ostream &OS,

3
contrib/llvm/lib/Target/Hexagon/HexagonRemoveSZExtArgs.cpp
View File

@ -15,6 +15,7 @@
#include "Hexagon.h"
#include "HexagonTargetMachine.h"
#include "llvm/CodeGen/MachineFunctionAnalysis.h"
#include "llvm/CodeGen/StackProtector.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/Instructions.h"
#include "llvm/Pass.h"
@ -42,7 +43,7 @@ namespace {
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<MachineFunctionAnalysis>();
AU.addPreserved<MachineFunctionAnalysis>();
AU.addPreserved("stack-protector");
AU.addPreserved<StackProtector>();
FunctionPass::getAnalysisUsage(AU);
}
};

4
contrib/llvm/lib/Target/Mips/Mips16ISelLowering.cpp
View File

@ -497,14 +497,14 @@ getOpndList(SmallVectorImpl<SDValue> &Ops,
SDValue JumpTarget = Callee;
// T9 should contain the address of the callee function if
// -reloction-model=pic or it is an indirect call.
// -relocation-model=pic or it is an indirect call.
if (IsPICCall || !GlobalOrExternal) {
unsigned V0Reg = Mips::V0;
if (NeedMips16Helper) {
RegsToPass.push_front(std::make_pair(V0Reg, Callee));
JumpTarget = DAG.getExternalSymbol(Mips16HelperFunction, getPointerTy());
ExternalSymbolSDNode *S = cast<ExternalSymbolSDNode>(JumpTarget);
JumpTarget = getAddrGlobal(S, JumpTarget.getValueType(), DAG,
JumpTarget = getAddrGlobal(S, CLI.DL, JumpTarget.getValueType(), DAG,
MipsII::MO_GOT, Chain,
FuncInfo->callPtrInfo(S->getSymbol()));
} else

4
contrib/llvm/lib/Target/Mips/Mips32r6InstrInfo.td
View File

@ -756,7 +756,7 @@ def : MipsPat<(setge f32:$lhs, f32:$rhs), (CMP_LT_S f32:$rhs, f32:$lhs)>,
ISA_MIPS32R6;
def : MipsPat<(setlt f32:$lhs, f32:$rhs), (CMP_LT_S f32:$lhs, f32:$rhs)>,
ISA_MIPS32R6;
def : MipsPat<(setlt f32:$lhs, f32:$rhs), (CMP_LE_S f32:$lhs, f32:$rhs)>,
def : MipsPat<(setle f32:$lhs, f32:$rhs), (CMP_LE_S f32:$lhs, f32:$rhs)>,
ISA_MIPS32R6;
def : MipsPat<(setne f32:$lhs, f32:$rhs),
(NOR (CMP_EQ_S f32:$lhs, f32:$rhs), ZERO)>, ISA_MIPS32R6;
@ -776,7 +776,7 @@ def : MipsPat<(setge f64:$lhs, f64:$rhs), (CMP_LT_D f64:$rhs, f64:$lhs)>,
ISA_MIPS32R6;
def : MipsPat<(setlt f64:$lhs, f64:$rhs), (CMP_LT_D f64:$lhs, f64:$rhs)>,
ISA_MIPS32R6;
def : MipsPat<(setlt f64:$lhs, f64:$rhs), (CMP_LE_D f64:$lhs, f64:$rhs)>,
def : MipsPat<(setle f64:$lhs, f64:$rhs), (CMP_LE_D f64:$lhs, f64:$rhs)>,
ISA_MIPS32R6;
def : MipsPat<(setne f64:$lhs, f64:$rhs),
(NOR (CMP_EQ_D f64:$lhs, f64:$rhs), ZERO)>, ISA_MIPS32R6;

38
contrib/llvm/lib/Target/Mips/MipsISelLowering.cpp
View File

@ -1613,22 +1613,22 @@ SDValue MipsTargetLowering::lowerGlobalAddress(SDValue Op,
if (TLOF.IsGlobalInSmallSection(GV, getTargetMachine()))
// %gp_rel relocation
return getAddrGPRel(N, Ty, DAG);
return getAddrGPRel(N, SDLoc(N), Ty, DAG);
// %hi/%lo relocation
return getAddrNonPIC(N, Ty, DAG);
return getAddrNonPIC(N, SDLoc(N), Ty, DAG);
}
if (GV->hasInternalLinkage() || (GV->hasLocalLinkage() && !isa<Function>(GV)))
return getAddrLocal(N, Ty, DAG,
return getAddrLocal(N, SDLoc(N), Ty, DAG,
Subtarget.isABI_N32() || Subtarget.isABI_N64());
if (LargeGOT)
return getAddrGlobalLargeGOT(N, Ty, DAG, MipsII::MO_GOT_HI16,
return getAddrGlobalLargeGOT(N, SDLoc(N), Ty, DAG, MipsII::MO_GOT_HI16,
MipsII::MO_GOT_LO16, DAG.getEntryNode(),
MachinePointerInfo::getGOT());
return getAddrGlobal(N, Ty, DAG,
return getAddrGlobal(N, SDLoc(N), Ty, DAG,
(Subtarget.isABI_N32() || Subtarget.isABI_N64())
? MipsII::MO_GOT_DISP
: MipsII::MO_GOT16,
@ -1642,9 +1642,9 @@ SDValue MipsTargetLowering::lowerBlockAddress(SDValue Op,
if (getTargetMachine().getRelocationModel() != Reloc::PIC_ &&
!Subtarget.isABI_N64())
return getAddrNonPIC(N, Ty, DAG);
return getAddrNonPIC(N, SDLoc(N), Ty, DAG);
return getAddrLocal(N, Ty, DAG,
return getAddrLocal(N, SDLoc(N), Ty, DAG,
Subtarget.isABI_N32() || Subtarget.isABI_N64());
}
@ -1735,9 +1735,9 @@ lowerJumpTable(SDValue Op, SelectionDAG &DAG) const
if (getTargetMachine().getRelocationModel() != Reloc::PIC_ &&
!Subtarget.isABI_N64())
return getAddrNonPIC(N, Ty, DAG);
return getAddrNonPIC(N, SDLoc(N), Ty, DAG);
return getAddrLocal(N, Ty, DAG,
return getAddrLocal(N, SDLoc(N), Ty, DAG,
Subtarget.isABI_N32() || Subtarget.isABI_N64());
}
@ -1754,12 +1754,12 @@ lowerConstantPool(SDValue Op, SelectionDAG &DAG) const
if (TLOF.IsConstantInSmallSection(N->getConstVal(), getTargetMachine()))
// %gp_rel relocation
return getAddrGPRel(N, Ty, DAG);
return getAddrGPRel(N, SDLoc(N), Ty, DAG);
return getAddrNonPIC(N, Ty, DAG);
return getAddrNonPIC(N, SDLoc(N), Ty, DAG);
}
return getAddrLocal(N, Ty, DAG,
return getAddrLocal(N, SDLoc(N), Ty, DAG,
Subtarget.isABI_N32() || Subtarget.isABI_N64());
}
@ -2681,15 +2681,15 @@ MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
InternalLinkage = Val->hasInternalLinkage();
if (InternalLinkage)
Callee = getAddrLocal(G, Ty, DAG,
Callee = getAddrLocal(G, DL, Ty, DAG,
Subtarget.isABI_N32() || Subtarget.isABI_N64());
else if (LargeGOT) {
Callee = getAddrGlobalLargeGOT(G, Ty, DAG, MipsII::MO_CALL_HI16,
Callee = getAddrGlobalLargeGOT(G, DL, Ty, DAG, MipsII::MO_CALL_HI16,
MipsII::MO_CALL_LO16, Chain,
FuncInfo->callPtrInfo(Val));
IsCallReloc = true;
} else {
Callee = getAddrGlobal(G, Ty, DAG, MipsII::MO_GOT_CALL, Chain,
Callee = getAddrGlobal(G, DL, Ty, DAG, MipsII::MO_GOT_CALL, Chain,
FuncInfo->callPtrInfo(Val));
IsCallReloc = true;
}
@ -2702,15 +2702,15 @@ MipsTargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
const char *Sym = S->getSymbol();
if (!Subtarget.isABI_N64() && !IsPIC) // !N64 && static
Callee = DAG.getTargetExternalSymbol(Sym, getPointerTy(),
MipsII::MO_NO_FLAG);
Callee =
DAG.getTargetExternalSymbol(Sym, getPointerTy(), MipsII::MO_NO_FLAG);
else if (LargeGOT) {
Callee = getAddrGlobalLargeGOT(S, Ty, DAG, MipsII::MO_CALL_HI16,
Callee = getAddrGlobalLargeGOT(S, DL, Ty, DAG, MipsII::MO_CALL_HI16,
MipsII::MO_CALL_LO16, Chain,
FuncInfo->callPtrInfo(Sym));
IsCallReloc = true;
} else { // N64 || PIC
Callee = getAddrGlobal(S, Ty, DAG, MipsII::MO_GOT_CALL, Chain,
Callee = getAddrGlobal(S, DL, Ty, DAG, MipsII::MO_GOT_CALL, Chain,
FuncInfo->callPtrInfo(Sym));
IsCallReloc = true;
}

32
contrib/llvm/lib/Target/Mips/MipsISelLowering.h
View File

@ -272,9 +272,8 @@ namespace llvm {
//
// (add (load (wrapper $gp, %got(sym)), %lo(sym))
template <class NodeTy>
SDValue getAddrLocal(NodeTy *N, EVT Ty, SelectionDAG &DAG,
SDValue getAddrLocal(NodeTy *N, SDLoc DL, EVT Ty, SelectionDAG &DAG,
bool IsN32OrN64) const {
SDLoc DL(N);
unsigned GOTFlag = IsN32OrN64 ? MipsII::MO_GOT_PAGE : MipsII::MO_GOT;
SDValue GOT = DAG.getNode(MipsISD::Wrapper, DL, Ty, getGlobalReg(DAG, Ty),
getTargetNode(N, Ty, DAG, GOTFlag));
@ -291,11 +290,10 @@ namespace llvm {
// computing a global symbol's address:
//
// (load (wrapper $gp, %got(sym)))
template<class NodeTy>
SDValue getAddrGlobal(NodeTy *N, EVT Ty, SelectionDAG &DAG,
template <class NodeTy>
SDValue getAddrGlobal(NodeTy *N, SDLoc DL, EVT Ty, SelectionDAG &DAG,
unsigned Flag, SDValue Chain,
const MachinePointerInfo &PtrInfo) const {
SDLoc DL(N);
SDValue Tgt = DAG.getNode(MipsISD::Wrapper, DL, Ty, getGlobalReg(DAG, Ty),
getTargetNode(N, Ty, DAG, Flag));
return DAG.getLoad(Ty, DL, Chain, Tgt, PtrInfo, false, false, false, 0);
@ -305,14 +303,13 @@ namespace llvm {
// computing a global symbol's address in large-GOT mode:
//
// (load (wrapper (add %hi(sym), $gp), %lo(sym)))
template<class NodeTy>
SDValue getAddrGlobalLargeGOT(NodeTy *N, EVT Ty, SelectionDAG &DAG,
unsigned HiFlag, unsigned LoFlag,
SDValue Chain,
template <class NodeTy>
SDValue getAddrGlobalLargeGOT(NodeTy *N, SDLoc DL, EVT Ty,
SelectionDAG &DAG, unsigned HiFlag,
unsigned LoFlag, SDValue Chain,
const MachinePointerInfo &PtrInfo) const {
SDLoc DL(N);
SDValue Hi = DAG.getNode(MipsISD::Hi, DL, Ty,
getTargetNode(N, Ty, DAG, HiFlag));
SDValue Hi =
DAG.getNode(MipsISD::Hi, DL, Ty, getTargetNode(N, Ty, DAG, HiFlag));
Hi = DAG.getNode(ISD::ADD, DL, Ty, Hi, getGlobalReg(DAG, Ty));
SDValue Wrapper = DAG.getNode(MipsISD::Wrapper, DL, Ty, Hi,
getTargetNode(N, Ty, DAG, LoFlag));
@ -324,9 +321,9 @@ namespace llvm {
// computing a symbol's address in non-PIC mode:
//
// (add %hi(sym), %lo(sym))
template<class NodeTy>
SDValue getAddrNonPIC(NodeTy *N, EVT Ty, SelectionDAG &DAG) const {
SDLoc DL(N);
template <class NodeTy>
SDValue getAddrNonPIC(NodeTy *N, SDLoc DL, EVT Ty,
SelectionDAG &DAG) const {
SDValue Hi = getTargetNode(N, Ty, DAG, MipsII::MO_ABS_HI);
SDValue Lo = getTargetNode(N, Ty, DAG, MipsII::MO_ABS_LO);
return DAG.getNode(ISD::ADD, DL, Ty,
@ -338,9 +335,8 @@ namespace llvm {
// computing a symbol's address using gp-relative addressing:
//
// (add $gp, %gp_rel(sym))
template<class NodeTy>
SDValue getAddrGPRel(NodeTy *N, EVT Ty, SelectionDAG &DAG) const {
SDLoc DL(N);
template <class NodeTy>
SDValue getAddrGPRel(NodeTy *N, SDLoc DL, EVT Ty, SelectionDAG &DAG) const {
assert(Ty == MVT::i32);
SDValue GPRel = getTargetNode(N, Ty, DAG, MipsII::MO_GPREL);
return DAG.getNode(ISD::ADD, DL, Ty,

3
contrib/llvm/lib/Target/NVPTX/NVPTXAllocaHoisting.h
View File

@ -15,6 +15,7 @@
#define LLVM_LIB_TARGET_NVPTX_NVPTXALLOCAHOISTING_H
#include "llvm/CodeGen/MachineFunctionAnalysis.h"
#include "llvm/CodeGen/StackProtector.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/Pass.h"
@ -32,8 +33,8 @@ public:
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<DataLayoutPass>();
AU.addPreserved("stack-protector");
AU.addPreserved<MachineFunctionAnalysis>();
AU.addPreserved<StackProtector>();
}
const char *getPassName() const override {

3
contrib/llvm/lib/Target/NVPTX/NVPTXLowerAggrCopies.h
View File

@ -16,6 +16,7 @@
#define LLVM_LIB_TARGET_NVPTX_NVPTXLOWERAGGRCOPIES_H
#include "llvm/CodeGen/MachineFunctionAnalysis.h"
#include "llvm/CodeGen/StackProtector.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/Pass.h"
@ -29,8 +30,8 @@ struct NVPTXLowerAggrCopies : public FunctionPass {
void getAnalysisUsage(AnalysisUsage &AU) const override {
AU.addRequired<DataLayoutPass>();
AU.addPreserved("stack-protector");
AU.addPreserved<MachineFunctionAnalysis>();
AU.addPreserved<StackProtector>();
}
bool runOnFunction(Function &F) override;

16
contrib/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMachObjectWriter.cpp
View File

@ -41,7 +41,7 @@ public:
: MCMachObjectTargetWriter(Is64Bit, CPUType, CPUSubtype,
/*UseAggressiveSymbolFolding=*/Is64Bit) {}
void RecordRelocation(MachObjectWriter *Writer, MCAssembler &Asm,
void RecordRelocation(MachObjectWriter *Writer, const MCAssembler &Asm,
const MCAsmLayout &Layout, const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) override {
@ -282,7 +282,7 @@ bool PPCMachObjectWriter::RecordScatteredRelocation(
MachO::any_relocation_info MRE;
makeScatteredRelocationInfo(MRE, other_half, MachO::GENERIC_RELOC_PAIR,
Log2Size, IsPCRel, Value2);
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
} else {
// If the offset is more than 24-bits, it won't fit in a scattered
// relocation offset field, so we fall back to using a non-scattered
@ -296,7 +296,7 @@ bool PPCMachObjectWriter::RecordScatteredRelocation(
}
MachO::any_relocation_info MRE;
makeScatteredRelocationInfo(MRE, FixupOffset, Type, Log2Size, IsPCRel, Value);
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
return true;
}
@ -331,9 +331,9 @@ void PPCMachObjectWriter::RecordPPCRelocation(
// See <reloc.h>.
const uint32_t FixupOffset = getFixupOffset(Layout, Fragment, Fixup);
unsigned Index = 0;
unsigned IsExtern = 0;
unsigned Type = RelocType;
const MCSymbolData *RelSymbol = nullptr;
if (Target.isAbsolute()) { // constant
// SymbolNum of 0 indicates the absolute section.
//
@ -355,7 +355,8 @@ void PPCMachObjectWriter::RecordPPCRelocation(
// Check whether we need an external or internal relocation.
if (Writer->doesSymbolRequireExternRelocation(SD)) {
RelSymbol = SD;
IsExtern = 1;
Index = SD->getIndex();
// For external relocations, make sure to offset the fixup value to
// compensate for the addend of the symbol address, if it was
// undefined. This occurs with weak definitions, for example.
@ -374,8 +375,9 @@ void PPCMachObjectWriter::RecordPPCRelocation(
// struct relocation_info (8 bytes)
MachO::any_relocation_info MRE;
makeRelocationInfo(MRE, FixupOffset, Index, IsPCRel, Log2Size, false, Type);
Writer->addRelocation(RelSymbol, Fragment->getParent(), MRE);
makeRelocationInfo(MRE, FixupOffset, Index, IsPCRel, Log2Size, IsExtern,
Type);
Writer->addRelocation(Fragment->getParent(), MRE);
}
MCObjectWriter *llvm::createPPCMachObjectWriter(raw_ostream &OS, bool Is64Bit,

6
contrib/llvm/lib/Target/R600/AMDGPU.h
View File

@ -77,7 +77,11 @@ extern Target TheGCNTarget;
namespace AMDGPU {
enum TargetIndex {
TI_CONSTDATA_START
TI_CONSTDATA_START,
TI_SCRATCH_RSRC_DWORD0,
TI_SCRATCH_RSRC_DWORD1,
TI_SCRATCH_RSRC_DWORD2,
TI_SCRATCH_RSRC_DWORD3
};
}

5
contrib/llvm/lib/Target/R600/AMDGPU.td
View File

@ -92,6 +92,11 @@ def FeatureFlatAddressSpace : SubtargetFeature<"flat-address-space",
"true",
"Support flat address space">;
def FeatureVGPRSpilling : SubtargetFeature<"vgpr-spilling",
"EnableVGPRSpilling",
"true",
"Enable spilling of VGPRs to scratch memory">;
class SubtargetFeatureFetchLimit <string Value> :
SubtargetFeature <"fetch"#Value,
"TexVTXClauseSize",

19
contrib/llvm/lib/Target/R600/AMDGPUAsmPrinter.cpp
View File

@ -116,7 +116,6 @@ bool AMDGPUAsmPrinter::runOnMachineFunction(MachineFunction &MF) {
const AMDGPUSubtarget &STM = TM.getSubtarget<AMDGPUSubtarget>();
SIProgramInfo KernelInfo;
if (STM.isAmdHsaOS()) {
OutStreamer.SwitchSection(getObjFileLowering().getTextSection());
getSIProgramInfo(KernelInfo, MF);
EmitAmdKernelCodeT(MF, KernelInfo);
OutStreamer.EmitCodeAlignment(2 << (MF.getAlignment() - 1));
@ -421,6 +420,7 @@ static unsigned getRsrcReg(unsigned ShaderType) {
void AMDGPUAsmPrinter::EmitProgramInfoSI(const MachineFunction &MF,
const SIProgramInfo &KernelInfo) {
const AMDGPUSubtarget &STM = TM.getSubtarget<AMDGPUSubtarget>();
const SIMachineFunctionInfo *MFI = MF.getInfo<SIMachineFunctionInfo>();
unsigned RsrcReg = getRsrcReg(MFI->getShaderType());
@ -441,6 +441,10 @@ void AMDGPUAsmPrinter::EmitProgramInfoSI(const MachineFunction &MF,
OutStreamer.EmitIntValue(RsrcReg, 4);
OutStreamer.EmitIntValue(S_00B028_VGPRS(KernelInfo.VGPRBlocks) |
S_00B028_SGPRS(KernelInfo.SGPRBlocks), 4);
if (STM.isVGPRSpillingEnabled(MFI)) {
OutStreamer.EmitIntValue(R_0286E8_SPI_TMPRING_SIZE, 4);
OutStreamer.EmitIntValue(S_0286E8_WAVESIZE(KernelInfo.ScratchBlocks), 4);
}
}
if (MFI->getShaderType() == ShaderType::PIXEL) {
@ -504,6 +508,19 @@ void AMDGPUAsmPrinter::EmitAmdKernelCodeT(const MachineFunction &MF,
header.wavefront_size = STM.getWavefrontSize();
const MCSectionELF *VersionSection = OutContext.getELFSection(".hsa.version",
ELF::SHT_PROGBITS, 0, SectionKind::getReadOnly());
OutStreamer.SwitchSection(VersionSection);
OutStreamer.EmitBytes(Twine("HSA Code Unit:" +
Twine(header.hsail_version_major) + "." +
Twine(header.hsail_version_minor) + ":" +
"AMD:" +
Twine(header.amd_code_version_major) + "." +
Twine(header.amd_code_version_minor) + ":" +
"GFX8.1:0").str());
OutStreamer.SwitchSection(getObjFileLowering().getTextSection());
if (isVerbose()) {
OutStreamer.emitRawComment("amd_code_version_major = " +
Twine(header.amd_code_version_major), false);

59
contrib/llvm/lib/Target/R600/AMDGPUISelDAGToDAG.cpp
View File

@ -417,6 +417,28 @@ SDNode *AMDGPUDAGToDAGISel::Select(SDNode *N) {
N->getValueType(0), Ops);
}
case ISD::LOAD: {
// To simplify the TableGen patters, we replace all i64 loads with
// v2i32 loads. Alternatively, we could promote i64 loads to v2i32
// during DAG legalization, however, so places (ExpandUnalignedLoad)
// in the DAG legalizer assume that if i64 is legal, so doing this
// promotion early can cause problems.
EVT VT = N->getValueType(0);
LoadSDNode *LD = cast<LoadSDNode>(N);
if (VT != MVT::i64 || LD->getExtensionType() != ISD::NON_EXTLOAD)
break;
SDValue NewLoad = CurDAG->getLoad(MVT::v2i32, SDLoc(N), LD->getChain(),
LD->getBasePtr(), LD->getMemOperand());
SDValue BitCast = CurDAG->getNode(ISD::BITCAST, SDLoc(N),
MVT::i64, NewLoad);
CurDAG->ReplaceAllUsesOfValueWith(SDValue(N, 1), NewLoad.getValue(1));
CurDAG->ReplaceAllUsesOfValueWith(SDValue(N, 0), BitCast);
SelectCode(NewLoad.getNode());
N = BitCast.getNode();
break;
}
case AMDGPUISD::REGISTER_LOAD: {
if (ST.getGeneration() <= AMDGPUSubtarget::NORTHERN_ISLANDS)
break;
@ -962,16 +984,27 @@ bool AMDGPUDAGToDAGISel::SelectMUBUFScratch(SDValue Addr, SDValue &Rsrc,
const SITargetLowering& Lowering =
*static_cast<const SITargetLowering*>(getTargetLowering());
unsigned ScratchPtrReg =
TRI->getPreloadedValue(MF, SIRegisterInfo::SCRATCH_PTR);
unsigned ScratchOffsetReg =
TRI->getPreloadedValue(MF, SIRegisterInfo::SCRATCH_WAVE_OFFSET);
Lowering.CreateLiveInRegister(*CurDAG, &AMDGPU::SReg_32RegClass,
ScratchOffsetReg, MVT::i32);
SDValue Sym0 = CurDAG->getExternalSymbol("SCRATCH_RSRC_DWORD0", MVT::i32);
SDValue ScratchRsrcDword0 =
SDValue(CurDAG->getMachineNode(AMDGPU::S_MOV_B32, DL, MVT::i32, Sym0), 0);
SDValue ScratchPtr =
CurDAG->getCopyFromReg(CurDAG->getEntryNode(), DL,
MRI.getLiveInVirtReg(ScratchPtrReg), MVT::i64);
SDValue Sym1 = CurDAG->getExternalSymbol("SCRATCH_RSRC_DWORD1", MVT::i32);
SDValue ScratchRsrcDword1 =
SDValue(CurDAG->getMachineNode(AMDGPU::S_MOV_B32, DL, MVT::i32, Sym1), 0);
const SDValue RsrcOps[] = {
CurDAG->getTargetConstant(AMDGPU::SReg_64RegClassID, MVT::i32),
ScratchRsrcDword0,
CurDAG->getTargetConstant(AMDGPU::sub0, MVT::i32),
ScratchRsrcDword1,
CurDAG->getTargetConstant(AMDGPU::sub1, MVT::i32),
};
SDValue ScratchPtr = SDValue(CurDAG->getMachineNode(AMDGPU::REG_SEQUENCE, DL,
MVT::v2i32, RsrcOps), 0);
Rsrc = SDValue(Lowering.buildScratchRSRC(*CurDAG, DL, ScratchPtr), 0);
SOffset = CurDAG->getCopyFromReg(CurDAG->getEntryNode(), DL,
MRI.getLiveInVirtReg(ScratchOffsetReg), MVT::i32);
@ -988,22 +1021,6 @@ bool AMDGPUDAGToDAGISel::SelectMUBUFScratch(SDValue Addr, SDValue &Rsrc,
}
}
// (add FI, n0)
if ((Addr.getOpcode() == ISD::ADD || Addr.getOpcode() == ISD::OR) &&
isa<FrameIndexSDNode>(Addr.getOperand(0))) {
VAddr = Addr.getOperand(1);
ImmOffset = Addr.getOperand(0);
return true;
}
// (FI)
if (isa<FrameIndexSDNode>(Addr)) {
VAddr = SDValue(CurDAG->getMachineNode(AMDGPU::V_MOV_B32_e32, DL, MVT::i32,
CurDAG->getConstant(0, MVT::i32)), 0);
ImmOffset = Addr;
return true;
}
// (node)
VAddr = Addr;
ImmOffset = CurDAG->getTargetConstant(0, MVT::i16);

3
contrib/llvm/lib/Target/R600/AMDGPUISelLowering.cpp
View File

@ -187,9 +187,6 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(TargetMachine &TM) :
setOperationAction(ISD::LOAD, MVT::v2f32, Promote);
AddPromotedToType(ISD::LOAD, MVT::v2f32, MVT::v2i32);
setOperationAction(ISD::LOAD, MVT::i64, Promote);
AddPromotedToType(ISD::LOAD, MVT::i64, MVT::v2i32);
setOperationAction(ISD::LOAD, MVT::v4f32, Promote);
AddPromotedToType(ISD::LOAD, MVT::v4f32, MVT::v4i32);

33
contrib/llvm/lib/Target/R600/AMDGPUInstrInfo.cpp
View File

@ -341,8 +341,39 @@ int AMDGPUInstrInfo::getMaskedMIMGOp(uint16_t Opcode, unsigned Channels) const {
// instead.
namespace llvm {
namespace AMDGPU {
int getMCOpcode(uint16_t Opcode, unsigned Gen) {
static int getMCOpcode(uint16_t Opcode, unsigned Gen) {
return getMCOpcodeGen(Opcode, (enum Subtarget)Gen);
}
}
}
// This must be kept in sync with the SISubtarget class in SIInstrInfo.td
enum SISubtarget {
SI = 0,
VI = 1
};
enum SISubtarget AMDGPUSubtargetToSISubtarget(unsigned Gen) {
switch (Gen) {
default:
return SI;
case AMDGPUSubtarget::VOLCANIC_ISLANDS:
return VI;
}
}
int AMDGPUInstrInfo::pseudoToMCOpcode(int Opcode) const {
int MCOp = AMDGPU::getMCOpcode(Opcode,
AMDGPUSubtargetToSISubtarget(RI.ST.getGeneration()));
// -1 means that Opcode is already a native instruction.
if (MCOp == -1)
return Opcode;
// (uint16_t)-1 means that Opcode is a pseudo instruction that has
// no encoding in the given subtarget generation.
if (MCOp == (uint16_t)-1)
return -1;
return MCOp;
}

5
contrib/llvm/lib/Target/R600/AMDGPUInstrInfo.h
View File

@ -135,6 +135,11 @@ public:
bool isRegisterStore(const MachineInstr &MI) const;
bool isRegisterLoad(const MachineInstr &MI) const;
/// \brief Return a target-specific opcode if Opcode is a pseudo instruction.
/// Return -1 if the target-specific opcode for the pseudo instruction does
/// not exist. If Opcode is not a pseudo instruction, this is identity.
int pseudoToMCOpcode(int Opcode) const;
//===---------------------------------------------------------------------===//
// Pure virtual funtions to be implemented by sub-classes.
//===---------------------------------------------------------------------===//

37
contrib/llvm/lib/Target/R600/AMDGPUMCInstLower.cpp
View File

@ -22,6 +22,7 @@
#include "llvm/CodeGen/MachineBasicBlock.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/MC/MCCodeEmitter.h"
#include "llvm/MC/MCContext.h"
@ -39,29 +40,17 @@ AMDGPUMCInstLower::AMDGPUMCInstLower(MCContext &ctx, const AMDGPUSubtarget &st):
Ctx(ctx), ST(st)
{ }
enum AMDGPUMCInstLower::SISubtarget
AMDGPUMCInstLower::AMDGPUSubtargetToSISubtarget(unsigned Gen) const {
switch (Gen) {
default:
return AMDGPUMCInstLower::SI;
case AMDGPUSubtarget::VOLCANIC_ISLANDS:
return AMDGPUMCInstLower::VI;
}
}
unsigned AMDGPUMCInstLower::getMCOpcode(unsigned MIOpcode) const {
int MCOpcode = AMDGPU::getMCOpcode(MIOpcode,
AMDGPUSubtargetToSISubtarget(ST.getGeneration()));
if (MCOpcode == -1)
MCOpcode = MIOpcode;
return MCOpcode;
}
void AMDGPUMCInstLower::lower(const MachineInstr *MI, MCInst &OutMI) const {
OutMI.setOpcode(getMCOpcode(MI->getOpcode()));
int MCOpcode = ST.getInstrInfo()->pseudoToMCOpcode(MI->getOpcode());
if (MCOpcode == -1) {
LLVMContext &C = MI->getParent()->getParent()->getFunction()->getContext();
C.emitError("AMDGPUMCInstLower::lower - Pseudo instruction doesn't have "
"a target-specific version: " + Twine(MI->getOpcode()));
}
OutMI.setOpcode(MCOpcode);
for (const MachineOperand &MO : MI->explicit_operands()) {
MCOperand MCOp;
@ -91,6 +80,12 @@ void AMDGPUMCInstLower::lower(const MachineInstr *MI, MCInst &OutMI) const {
MCOp = MCOperand::CreateExpr(Expr);
break;
}
case MachineOperand::MO_ExternalSymbol: {
MCSymbol *Sym = Ctx.GetOrCreateSymbol(StringRef(MO.getSymbolName()));
const MCSymbolRefExpr *Expr = MCSymbolRefExpr::Create(Sym, Ctx);
MCOp = MCOperand::CreateExpr(Expr);
break;
}
}
OutMI.addOperand(MCOp);
}

14
contrib/llvm/lib/Target/R600/AMDGPUMCInstLower.h
View File

@ -19,23 +19,9 @@ class MCContext;
class MCInst;
class AMDGPUMCInstLower {
// This must be kept in sync with the SISubtarget class in SIInstrInfo.td
enum SISubtarget {
SI = 0,
VI = 1
};
MCContext &Ctx;
const AMDGPUSubtarget &ST;
/// Convert a member of the AMDGPUSubtarget::Generation enum to the
/// SISubtarget enum.
enum SISubtarget AMDGPUSubtargetToSISubtarget(unsigned Gen) const;
/// Get the MC opcode for this MachineInstr.
unsigned getMCOpcode(unsigned MIOpcode) const;
public:
AMDGPUMCInstLower(MCContext &ctx, const AMDGPUSubtarget &ST);

26
contrib/llvm/lib/Target/R600/AMDGPUSubtarget.cpp
View File

@ -18,7 +18,9 @@
#include "R600MachineScheduler.h"
#include "SIISelLowering.h"
#include "SIInstrInfo.h"
#include "SIMachineFunctionInfo.h"
#include "llvm/ADT/SmallString.h"
#include "llvm/CodeGen/MachineScheduler.h"
using namespace llvm;
@ -78,6 +80,7 @@ AMDGPUSubtarget::AMDGPUSubtarget(StringRef TT, StringRef GPU, StringRef FS,
FlatAddressSpace(false), EnableIRStructurizer(true),
EnablePromoteAlloca(false), EnableIfCvt(true),
EnableLoadStoreOpt(false), WavefrontSize(0), CFALUBug(false), LocalMemorySize(0),
EnableVGPRSpilling(false),
DL(computeDataLayout(initializeSubtargetDependencies(GPU, FS))),
FrameLowering(TargetFrameLowering::StackGrowsUp,
64 * 16, // Maximum stack alignment (long16)
@ -113,3 +116,26 @@ unsigned AMDGPUSubtarget::getAmdKernelCodeChipID() const {
case SEA_ISLANDS: return 12;
}
}
bool AMDGPUSubtarget::isVGPRSpillingEnabled(
const SIMachineFunctionInfo *MFI) const {
return MFI->getShaderType() == ShaderType::COMPUTE || EnableVGPRSpilling;
}
void AMDGPUSubtarget::overrideSchedPolicy(MachineSchedPolicy &Policy,
MachineInstr *begin,
MachineInstr *end,
unsigned NumRegionInstrs) const {
if (getGeneration() >= SOUTHERN_ISLANDS) {
// Track register pressure so the scheduler can try to decrease
// pressure once register usage is above the threshold defined by
// SIRegisterInfo::getRegPressureSetLimit()
Policy.ShouldTrackPressure = true;
// Enabling both top down and bottom up scheduling seems to give us less
// register spills than just using one of these approaches on its own.
Policy.OnlyTopDown = false;
Policy.OnlyBottomUp = false;
}
}

16
contrib/llvm/lib/Target/R600/AMDGPUSubtarget.h
View File

@ -30,6 +30,8 @@
namespace llvm {
class SIMachineFunctionInfo;
class AMDGPUSubtarget : public AMDGPUGenSubtargetInfo {
public:
@ -63,6 +65,7 @@ private:
unsigned WavefrontSize;
bool CFALUBug;
int LocalMemorySize;
bool EnableVGPRSpilling;
const DataLayout DL;
AMDGPUFrameLowering FrameLowering;
@ -206,6 +209,10 @@ public:
return getGeneration() <= NORTHERN_ISLANDS;
}
void overrideSchedPolicy(MachineSchedPolicy &Policy,
MachineInstr *begin, MachineInstr *end,
unsigned NumRegionInstrs) const override;
// Helper functions to simplify if statements
bool isTargetELF() const {
return false;
@ -224,6 +231,15 @@ public:
bool isAmdHsaOS() const {
return TargetTriple.getOS() == Triple::AMDHSA;
}
bool isVGPRSpillingEnabled(const SIMachineFunctionInfo *MFI) const;
unsigned getMaxWavesPerCU() const {
if (getGeneration() >= AMDGPUSubtarget::SOUTHERN_ISLANDS)
return 10;
// FIXME: Not sure what this is for other subtagets.
llvm_unreachable("do not know max waves per CU for this subtarget.");
}
};
} // End namespace llvm

2
contrib/llvm/lib/Target/R600/MCTargetDesc/AMDGPUAsmBackend.cpp
View File

@ -29,7 +29,7 @@ public:
const MCAsmLayout &Layout) override {
//XXX: Implement if necessary.
}
void RecordRelocation(MCAssembler &Asm, const MCAsmLayout &Layout,
void RecordRelocation(const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, bool &IsPCRel,
uint64_t &FixedValue) override {

4
contrib/llvm/lib/Target/R600/SIDefines.h
View File

@ -163,4 +163,8 @@ namespace SIOutMods {
#define R_00B860_COMPUTE_TMPRING_SIZE 0x00B860
#define S_00B860_WAVESIZE(x) (((x) & 0x1FFF) << 12)
#define R_0286E8_SPI_TMPRING_SIZE 0x0286E8
#define S_0286E8_WAVESIZE(x) (((x) & 0x1FFF) << 12)
#endif

6
contrib/llvm/lib/Target/R600/SIISelLowering.cpp
View File

@ -588,6 +588,12 @@ SDValue SITargetLowering::LowerFormalArguments(
InVals.push_back(Val);
}
if (Info->getShaderType() != ShaderType::COMPUTE) {
unsigned ScratchIdx = CCInfo.getFirstUnallocated(
AMDGPU::SGPR_32RegClass.begin(), AMDGPU::SGPR_32RegClass.getNumRegs());
Info->ScratchOffsetReg = AMDGPU::SGPR_32RegClass.getRegister(ScratchIdx);
}
return Chain;
}

36
contrib/llvm/lib/Target/R600/SIInstrFormats.td
View File

@ -85,49 +85,41 @@ class Enc64 {
let Uses = [EXEC] in {
class VOPCCommon <dag ins, string asm, list<dag> pattern> :
InstSI <(outs VCCReg:$dst), ins, asm, pattern> {
class VOPAnyCommon <dag outs, dag ins, string asm, list<dag> pattern> :
InstSI <outs, ins, asm, pattern> {
let DisableEncoding = "$dst";
let mayLoad = 0;
let mayStore = 0;
let hasSideEffects = 0;
let UseNamedOperandTable = 1;
let VOPC = 1;
let VALU = 1;
}
class VOPCCommon <dag ins, string asm, list<dag> pattern> :
VOPAnyCommon <(outs VCCReg:$dst), ins, asm, pattern> {
let DisableEncoding = "$dst";
let VOPC = 1;
let Size = 4;
}
class VOP1Common <dag outs, dag ins, string asm, list<dag> pattern> :
InstSI <outs, ins, asm, pattern> {
let mayLoad = 0;
let mayStore = 0;
let hasSideEffects = 0;
let UseNamedOperandTable = 1;
VOPAnyCommon <outs, ins, asm, pattern> {
let VOP1 = 1;
let VALU = 1;
let Size = 4;
}
class VOP2Common <dag outs, dag ins, string asm, list<dag> pattern> :
InstSI <outs, ins, asm, pattern> {
VOPAnyCommon <outs, ins, asm, pattern> {
let mayLoad = 0;
let mayStore = 0;
let hasSideEffects = 0;
let UseNamedOperandTable = 1;
let VOP2 = 1;
let VALU = 1;
let Size = 4;
}
class VOP3Common <dag outs, dag ins, string asm, list<dag> pattern> :
InstSI <outs, ins, asm, pattern> {
VOPAnyCommon <outs, ins, asm, pattern> {
let mayLoad = 0;
let mayStore = 0;
let hasSideEffects = 0;
let UseNamedOperandTable = 1;
// Using complex patterns gives VOP3 patterns a very high complexity rating,
// but standalone patterns are almost always prefered, so we need to adjust the
// priority lower. The goal is to use a high number to reduce complexity to
@ -135,8 +127,6 @@ class VOP3Common <dag outs, dag ins, string asm, list<dag> pattern> :
let AddedComplexity = -1000;
let VOP3 = 1;
let VALU = 1;
int Size = 8;
}

48
contrib/llvm/lib/Target/R600/SIInstrInfo.cpp
View File

@ -430,15 +430,6 @@ unsigned SIInstrInfo::getMovOpcode(const TargetRegisterClass *DstRC) const {
return AMDGPU::COPY;
}
static bool shouldTryToSpillVGPRs(MachineFunction *MF) {
SIMachineFunctionInfo *MFI = MF->getInfo<SIMachineFunctionInfo>();
// FIXME: Implement spilling for other shader types.
return MFI->getShaderType() == ShaderType::COMPUTE;
}
void SIInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
unsigned SrcReg, bool isKill,
@ -462,7 +453,7 @@ void SIInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
case 256: Opcode = AMDGPU::SI_SPILL_S256_SAVE; break;
case 512: Opcode = AMDGPU::SI_SPILL_S512_SAVE; break;
}
} else if(shouldTryToSpillVGPRs(MF) && RI.hasVGPRs(RC)) {
} else if(RI.hasVGPRs(RC) && ST.isVGPRSpillingEnabled(MFI)) {
MFI->setHasSpilledVGPRs();
switch(RC->getSize() * 8) {
@ -482,7 +473,7 @@ void SIInstrInfo::storeRegToStackSlot(MachineBasicBlock &MBB,
.addFrameIndex(FrameIndex)
// Place-holder registers, these will be filled in by
// SIPrepareScratchRegs.
.addReg(AMDGPU::SGPR0_SGPR1, RegState::Undef)
.addReg(AMDGPU::SGPR0_SGPR1_SGPR2_SGPR3, RegState::Undef)
.addReg(AMDGPU::SGPR0, RegState::Undef);
} else {
LLVMContext &Ctx = MF->getFunction()->getContext();
@ -499,6 +490,7 @@ void SIInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
const TargetRegisterClass *RC,
const TargetRegisterInfo *TRI) const {
MachineFunction *MF = MBB.getParent();
const SIMachineFunctionInfo *MFI = MF->getInfo<SIMachineFunctionInfo>();
MachineFrameInfo *FrameInfo = MF->getFrameInfo();
DebugLoc DL = MBB.findDebugLoc(MI);
int Opcode = -1;
@ -511,7 +503,7 @@ void SIInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
case 256: Opcode = AMDGPU::SI_SPILL_S256_RESTORE; break;
case 512: Opcode = AMDGPU::SI_SPILL_S512_RESTORE; break;
}
} else if(shouldTryToSpillVGPRs(MF) && RI.hasVGPRs(RC)) {
} else if(RI.hasVGPRs(RC) && ST.isVGPRSpillingEnabled(MFI)) {
switch(RC->getSize() * 8) {
case 32: Opcode = AMDGPU::SI_SPILL_V32_RESTORE; break;
case 64: Opcode = AMDGPU::SI_SPILL_V64_RESTORE; break;
@ -528,7 +520,7 @@ void SIInstrInfo::loadRegFromStackSlot(MachineBasicBlock &MBB,
.addFrameIndex(FrameIndex)
// Place-holder registers, these will be filled in by
// SIPrepareScratchRegs.
.addReg(AMDGPU::SGPR0_SGPR1, RegState::Undef)
.addReg(AMDGPU::SGPR0_SGPR1_SGPR2_SGPR3, RegState::Undef)
.addReg(AMDGPU::SGPR0, RegState::Undef);
} else {
@ -615,7 +607,7 @@ unsigned SIInstrInfo::calculateLDSSpillAddress(MachineBasicBlock &MBB,
.addImm(-1)
.addImm(0);
BuildMI(Entry, Insert, DL, get(AMDGPU::V_MBCNT_HI_U32_B32_e32),
BuildMI(Entry, Insert, DL, get(AMDGPU::V_MBCNT_HI_U32_B32_e64),
TIDReg)
.addImm(-1)
.addReg(TIDReg);
@ -1053,7 +1045,11 @@ bool SIInstrInfo::canFoldOffset(unsigned OffsetSize, unsigned AS) const {
}
bool SIInstrInfo::hasVALU32BitEncoding(unsigned Opcode) const {
return AMDGPU::getVOPe32(Opcode) != -1;
int Op32 = AMDGPU::getVOPe32(Opcode);
if (Op32 == -1)
return false;
return pseudoToMCOpcode(Op32) != -1;
}
bool SIInstrInfo::hasModifiers(unsigned Opcode) const {
@ -1126,12 +1122,18 @@ bool SIInstrInfo::verifyInstruction(const MachineInstr *MI,
}
switch (Desc.OpInfo[i].OperandType) {
case MCOI::OPERAND_REGISTER: {
if (MI->getOperand(i).isImm() &&
!isImmOperandLegal(MI, i, MI->getOperand(i))) {
ErrInfo = "Illegal immediate value for operand.";
return false;
}
case MCOI::OPERAND_REGISTER:
if (MI->getOperand(i).isImm() || MI->getOperand(i).isFPImm()) {
ErrInfo = "Illegal immediate value for operand.";
return false;
}
break;
case AMDGPU::OPERAND_REG_IMM32:
break;
case AMDGPU::OPERAND_REG_INLINE_C:
if (MI->getOperand(i).isImm() && !isInlineConstant(MI->getOperand(i))) {
ErrInfo = "Illegal immediate value for operand.";
return false;
}
break;
case MCOI::OPERAND_IMMEDIATE:
@ -1287,7 +1289,7 @@ unsigned SIInstrInfo::getVALUOp(const MachineInstr &MI) {
case AMDGPU::S_LOAD_DWORDX2_SGPR: return AMDGPU::BUFFER_LOAD_DWORDX2_ADDR64;
case AMDGPU::S_LOAD_DWORDX4_IMM:
case AMDGPU::S_LOAD_DWORDX4_SGPR: return AMDGPU::BUFFER_LOAD_DWORDX4_ADDR64;
case AMDGPU::S_BCNT1_I32_B32: return AMDGPU::V_BCNT_U32_B32_e32;
case AMDGPU::S_BCNT1_I32_B32: return AMDGPU::V_BCNT_U32_B32_e64;
case AMDGPU::S_FF1_I32_B32: return AMDGPU::V_FFBL_B32_e32;
case AMDGPU::S_FLBIT_I32_B32: return AMDGPU::V_FFBH_U32_e32;
}
@ -2278,7 +2280,7 @@ void SIInstrInfo::splitScalar64BitBCNT(SmallVectorImpl<MachineInstr *> &Worklist
MachineOperand &Dest = Inst->getOperand(0);
MachineOperand &Src = Inst->getOperand(1);
const MCInstrDesc &InstDesc = get(AMDGPU::V_BCNT_U32_B32_e32);
const MCInstrDesc &InstDesc = get(AMDGPU::V_BCNT_U32_B32_e64);
const TargetRegisterClass *SrcRC = Src.isReg() ?
MRI.getRegClass(Src.getReg()) :
&AMDGPU::SGPR_32RegClass;

1
contrib/llvm/lib/Target/R600/SIInstrInfo.h
View File

@ -325,7 +325,6 @@ namespace AMDGPU {
int getVOPe32(uint16_t Opcode);
int getCommuteRev(uint16_t Opcode);
int getCommuteOrig(uint16_t Opcode);
int getMCOpcode(uint16_t Opcode, unsigned Gen);
int getAddr64Inst(uint16_t Opcode);
int getAtomicRetOp(uint16_t Opcode);
int getAtomicNoRetOp(uint16_t Opcode);

102
contrib/llvm/lib/Target/R600/SIInstrInfo.td
View File

@ -36,6 +36,12 @@ class vop2 <bits<6> si, bits<6> vi = si> : vop {
field bits<10> VI3 = {0, 1, 0, 0, vi{5-0}};
}
// Specify a VOP2 opcode for SI and VOP3 opcode for VI
// that doesn't have VOP2 encoding on VI
class vop23 <bits<6> si, bits<10> vi> : vop2 <si> {
let VI3 = vi;
}
class vop3 <bits<9> si, bits<10> vi = {0, si}> : vop {
let SI3 = si;
let VI3 = vi;
@ -57,7 +63,7 @@ class sopk <bits<5> si, bits<5> vi = si> {
}
// Execpt for the NONE field, this must be kept in sync with the SISubtarget enum
// in AMDGPUMCInstLower.h
// in AMDGPUInstrInfo.cpp
def SISubtarget {
int NONE = -1;
int SI = 0;
@ -731,7 +737,7 @@ class getAsm32 <int NumSrcArgs> {
// Returns the assembly string for the inputs and outputs of a VOP3
// instruction.
class getAsm64 <int NumSrcArgs, bit HasModifiers> {
string src0 = "$src0_modifiers,";
string src0 = !if(!eq(NumSrcArgs, 1), "$src0_modifiers", "$src0_modifiers,");
string src1 = !if(!eq(NumSrcArgs, 1), "",
!if(!eq(NumSrcArgs, 2), " $src1_modifiers",
" $src1_modifiers,"));
@ -848,6 +854,16 @@ class VOP2_Pseudo <dag outs, dag ins, list<dag> pattern, string opName> :
let isPseudo = 1;
}
multiclass VOP2SI_m <vop2 op, dag outs, dag ins, string asm, list<dag> pattern,
string opName, string revOpSI> {
def "" : VOP2_Pseudo <outs, ins, pattern, opName>,
VOP2_REV<revOpSI#"_e32", !eq(revOpSI, opName)>;
def _si : VOP2 <op.SI, outs, ins, opName#asm, []>,
VOP2_REV<revOpSI#"_e32_si", !eq(revOpSI, opName)>,
SIMCInstr <opName#"_e32", SISubtarget.SI>;
}
multiclass VOP2_m <vop2 op, dag outs, dag ins, string asm, list<dag> pattern,
string opName, string revOpSI, string revOpVI> {
def "" : VOP2_Pseudo <outs, ins, pattern, opName>,
@ -889,16 +905,6 @@ class VOP3_Real_vi <bits<10> op, dag outs, dag ins, string asm, string opName> :
VOP3e_vi <op>,
SIMCInstr <opName#"_e64", SISubtarget.VI>;
// VI only instruction
class VOP3_vi <bits<10> op, string opName, dag outs, dag ins, string asm,
list<dag> pattern, int NumSrcArgs, bit HasMods = 1> :
VOP3Common <outs, ins, asm, pattern>,
VOP <opName>,
VOP3e_vi <op>,
VOP3DisableFields<!if(!eq(NumSrcArgs, 1), 0, 1),
!if(!eq(NumSrcArgs, 2), 0, 1),
HasMods>;
multiclass VOP3_m <vop op, dag outs, dag ins, string asm, list<dag> pattern,
string opName, int NumSrcArgs, bit HasMods = 1> {
@ -998,6 +1004,23 @@ multiclass VOP3_C_m <vop op, dag outs, dag ins, string asm,
}
}
// An instruction that is VOP2 on SI and VOP3 on VI, no modifiers.
multiclass VOP2SI_3VI_m <vop3 op, string opName, dag outs, dag ins,
string asm, list<dag> pattern = []> {
let isPseudo = 1 in {
def "" : VOPAnyCommon <outs, ins, "", pattern>,
SIMCInstr<opName, SISubtarget.NONE>;
}
def _si : VOP2 <op.SI3{5-0}, outs, ins, asm, []>,
SIMCInstr <opName, SISubtarget.SI>;
def _vi : VOP3Common <outs, ins, asm, []>,
VOP3e_vi <op.VI3>,
VOP3DisableFields <1, 0, 0>,
SIMCInstr <opName, SISubtarget.VI>;
}
multiclass VOP1_Helper <vop1 op, string opName, dag outs,
dag ins32, string asm32, list<dag> pat32,
dag ins64, string asm64, list<dag> pat64,
@ -1089,6 +1112,33 @@ multiclass VOP2bInst <vop2 op, string opName, VOPProfile P,
revOp, P.HasModifiers
>;
// A VOP2 instruction that is VOP3-only on VI.
multiclass VOP2_VI3_Helper <vop23 op, string opName, dag outs,
dag ins32, string asm32, list<dag> pat32,
dag ins64, string asm64, list<dag> pat64,
string revOpSI, string revOpVI, bit HasMods> {
defm _e32 : VOP2SI_m <op, outs, ins32, asm32, pat32, opName, revOpSI>;
defm _e64 : VOP3_2_m <op, outs, ins64, opName#"_e64"#asm64, pat64, opName,
revOpSI, revOpVI, HasMods>;
}
multiclass VOP2_VI3_Inst <vop23 op, string opName, VOPProfile P,
SDPatternOperator node = null_frag,
string revOpSI = opName, string revOpVI = revOpSI>
: VOP2_VI3_Helper <
op, opName, P.Outs,
P.Ins32, P.Asm32, [],
P.Ins64, P.Asm64,
!if(P.HasModifiers,
[(set P.DstVT:$dst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
[(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))]),
revOpSI, revOpVI, P.HasModifiers
>;
class VOPC_Pseudo <dag outs, dag ins, list<dag> pattern, string opName> :
VOPCCommon <ins, "", pattern>,
VOP <opName>,
@ -1224,34 +1274,6 @@ multiclass VOP3Inst <vop3 op, string opName, VOPProfile P,
P.NumSrcArgs, P.HasModifiers
>;
class VOP3InstVI <bits<10> op, string opName, VOPProfile P,
SDPatternOperator node = null_frag> : VOP3_vi <
op, opName#"_vi", P.Outs, P.Ins64, opName#P.Asm64,
!if(!eq(P.NumSrcArgs, 3),
!if(P.HasModifiers,
[(set P.DstVT:$dst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers)),
(P.Src2VT (VOP3Mods P.Src2VT:$src2, i32:$src2_modifiers))))],
[(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1,
P.Src2VT:$src2))]),
!if(!eq(P.NumSrcArgs, 2),
!if(P.HasModifiers,
[(set P.DstVT:$dst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod)),
(P.Src1VT (VOP3Mods P.Src1VT:$src1, i32:$src1_modifiers))))],
[(set P.DstVT:$dst, (node P.Src0VT:$src0, P.Src1VT:$src1))])
/* P.NumSrcArgs == 1 */,
!if(P.HasModifiers,
[(set P.DstVT:$dst,
(node (P.Src0VT (VOP3Mods0 P.Src0VT:$src0, i32:$src0_modifiers,
i1:$clamp, i32:$omod))))],
[(set P.DstVT:$dst, (node P.Src0VT:$src0))]))),
P.NumSrcArgs, P.HasModifiers
>;
multiclass VOP3b_Helper <vop op, RegisterClass vrc, RegisterOperand arc,
string opName, list<dag> pattern> :
VOP3b_2_m <

92
contrib/llvm/lib/Target/R600/SIInstructions.td
View File

@ -1525,25 +1525,25 @@ defm V_SUBBREV_U32 : VOP2bInst <vop2<0x2a, 0x1e>, "v_subbrev_u32",
} // End Uses = [VCC]
} // End isCommutable = 1, Defs = [VCC]
// These instructions only exist on SI and CI
let SubtargetPredicate = isSICI in {
def V_READLANE_B32 : VOP2 <
0x00000001,
defm V_READLANE_B32 : VOP2SI_3VI_m <
vop3 <0x001, 0x289>,
"v_readlane_b32",
(outs SReg_32:$vdst),
(ins VGPR_32:$src0, SSrc_32:$vsrc1),
"v_readlane_b32 $vdst, $src0, $vsrc1",
[]
"v_readlane_b32 $vdst, $src0, $vsrc1"
>;
def V_WRITELANE_B32 : VOP2 <
0x00000002,
defm V_WRITELANE_B32 : VOP2SI_3VI_m <
vop3 <0x002, 0x28a>,
"v_writelane_b32",
(outs VGPR_32:$vdst),
(ins SReg_32:$src0, SSrc_32:$vsrc1),
"v_writelane_b32 $vdst, $src0, $vsrc1",
[]
"v_writelane_b32 $vdst, $src0, $vsrc1"
>;
// These instructions only exist on SI and CI
let SubtargetPredicate = isSICI in {
let isCommutable = 1 in {
defm V_MAC_LEGACY_F32 : VOP2Inst <vop2<0x6>, "v_mac_legacy_f32",
VOP_F32_F32_F32
@ -1568,30 +1568,33 @@ defm V_LSHL_B32 : VOP2Inst <vop2<0x19>, "v_lshl_b32", VOP_I32_I32_I32, shl>;
}
} // End isCommutable = 1
} // End let SubtargetPredicate = SICI
defm V_BFM_B32 : VOP2Inst <vop2<0x1e>, "v_bfm_b32", VOP_I32_I32_I32,
AMDGPUbfm>;
defm V_BCNT_U32_B32 : VOP2Inst <vop2<0x22>, "v_bcnt_u32_b32", VOP_I32_I32_I32>;
defm V_MBCNT_LO_U32_B32 : VOP2Inst <vop2<0x23>, "v_mbcnt_lo_u32_b32",
defm V_BFM_B32 : VOP2_VI3_Inst <vop23<0x1e, 0x293>, "v_bfm_b32", VOP_I32_I32_I32,
AMDGPUbfm
>;
defm V_BCNT_U32_B32 : VOP2_VI3_Inst <vop23<0x22, 0x28b>, "v_bcnt_u32_b32",
VOP_I32_I32_I32
>;
defm V_MBCNT_HI_U32_B32 : VOP2Inst <vop2<0x24>, "v_mbcnt_hi_u32_b32",
defm V_MBCNT_LO_U32_B32 : VOP2_VI3_Inst <vop23<0x23, 0x28c>, "v_mbcnt_lo_u32_b32",
VOP_I32_I32_I32
>;
defm V_LDEXP_F32 : VOP2Inst <vop2<0x2b>, "v_ldexp_f32",
defm V_MBCNT_HI_U32_B32 : VOP2_VI3_Inst <vop23<0x24, 0x28d>, "v_mbcnt_hi_u32_b32",
VOP_I32_I32_I32
>;
defm V_LDEXP_F32 : VOP2_VI3_Inst <vop23<0x2b, 0x288>, "v_ldexp_f32",
VOP_F32_F32_I32, AMDGPUldexp
>;
////def V_CVT_PKACCUM_U8_F32 : VOP2_U8 <0x0000002c, "v_cvt_pkaccum_u8_f32", []>;
////def V_CVT_PKNORM_I16_F32 : VOP2_I16 <0x0000002d, "v_cvt_pknorm_i16_f32", []>;
////def V_CVT_PKNORM_U16_F32 : VOP2_U16 <0x0000002e, "v_cvt_pknorm_u16_f32", []>;
defm V_CVT_PKRTZ_F16_F32 : VOP2Inst <vop2<0x2f>, "v_cvt_pkrtz_f16_f32",
defm V_CVT_PKRTZ_F16_F32 : VOP2_VI3_Inst <vop23<0x2f, 0x296>, "v_cvt_pkrtz_f16_f32",
VOP_I32_F32_F32, int_SI_packf16
>;
////def V_CVT_PK_U16_U32 : VOP2_U16 <0x00000030, "v_cvt_pk_u16_u32", []>;
////def V_CVT_PK_I16_I32 : VOP2_I16 <0x00000031, "v_cvt_pk_i16_i32", []>;
} // End let SubtargetPredicate = SICI
//===----------------------------------------------------------------------===//
// VOP3 Instructions
//===----------------------------------------------------------------------===//
@ -1656,9 +1659,6 @@ defm V_ALIGNBYTE_B32 : VOP3Inst <vop3<0x14f, 0x1cf>, "v_alignbyte_b32",
VOP_I32_I32_I32_I32
>;
// Only on SI
defm V_MULLIT_F32 : VOP3Inst <vop3<0x150>, "v_mullit_f32",
VOP_F32_F32_F32_F32>;
defm V_MIN3_F32 : VOP3Inst <vop3<0x151>, "v_min3_f32",
VOP_F32_F32_F32_F32, AMDGPUfmin3>;
@ -1699,20 +1699,6 @@ defm V_DIV_FIXUP_F64 : VOP3Inst <
} // let SchedRW = [WriteDouble]
defm V_LSHL_B64 : VOP3Inst <vop3<0x161>, "v_lshl_b64",
VOP_I64_I64_I32, shl
>;
// Only on SI
defm V_LSHR_B64 : VOP3Inst <vop3<0x162>, "v_lshr_b64",
VOP_I64_I64_I32, srl
>;
// Only on SI
defm V_ASHR_I64 : VOP3Inst <vop3<0x163>, "v_ashr_i64",
VOP_I64_I64_I32, sra
>;
let SchedRW = [WriteDouble] in {
let isCommutable = 1 in {
@ -1785,6 +1771,26 @@ defm V_TRIG_PREOP_F64 : VOP3Inst <
} // let SchedRW = [WriteDouble]
// These instructions only exist on SI and CI
let SubtargetPredicate = isSICI in {
defm V_LSHL_B64 : VOP3Inst <vop3<0x161>, "v_lshl_b64",
VOP_I64_I64_I32, shl
>;
defm V_LSHR_B64 : VOP3Inst <vop3<0x162>, "v_lshr_b64",
VOP_I64_I64_I32, srl
>;
defm V_ASHR_I64 : VOP3Inst <vop3<0x163>, "v_ashr_i64",
VOP_I64_I64_I32, sra
>;
defm V_MULLIT_F32 : VOP3Inst <vop3<0x150>, "v_mullit_f32",
VOP_F32_F32_F32_F32>;
} // End SubtargetPredicate = isSICI
//===----------------------------------------------------------------------===//
// Pseudo Instructions
//===----------------------------------------------------------------------===//
@ -1943,14 +1949,14 @@ multiclass SI_SPILL_SGPR <RegisterClass sgpr_class> {
let UseNamedOperandTable = 1 in {
def _SAVE : InstSI <
(outs),
(ins sgpr_class:$src, i32imm:$frame_idx, SReg_64:$scratch_ptr,
(ins sgpr_class:$src, i32imm:$frame_idx, SReg_128:$scratch_rsrc,
SReg_32:$scratch_offset),
"", []
>;
def _RESTORE : InstSI <
(outs sgpr_class:$dst),
(ins i32imm:$frame_idx, SReg_64:$scratch_ptr, SReg_32:$scratch_offset),
(ins i32imm:$frame_idx, SReg_128:$scratch_rsrc, SReg_32:$scratch_offset),
"", []
>;
} // End UseNamedOperandTable = 1
@ -1966,14 +1972,14 @@ multiclass SI_SPILL_VGPR <RegisterClass vgpr_class> {
let UseNamedOperandTable = 1 in {
def _SAVE : InstSI <
(outs),
(ins vgpr_class:$src, i32imm:$frame_idx, SReg_64:$scratch_ptr,
(ins vgpr_class:$src, i32imm:$frame_idx, SReg_128:$scratch_rsrc,
SReg_32:$scratch_offset),
"", []
>;
def _RESTORE : InstSI <
(outs vgpr_class:$dst),
(ins i32imm:$frame_idx, SReg_64:$scratch_ptr, SReg_32:$scratch_offset),
(ins i32imm:$frame_idx, SReg_128:$scratch_rsrc, SReg_32:$scratch_offset),
"", []
>;
} // End UseNamedOperandTable = 1
@ -2728,16 +2734,12 @@ def : Pat <
(V_RCP_IFLAG_F32_e32 (V_CVT_F32_U32_e32 $src0))))
>;
let Predicates = [isSICI] in {
def : Pat <
(int_SI_tid),
(V_MBCNT_HI_U32_B32_e32 0xffffffff,
(V_MBCNT_HI_U32_B32_e64 0xffffffff,
(V_MBCNT_LO_U32_B32_e64 0xffffffff, 0))
>;
}
//===----------------------------------------------------------------------===//
// VOP3 Patterns
//===----------------------------------------------------------------------===//

1
contrib/llvm/lib/Target/R600/SIMachineFunctionInfo.h
View File

@ -50,6 +50,7 @@ public:
unsigned NumUserSGPRs;
std::map<unsigned, unsigned> LaneVGPRs;
unsigned LDSWaveSpillSize;
unsigned ScratchOffsetReg;
bool hasCalculatedTID() const { return TIDReg != AMDGPU::NoRegister; };
unsigned getTIDReg() const { return TIDReg; };
void setTIDReg(unsigned Reg) { TIDReg = Reg; }

100
contrib/llvm/lib/Target/R600/SIPrepareScratchRegs.cpp
View File

@ -84,28 +84,10 @@ bool SIPrepareScratchRegs::runOnMachineFunction(MachineFunction &MF) {
if (!Entry->isLiveIn(ScratchOffsetPreloadReg))
Entry->addLiveIn(ScratchOffsetPreloadReg);
// Load the scratch pointer
unsigned ScratchPtrReg =
TRI->findUnusedRegister(MRI, &AMDGPU::SGPR_64RegClass);
int ScratchPtrFI = -1;
if (ScratchPtrReg != AMDGPU::NoRegister) {
// Found an SGPR to use.
MRI.setPhysRegUsed(ScratchPtrReg);
BuildMI(*Entry, I, DL, TII->get(AMDGPU::S_MOV_B64), ScratchPtrReg)
.addReg(ScratchPtrPreloadReg);
} else {
// No SGPR is available, we must spill.
ScratchPtrFI = FrameInfo->CreateSpillStackObject(8, 4);
BuildMI(*Entry, I, DL, TII->get(AMDGPU::SI_SPILL_S64_SAVE))
.addReg(ScratchPtrPreloadReg)
.addFrameIndex(ScratchPtrFI);
}
// Load the scratch offset.
unsigned ScratchOffsetReg =
TRI->findUnusedRegister(MRI, &AMDGPU::SGPR_32RegClass);
int ScratchOffsetFI = ~0;
int ScratchOffsetFI = -1;
if (ScratchOffsetReg != AMDGPU::NoRegister) {
// Found an SGPR to use
@ -117,7 +99,9 @@ bool SIPrepareScratchRegs::runOnMachineFunction(MachineFunction &MF) {
ScratchOffsetFI = FrameInfo->CreateSpillStackObject(4,4);
BuildMI(*Entry, I, DL, TII->get(AMDGPU::SI_SPILL_S32_SAVE))
.addReg(ScratchOffsetPreloadReg)
.addFrameIndex(ScratchOffsetFI);
.addFrameIndex(ScratchOffsetFI)
.addReg(AMDGPU::SGPR0_SGPR1_SGPR2_SGPR3, RegState::Undef)
.addReg(AMDGPU::SGPR0, RegState::Undef);
}
@ -125,22 +109,27 @@ bool SIPrepareScratchRegs::runOnMachineFunction(MachineFunction &MF) {
// add them to all the SI_SPILL_V* instructions.
RegScavenger RS;
bool UseRegScavenger =
(ScratchPtrReg == AMDGPU::NoRegister ||
ScratchOffsetReg == AMDGPU::NoRegister);
unsigned ScratchRsrcFI = FrameInfo->CreateSpillStackObject(16, 4);
RS.addScavengingFrameIndex(ScratchRsrcFI);
for (MachineFunction::iterator BI = MF.begin(), BE = MF.end();
BI != BE; ++BI) {
MachineBasicBlock &MBB = *BI;
if (UseRegScavenger)
RS.enterBasicBlock(&MBB);
// Add the scratch offset reg as a live-in so that the register scavenger
// doesn't re-use it.
if (!MBB.isLiveIn(ScratchOffsetReg) &&
ScratchOffsetReg != AMDGPU::NoRegister)
MBB.addLiveIn(ScratchOffsetReg);
RS.enterBasicBlock(&MBB);
for (MachineBasicBlock::iterator I = MBB.begin(), E = MBB.end();
I != E; ++I) {
MachineInstr &MI = *I;
RS.forward(I);
DebugLoc DL = MI.getDebugLoc();
switch(MI.getOpcode()) {
default: break;;
default: break;
case AMDGPU::SI_SPILL_V512_SAVE:
case AMDGPU::SI_SPILL_V256_SAVE:
case AMDGPU::SI_SPILL_V128_SAVE:
@ -153,43 +142,66 @@ bool SIPrepareScratchRegs::runOnMachineFunction(MachineFunction &MF) {
case AMDGPU::SI_SPILL_V256_RESTORE:
case AMDGPU::SI_SPILL_V512_RESTORE:
// Scratch Pointer
if (ScratchPtrReg == AMDGPU::NoRegister) {
ScratchPtrReg = RS.scavengeRegister(&AMDGPU::SGPR_64RegClass, 0);
BuildMI(MBB, I, DL, TII->get(AMDGPU::SI_SPILL_S64_RESTORE),
ScratchPtrReg)
.addFrameIndex(ScratchPtrFI)
.addReg(AMDGPU::NoRegister)
.addReg(AMDGPU::NoRegister);
} else if (!MBB.isLiveIn(ScratchPtrReg)) {
MBB.addLiveIn(ScratchPtrReg);
}
// Scratch resource
unsigned ScratchRsrcReg =
RS.scavengeRegister(&AMDGPU::SReg_128RegClass, 0);
uint64_t Rsrc = AMDGPU::RSRC_DATA_FORMAT | AMDGPU::RSRC_TID_ENABLE |
0xffffffff; // Size
unsigned Rsrc0 = TRI->getSubReg(ScratchRsrcReg, AMDGPU::sub0);
unsigned Rsrc1 = TRI->getSubReg(ScratchRsrcReg, AMDGPU::sub1);
unsigned Rsrc2 = TRI->getSubReg(ScratchRsrcReg, AMDGPU::sub2);
unsigned Rsrc3 = TRI->getSubReg(ScratchRsrcReg, AMDGPU::sub3);
BuildMI(MBB, I, DL, TII->get(AMDGPU::S_MOV_B32), Rsrc0)
.addExternalSymbol("SCRATCH_RSRC_DWORD0")
.addReg(ScratchRsrcReg, RegState::ImplicitDefine);
BuildMI(MBB, I, DL, TII->get(AMDGPU::S_MOV_B32), Rsrc1)
.addExternalSymbol("SCRATCH_RSRC_DWORD1")
.addReg(ScratchRsrcReg, RegState::ImplicitDefine);
BuildMI(MBB, I, DL, TII->get(AMDGPU::S_MOV_B32), Rsrc2)
.addImm(Rsrc & 0xffffffff)
.addReg(ScratchRsrcReg, RegState::ImplicitDefine);
BuildMI(MBB, I, DL, TII->get(AMDGPU::S_MOV_B32), Rsrc3)
.addImm(Rsrc >> 32)
.addReg(ScratchRsrcReg, RegState::ImplicitDefine);
// Scratch Offset
if (ScratchOffsetReg == AMDGPU::NoRegister) {
ScratchOffsetReg = RS.scavengeRegister(&AMDGPU::SGPR_32RegClass, 0);
BuildMI(MBB, I, DL, TII->get(AMDGPU::SI_SPILL_S32_RESTORE),
ScratchOffsetReg)
.addFrameIndex(ScratchOffsetFI)
.addReg(AMDGPU::NoRegister)
.addReg(AMDGPU::NoRegister);
.addReg(AMDGPU::SGPR0_SGPR1_SGPR2_SGPR3, RegState::Undef)
.addReg(AMDGPU::SGPR0, RegState::Undef);
} else if (!MBB.isLiveIn(ScratchOffsetReg)) {
MBB.addLiveIn(ScratchOffsetReg);
}
if (ScratchPtrReg == AMDGPU::NoRegister ||
if (ScratchRsrcReg == AMDGPU::NoRegister ||
ScratchOffsetReg == AMDGPU::NoRegister) {
LLVMContext &Ctx = MF.getFunction()->getContext();
Ctx.emitError("ran out of SGPRs for spilling VGPRs");
ScratchPtrReg = AMDGPU::SGPR0;
ScratchRsrcReg = AMDGPU::SGPR0;
ScratchOffsetReg = AMDGPU::SGPR0;
}
MI.getOperand(2).setReg(ScratchPtrReg);
MI.getOperand(2).setReg(ScratchRsrcReg);
MI.getOperand(2).setIsKill(true);
MI.getOperand(2).setIsUndef(false);
MI.getOperand(3).setReg(ScratchOffsetReg);
MI.getOperand(3).setIsUndef(false);
MI.getOperand(3).setIsKill(false);
MI.addOperand(MachineOperand::CreateReg(Rsrc0, false, true, true));
MI.addOperand(MachineOperand::CreateReg(Rsrc1, false, true, true));
MI.addOperand(MachineOperand::CreateReg(Rsrc2, false, true, true));
MI.addOperand(MachineOperand::CreateReg(Rsrc3, false, true, true));
break;
}
if (UseRegScavenger)
RS.forward();
}
}
return true;

97
contrib/llvm/lib/Target/R600/SIRegisterInfo.cpp
View File

@ -23,7 +23,6 @@
#include "llvm/IR/Function.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/Support/Debug.h"
using namespace llvm;
SIRegisterInfo::SIRegisterInfo(const AMDGPUSubtarget &st)
@ -51,9 +50,32 @@ BitVector SIRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
return Reserved;
}
unsigned SIRegisterInfo::getRegPressureLimit(const TargetRegisterClass *RC,
MachineFunction &MF) const {
return RC->getNumRegs();
unsigned SIRegisterInfo::getRegPressureSetLimit(unsigned Idx) const {
// FIXME: We should adjust the max number of waves based on LDS size.
unsigned SGPRLimit = getNumSGPRsAllowed(ST.getMaxWavesPerCU());
unsigned VGPRLimit = getNumVGPRsAllowed(ST.getMaxWavesPerCU());
for (regclass_iterator I = regclass_begin(), E = regclass_end();
I != E; ++I) {
unsigned NumSubRegs = std::max((int)(*I)->getSize() / 4, 1);
unsigned Limit;
if (isSGPRClass(*I)) {
Limit = SGPRLimit / NumSubRegs;
} else {
Limit = VGPRLimit / NumSubRegs;
}
const int *Sets = getRegClassPressureSets(*I);
assert(Sets);
for (unsigned i = 0; Sets[i] != -1; ++i) {
if (Sets[i] == (int)Idx)
return Limit;
}
}
return 256;
}
bool SIRegisterInfo::requiresRegisterScavenging(const MachineFunction &Fn) const {
@ -98,7 +120,7 @@ static unsigned getNumSubRegsForSpillOp(unsigned Op) {
void SIRegisterInfo::buildScratchLoadStore(MachineBasicBlock::iterator MI,
unsigned LoadStoreOp,
unsigned Value,
unsigned ScratchPtr,
unsigned ScratchRsrcReg,
unsigned ScratchOffset,
int64_t Offset,
RegScavenger *RS) const {
@ -113,33 +135,9 @@ void SIRegisterInfo::buildScratchLoadStore(MachineBasicBlock::iterator MI,
bool RanOutOfSGPRs = false;
unsigned SOffset = ScratchOffset;
unsigned RsrcReg = RS->scavengeRegister(&AMDGPU::SReg_128RegClass, MI, 0);
if (RsrcReg == AMDGPU::NoRegister) {
RanOutOfSGPRs = true;
RsrcReg = AMDGPU::SGPR0_SGPR1_SGPR2_SGPR3;
}
unsigned NumSubRegs = getNumSubRegsForSpillOp(MI->getOpcode());
unsigned Size = NumSubRegs * 4;
uint64_t Rsrc = AMDGPU::RSRC_DATA_FORMAT | AMDGPU::RSRC_TID_ENABLE |
0xffffffff; // Size
BuildMI(*MBB, MI, DL, TII->get(AMDGPU::S_MOV_B64),
getSubReg(RsrcReg, AMDGPU::sub0_sub1))
.addReg(ScratchPtr)
.addReg(RsrcReg, RegState::ImplicitDefine);
BuildMI(*MBB, MI, DL, TII->get(AMDGPU::S_MOV_B32),
getSubReg(RsrcReg, AMDGPU::sub2))
.addImm(Rsrc & 0xffffffff)
.addReg(RsrcReg, RegState::ImplicitDefine);
BuildMI(*MBB, MI, DL, TII->get(AMDGPU::S_MOV_B32),
getSubReg(RsrcReg, AMDGPU::sub3))
.addImm(Rsrc >> 32)
.addReg(RsrcReg, RegState::ImplicitDefine);
if (!isUInt<12>(Offset + Size)) {
SOffset = RS->scavengeRegister(&AMDGPU::SGPR_32RegClass, MI, 0);
if (SOffset == AMDGPU::NoRegister) {
@ -163,9 +161,9 @@ void SIRegisterInfo::buildScratchLoadStore(MachineBasicBlock::iterator MI,
BuildMI(*MBB, MI, DL, TII->get(LoadStoreOp))
.addReg(SubReg, getDefRegState(IsLoad))
.addReg(RsrcReg, getKillRegState(IsKill))
.addReg(ScratchRsrcReg, getKillRegState(IsKill))
.addImm(Offset)
.addReg(SOffset, getKillRegState(IsKill))
.addReg(SOffset)
.addImm(0) // glc
.addImm(0) // slc
.addImm(0) // tfe
@ -235,9 +233,8 @@ void SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
Ctx.emitError("Ran out of VGPRs for spilling SGPR");
}
if (isM0) {
if (isM0)
SubReg = RS->scavengeRegister(&AMDGPU::SGPR_32RegClass, MI, 0);
}
BuildMI(*MBB, MI, DL, TII->get(AMDGPU::V_READLANE_B32), SubReg)
.addReg(Spill.VGPR)
@ -262,7 +259,7 @@ void SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
case AMDGPU::SI_SPILL_V32_SAVE:
buildScratchLoadStore(MI, AMDGPU::BUFFER_STORE_DWORD_OFFSET,
TII->getNamedOperand(*MI, AMDGPU::OpName::src)->getReg(),
TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_ptr)->getReg(),
TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_rsrc)->getReg(),
TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_offset)->getReg(),
FrameInfo->getObjectOffset(Index), RS);
MI->eraseFromParent();
@ -274,7 +271,7 @@ void SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
case AMDGPU::SI_SPILL_V512_RESTORE: {
buildScratchLoadStore(MI, AMDGPU::BUFFER_LOAD_DWORD_OFFSET,
TII->getNamedOperand(*MI, AMDGPU::OpName::dst)->getReg(),
TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_ptr)->getReg(),
TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_rsrc)->getReg(),
TII->getNamedOperand(*MI, AMDGPU::OpName::scratch_offset)->getReg(),
FrameInfo->getObjectOffset(Index), RS);
MI->eraseFromParent();
@ -289,7 +286,7 @@ void SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
BuildMI(*MBB, MI, MI->getDebugLoc(),
TII->get(AMDGPU::V_MOV_B32_e32), TmpReg)
.addImm(Offset);
FIOp.ChangeToRegister(TmpReg, false);
FIOp.ChangeToRegister(TmpReg, false, false, true);
}
}
}
@ -446,6 +443,8 @@ unsigned SIRegisterInfo::getPreloadedValue(const MachineFunction &MF,
case SIRegisterInfo::TGID_Z:
return AMDGPU::SReg_32RegClass.getRegister(MFI->NumUserSGPRs + 2);
case SIRegisterInfo::SCRATCH_WAVE_OFFSET:
if (MFI->getShaderType() != ShaderType::COMPUTE)
return MFI->ScratchOffsetReg;
return AMDGPU::SReg_32RegClass.getRegister(MFI->NumUserSGPRs + 4);
case SIRegisterInfo::SCRATCH_PTR:
return AMDGPU::SGPR2_SGPR3;
@ -475,3 +474,29 @@ unsigned SIRegisterInfo::findUnusedRegister(const MachineRegisterInfo &MRI,
return AMDGPU::NoRegister;
}
unsigned SIRegisterInfo::getNumVGPRsAllowed(unsigned WaveCount) const {
switch(WaveCount) {
case 10: return 24;
case 9: return 28;
case 8: return 32;
case 7: return 36;
case 6: return 40;
case 5: return 48;
case 4: return 64;
case 3: return 84;
case 2: return 128;
default: return 256;
}
}
unsigned SIRegisterInfo::getNumSGPRsAllowed(unsigned WaveCount) const {
switch(WaveCount) {
case 10: return 48;
case 9: return 56;
case 8: return 64;
case 7: return 72;
case 6: return 80;
case 5: return 96;
default: return 103;
}
}

14
contrib/llvm/lib/Target/R600/SIRegisterInfo.h
View File

@ -17,6 +17,7 @@
#define LLVM_LIB_TARGET_R600_SIREGISTERINFO_H
#include "AMDGPURegisterInfo.h"
#include "llvm/Support/Debug.h"
namespace llvm {
@ -26,8 +27,7 @@ struct SIRegisterInfo : public AMDGPURegisterInfo {
BitVector getReservedRegs(const MachineFunction &MF) const override;
unsigned getRegPressureLimit(const TargetRegisterClass *RC,
MachineFunction &MF) const override;
unsigned getRegPressureSetLimit(unsigned Idx) const override;
bool requiresRegisterScavenging(const MachineFunction &Fn) const override;
@ -105,13 +105,21 @@ struct SIRegisterInfo : public AMDGPURegisterInfo {
unsigned getPreloadedValue(const MachineFunction &MF,
enum PreloadedValue Value) const;
/// \brief Give the maximum number of VGPRs that can be used by \p WaveCount
/// concurrent waves.
unsigned getNumVGPRsAllowed(unsigned WaveCount) const;
/// \brief Give the maximum number of SGPRs that can be used by \p WaveCount
/// concurrent waves.
unsigned getNumSGPRsAllowed(unsigned WaveCount) const;
unsigned findUnusedRegister(const MachineRegisterInfo &MRI,
const TargetRegisterClass *RC) const;
private:
void buildScratchLoadStore(MachineBasicBlock::iterator MI,
unsigned LoadStoreOp, unsigned Value,
unsigned ScratchPtr, unsigned ScratchOffset,
unsigned ScratchRsrcReg, unsigned ScratchOffset,
int64_t Offset, RegScavenger *RS) const;
};

9
contrib/llvm/lib/Target/R600/SIShrinkInstructions.cpp
View File

@ -10,6 +10,7 @@
//
#include "AMDGPU.h"
#include "AMDGPUMCInstLower.h"
#include "AMDGPUSubtarget.h"
#include "SIInstrInfo.h"
#include "llvm/ADT/Statistic.h"
@ -206,13 +207,13 @@ bool SIShrinkInstructions::runOnMachineFunction(MachineFunction &MF) {
continue;
}
int Op32 = AMDGPU::getVOPe32(MI.getOpcode());
// Op32 could be -1 here if we started with an instruction that had a
// getVOPe32 could be -1 here if we started with an instruction that had
// a 32-bit encoding and then commuted it to an instruction that did not.
if (Op32 == -1)
if (!TII->hasVALU32BitEncoding(MI.getOpcode()))
continue;
int Op32 = AMDGPU::getVOPe32(MI.getOpcode());
if (TII->isVOPC(Op32)) {
unsigned DstReg = MI.getOperand(0).getReg();
if (TargetRegisterInfo::isVirtualRegister(DstReg)) {

29
contrib/llvm/lib/Target/R600/VIInstructions.td
View File

@ -11,22 +11,6 @@
let SubtargetPredicate = isVI in {
def V_LDEXP_F32 : VOP3InstVI <0x288, "v_ldexp_f32", VOP_F32_F32_I32,
AMDGPUldexp
>;
def V_BFM_B32 : VOP3InstVI <0x293, "v_bfm_b32", VOP_I32_I32_I32, AMDGPUbfm>;
def V_BCNT_U32_B32 : VOP3InstVI <0x28b, "v_bcnt_u32_b32", VOP_I32_I32_I32>;
def V_MBCNT_LO_U32_B32 : VOP3InstVI <0x28c, "v_mbcnt_lo_u32_b32",
VOP_I32_I32_I32
>;
def V_MBCNT_HI_U32_B32 : VOP3InstVI <0x28d, "v_mbcnt_hi_u32_b32",
VOP_I32_I32_I32
>;
def V_CVT_PKRTZ_F16_F32 : VOP3InstVI <0x296, "v_cvt_pkrtz_f16_f32",
VOP_I32_F32_F32, int_SI_packf16
>;
defm BUFFER_LOAD_DWORD_VI : MUBUF_Load_Helper_vi <
0x14, "buffer_load_dword", VGPR_32, i32, global_load
>;
@ -37,22 +21,13 @@ defm BUFFER_LOAD_FORMAT_XYZW_VI : MUBUF_Load_Helper_vi <
} // End SubtargetPredicate = isVI
//===----------------------------------------------------------------------===//
// VOP2 Patterns
//===----------------------------------------------------------------------===//
let Predicates = [isVI] in {
def : Pat <
(int_SI_tid),
(V_MBCNT_HI_U32_B32 0xffffffff,
(V_MBCNT_LO_U32_B32 0xffffffff, 0))
>;
//===----------------------------------------------------------------------===//
// SMEM Patterns
//===----------------------------------------------------------------------===//
let Predicates = [isVI] in {
// 1. Offset as 8bit DWORD immediate
def : Pat <
(SIload_constant v4i32:$sbase, IMM20bit:$offset),

3
contrib/llvm/lib/Target/Sparc/MCTargetDesc/SparcMCAsmInfo.cpp
View File

@ -42,7 +42,8 @@ SparcELFMCAsmInfo::SparcELFMCAsmInfo(StringRef TT) {
SunStyleELFSectionSwitchSyntax = true;
UsesELFSectionDirectiveForBSS = true;
UseIntegratedAssembler = true;
if (TheTriple.isOSSolaris() || TheTriple.isOSOpenBSD())
UseIntegratedAssembler = true;
}
const MCExpr*

13
contrib/llvm/lib/Target/X86/MCTargetDesc/X86AsmBackend.cpp
View File

@ -777,6 +777,19 @@ public:
MachO::CPU_TYPE_X86_64, Subtype);
}
bool doesSectionRequireSymbols(const MCSection &Section) const override {
// Temporary labels in the string literals sections require symbols. The
// issue is that the x86_64 relocation format does not allow symbol +
// offset, and so the linker does not have enough information to resolve the
// access to the appropriate atom unless an external relocation is used. For
// non-cstring sections, we expect the compiler to use a non-temporary label
// for anything that could have an addend pointing outside the symbol.
//
// See <rdar://problem/4765733>.
const MCSectionMachO &SMO = static_cast<const MCSectionMachO&>(Section);
return SMO.getType() == MachO::S_CSTRING_LITERALS;
}
/// \brief Generate the compact unwind encoding for the CFI instructions.
uint32_t generateCompactUnwindEncoding(
ArrayRef<MCCFIInstruction> Instrs) const override {

3
contrib/llvm/lib/Target/X86/MCTargetDesc/X86ELFObjectWriter.cpp
View File

@ -222,6 +222,9 @@ unsigned X86ELFObjectWriter::GetRelocType(const MCValue &Target,
case MCSymbolRefExpr::VK_GOT:
Type = ELF::R_386_GOT32;
break;
case MCSymbolRefExpr::VK_PLT:
Type = ELF::R_386_PLT32;
break;
case MCSymbolRefExpr::VK_GOTOFF:
Type = ELF::R_386_GOTOFF;
break;

114
contrib/llvm/lib/Target/X86/MCTargetDesc/X86MachObjectWriter.cpp
View File

@ -10,7 +10,6 @@
#include "MCTargetDesc/X86MCTargetDesc.h"
#include "MCTargetDesc/X86FixupKinds.h"
#include "llvm/ADT/Twine.h"
#include "llvm/MC/MCAsmInfo.h"
#include "llvm/MC/MCAsmLayout.h"
#include "llvm/MC/MCAssembler.h"
#include "llvm/MC/MCContext.h"
@ -48,21 +47,23 @@ class X86MachObjectWriter : public MCMachObjectTargetWriter {
const MCFixup &Fixup,
MCValue Target,
uint64_t &FixedValue);
void RecordX86_64Relocation(MachObjectWriter *Writer, MCAssembler &Asm,
void RecordX86_64Relocation(MachObjectWriter *Writer,
const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, uint64_t &FixedValue);
const MCFragment *Fragment,
const MCFixup &Fixup,
MCValue Target,
uint64_t &FixedValue);
public:
X86MachObjectWriter(bool Is64Bit, uint32_t CPUType,
uint32_t CPUSubtype)
: MCMachObjectTargetWriter(Is64Bit, CPUType, CPUSubtype,
/*UseAggressiveSymbolFolding=*/Is64Bit) {}
void RecordRelocation(MachObjectWriter *Writer, MCAssembler &Asm,
const MCAsmLayout &Layout, const MCFragment *Fragment,
const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) override {
void RecordRelocation(MachObjectWriter *Writer,
const MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup,
MCValue Target, uint64_t &FixedValue) override {
if (Writer->is64Bit())
RecordX86_64Relocation(Writer, Asm, Layout, Fragment, Fixup, Target,
FixedValue);
@ -96,10 +97,13 @@ static unsigned getFixupKindLog2Size(unsigned Kind) {
}
}
void X86MachObjectWriter::RecordX86_64Relocation(
MachObjectWriter *Writer, MCAssembler &Asm, const MCAsmLayout &Layout,
const MCFragment *Fragment, const MCFixup &Fixup, MCValue Target,
uint64_t &FixedValue) {
void X86MachObjectWriter::RecordX86_64Relocation(MachObjectWriter *Writer,
const MCAssembler &Asm,
const MCAsmLayout &Layout,
const MCFragment *Fragment,
const MCFixup &Fixup,
MCValue Target,
uint64_t &FixedValue) {
unsigned IsPCRel = Writer->isFixupKindPCRel(Asm, Fixup.getKind());
unsigned IsRIPRel = isFixupKindRIPRel(Fixup.getKind());
unsigned Log2Size = getFixupKindLog2Size(Fixup.getKind());
@ -113,7 +117,6 @@ void X86MachObjectWriter::RecordX86_64Relocation(
unsigned Index = 0;
unsigned IsExtern = 0;
unsigned Type = 0;
const MCSymbolData *RelSymbol = nullptr;
Value = Target.getConstant();
@ -129,6 +132,7 @@ void X86MachObjectWriter::RecordX86_64Relocation(
if (Target.isAbsolute()) { // constant
// SymbolNum of 0 indicates the absolute section.
Type = MachO::X86_64_RELOC_UNSIGNED;
Index = 0;
// FIXME: I believe this is broken, I don't think the linker can understand
// it. I think it would require a local relocation, but I'm not sure if that
@ -189,30 +193,36 @@ void X86MachObjectWriter::RecordX86_64Relocation(
Value -= Writer->getSymbolAddress(&B_SD, Layout) -
(!B_Base ? 0 : Writer->getSymbolAddress(B_Base, Layout));
if (!A_Base)
if (A_Base) {
Index = A_Base->getIndex();
IsExtern = 1;
} else {
Index = A_SD.getFragment()->getParent()->getOrdinal() + 1;
IsExtern = 0;
}
Type = MachO::X86_64_RELOC_UNSIGNED;
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 =
(Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
Writer->addRelocation(A_Base, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
(IsExtern << 27) |
(Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
if (B_Base)
RelSymbol = B_Base;
else
if (B_Base) {
Index = B_Base->getIndex();
IsExtern = 1;
} else {
Index = B_SD.getFragment()->getParent()->getOrdinal() + 1;
IsExtern = 0;
}
Type = MachO::X86_64_RELOC_SUBTRACTOR;
} else {
const MCSymbol *Symbol = &Target.getSymA()->getSymbol();
if (Symbol->isTemporary() && Value) {
const MCSection &Sec = Symbol->getSection();
if (!Asm.getContext().getAsmInfo()->isSectionAtomizableBySymbols(Sec))
Asm.addLocalUsedInReloc(*Symbol);
}
const MCSymbolData &SD = Asm.getSymbolData(*Symbol);
RelSymbol = Asm.getAtom(&SD);
const MCSymbolData *Base = Asm.getAtom(&SD);
// Relocations inside debug sections always use local relocations when
// possible. This seems to be done because the debugger doesn't fully
@ -222,20 +232,23 @@ void X86MachObjectWriter::RecordX86_64Relocation(
const MCSectionMachO &Section = static_cast<const MCSectionMachO&>(
Fragment->getParent()->getSection());
if (Section.hasAttribute(MachO::S_ATTR_DEBUG))
RelSymbol = nullptr;
Base = nullptr;
}
// x86_64 almost always uses external relocations, except when there is no
// symbol to use as a base address (a local symbol with no preceding
// non-local symbol).
if (RelSymbol) {
if (Base) {
Index = Base->getIndex();
IsExtern = 1;
// Add the local offset, if needed.
if (RelSymbol != &SD)
Value +=
Layout.getSymbolOffset(&SD) - Layout.getSymbolOffset(RelSymbol);
if (Base != &SD)
Value += Layout.getSymbolOffset(&SD) - Layout.getSymbolOffset(Base);
} else if (Symbol->isInSection() && !Symbol->isVariable()) {
// The index is the section ordinal (1-based).
Index = SD.getFragment()->getParent()->getOrdinal() + 1;
IsExtern = 0;
Value += Writer->getSymbolAddress(&SD, Layout);
if (IsPCRel)
@ -334,9 +347,12 @@ void X86MachObjectWriter::RecordX86_64Relocation(
// struct relocation_info (8 bytes)
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 = (Index << 0) | (IsPCRel << 24) | (Log2Size << 25) |
(IsExtern << 27) | (Type << 28);
Writer->addRelocation(RelSymbol, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
(IsExtern << 27) |
(Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
}
bool X86MachObjectWriter::RecordScatteredRelocation(MachObjectWriter *Writer,
@ -408,7 +424,7 @@ bool X86MachObjectWriter::RecordScatteredRelocation(MachObjectWriter *Writer,
(IsPCRel << 30) |
MachO::R_SCATTERED);
MRE.r_word1 = Value2;
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
} else {
// If the offset is more than 24-bits, it won't fit in a scattered
// relocation offset field, so we fall back to using a non-scattered
@ -430,7 +446,7 @@ bool X86MachObjectWriter::RecordScatteredRelocation(MachObjectWriter *Writer,
(IsPCRel << 30) |
MachO::R_SCATTERED);
MRE.r_word1 = Value;
Writer->addRelocation(nullptr, Fragment->getParent(), MRE);
Writer->addRelocation(Fragment->getParent(), MRE);
return true;
}
@ -451,6 +467,7 @@ void X86MachObjectWriter::RecordTLVPRelocation(MachObjectWriter *Writer,
// Get the symbol data.
const MCSymbolData *SD_A = &Asm.getSymbolData(Target.getSymA()->getSymbol());
unsigned Index = SD_A->getIndex();
// We're only going to have a second symbol in pic mode and it'll be a
// subtraction from the picbase. For 32-bit pic the addend is the difference
@ -473,9 +490,12 @@ void X86MachObjectWriter::RecordTLVPRelocation(MachObjectWriter *Writer,
// struct relocation_info (8 bytes)
MachO::any_relocation_info MRE;
MRE.r_word0 = Value;
MRE.r_word1 =
(IsPCRel << 24) | (Log2Size << 25) | (MachO::GENERIC_RELOC_TLV << 28);
Writer->addRelocation(SD_A, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
(1 << 27) | // r_extern
(MachO::GENERIC_RELOC_TLV << 28)); // r_type
Writer->addRelocation(Fragment->getParent(), MRE);
}
void X86MachObjectWriter::RecordX86Relocation(MachObjectWriter *Writer,
@ -526,8 +546,8 @@ void X86MachObjectWriter::RecordX86Relocation(MachObjectWriter *Writer,
// See <reloc.h>.
uint32_t FixupOffset = Layout.getFragmentOffset(Fragment)+Fixup.getOffset();
unsigned Index = 0;
unsigned IsExtern = 0;
unsigned Type = 0;
const MCSymbolData *RelSymbol = nullptr;
if (Target.isAbsolute()) { // constant
// SymbolNum of 0 indicates the absolute section.
@ -548,7 +568,8 @@ void X86MachObjectWriter::RecordX86Relocation(MachObjectWriter *Writer,
// Check whether we need an external or internal relocation.
if (Writer->doesSymbolRequireExternRelocation(SD)) {
RelSymbol = SD;
IsExtern = 1;
Index = SD->getIndex();
// For external relocations, make sure to offset the fixup value to
// compensate for the addend of the symbol address, if it was
// undefined. This occurs with weak definitions, for example.
@ -570,9 +591,12 @@ void X86MachObjectWriter::RecordX86Relocation(MachObjectWriter *Writer,
// struct relocation_info (8 bytes)
MachO::any_relocation_info MRE;
MRE.r_word0 = FixupOffset;
MRE.r_word1 =
(Index << 0) | (IsPCRel << 24) | (Log2Size << 25) | (Type << 28);
Writer->addRelocation(RelSymbol, Fragment->getParent(), MRE);
MRE.r_word1 = ((Index << 0) |
(IsPCRel << 24) |
(Log2Size << 25) |
(IsExtern << 27) |
(Type << 28));
Writer->addRelocation(Fragment->getParent(), MRE);
}
MCObjectWriter *llvm::createX86MachObjectWriter(raw_ostream &OS,

4
contrib/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
View File

@ -1376,6 +1376,10 @@ bool InstCombiner::transformConstExprCastCall(CallSite CS) {
dyn_cast<Function>(CS.getCalledValue()->stripPointerCasts());
if (!Callee)
return false;
// The prototype of thunks are a lie, don't try to directly call such
// functions.
if (Callee->hasFnAttribute("thunk"))
return false;
Instruction *Caller = CS.getInstruction();
const AttributeSet &CallerPAL = CS.getAttributes();

13
contrib/llvm/lib/Transforms/Scalar/GVN.cpp
View File

@ -2182,9 +2182,16 @@ bool GVN::propagateEquality(Value *LHS, Value *RHS,
// Handle the floating point versions of equality comparisons too.
if ((isKnownTrue && Cmp->getPredicate() == CmpInst::FCMP_OEQ) ||
(isKnownFalse && Cmp->getPredicate() == CmpInst::FCMP_UNE))
Worklist.push_back(std::make_pair(Op0, Op1));
(isKnownFalse && Cmp->getPredicate() == CmpInst::FCMP_UNE)) {
// Floating point -0.0 and 0.0 compare equal, so we can't
// propagate a constant based on that comparison.
// FIXME: We should do this optimization if 'no signed zeros' is
// applicable via an instruction-level fast-math-flag or some other
// indicator that relaxed FP semantics are being used.
if (!isa<ConstantFP>(Op1) || !cast<ConstantFP>(Op1)->isZero())
Worklist.push_back(std::make_pair(Op0, Op1));
}
// If "A >= B" is known true, replace "A < B" with false everywhere.
CmpInst::Predicate NotPred = Cmp->getInversePredicate();
Constant *NotVal = ConstantInt::get(Cmp->getType(), isKnownFalse);

3
contrib/llvm/lib/Transforms/Scalar/LoopInstSimplify.cpp
View File

@ -18,6 +18,7 @@
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopPass.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Instructions.h"
@ -47,7 +48,7 @@ namespace {
AU.addRequiredID(LoopSimplifyID);
AU.addPreservedID(LoopSimplifyID);
AU.addPreservedID(LCSSAID);
AU.addPreserved("scalar-evolution");
AU.addPreserved<ScalarEvolution>();
AU.addRequired<TargetLibraryInfo>();
}
};

1
contrib/llvm/lib/Transforms/Utils/LowerSwitch.cpp
View File

@ -46,7 +46,6 @@ namespace {
void getAnalysisUsage(AnalysisUsage &AU) const override {
// This is a cluster of orthogonal Transforms
AU.addPreserved<UnifyFunctionExitNodes>();
AU.addPreserved("mem2reg");
AU.addPreservedID(LowerInvokePassID);
}

12
contrib/llvm/lib/Transforms/Utils/SimplifyIndVar.cpp
View File

@ -278,9 +278,8 @@ bool SimplifyIndvar::strengthenOverflowingOperation(BinaryOperator *BO,
Value *IVOperand) {
// Currently we only handle instructions of the form "add <indvar> <value>"
// and "sub <indvar> <value>".
unsigned Op = BO->getOpcode();
if (!(Op == Instruction::Add || Op == Instruction::Sub))
if (Op != Instruction::Add)
return false;
// If BO is already both nuw and nsw then there is nothing left to do
@ -304,15 +303,6 @@ bool SimplifyIndvar::strengthenOverflowingOperation(BinaryOperator *BO,
if (OtherOpSCEV == SE->getCouldNotCompute())
return false;
if (Op == Instruction::Sub) {
// If the subtraction is of the form "sub <indvar>, <op>", then pretend it
// is "add <indvar>, -<op>" and continue, else bail out.
if (OtherOperandIdx != 1)
return false;
OtherOpSCEV = SE->getNegativeSCEV(OtherOpSCEV);
}
const SCEV *IVOpSCEV = SE->getSCEV(IVOperand);
const SCEV *ZeroSCEV = SE->getConstant(IVOpSCEV->getType(), 0);

24
contrib/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
View File

@ -1968,8 +1968,12 @@ Value *LibCallSimplifier::optimizeCall(CallInst *CI) {
// Try to further simplify the result.
CallInst *SimplifiedCI = dyn_cast<CallInst>(SimplifiedFortifiedCI);
if (SimplifiedCI && SimplifiedCI->getCalledFunction())
if (Value *V = optimizeStringMemoryLibCall(SimplifiedCI, Builder))
if (Value *V = optimizeStringMemoryLibCall(SimplifiedCI, Builder)) {
// If we were able to further simplify, remove the now redundant call.
SimplifiedCI->replaceAllUsesWith(V);
SimplifiedCI->eraseFromParent();
return V;
}
return SimplifiedFortifiedCI;
}
@ -2218,11 +2222,11 @@ Value *FortifiedLibCallSimplifier::optimizeMemSetChk(CallInst *CI, IRBuilder<> &
return nullptr;
}
Value *FortifiedLibCallSimplifier::optimizeStrCpyChk(CallInst *CI, IRBuilder<> &B) {
Value *FortifiedLibCallSimplifier::optimizeStrpCpyChk(CallInst *CI,
IRBuilder<> &B,
LibFunc::Func Func) {
Function *Callee = CI->getCalledFunction();
StringRef Name = Callee->getName();
LibFunc::Func Func =
Name.startswith("str") ? LibFunc::strcpy_chk : LibFunc::stpcpy_chk;
if (!checkStringCopyLibFuncSignature(Callee, Func, DL))
return nullptr;
@ -2231,7 +2235,7 @@ Value *FortifiedLibCallSimplifier::optimizeStrCpyChk(CallInst *CI, IRBuilder<> &
*ObjSize = CI->getArgOperand(2);
// __stpcpy_chk(x,x,...) -> x+strlen(x)
if (!OnlyLowerUnknownSize && Dst == Src) {
if (Func == LibFunc::stpcpy_chk && !OnlyLowerUnknownSize && Dst == Src) {
Value *StrLen = EmitStrLen(Src, B, DL, TLI);
return StrLen ? B.CreateInBoundsGEP(Dst, StrLen) : nullptr;
}
@ -2266,11 +2270,11 @@ Value *FortifiedLibCallSimplifier::optimizeStrCpyChk(CallInst *CI, IRBuilder<> &
return nullptr;
}
Value *FortifiedLibCallSimplifier::optimizeStrNCpyChk(CallInst *CI, IRBuilder<> &B) {
Value *FortifiedLibCallSimplifier::optimizeStrpNCpyChk(CallInst *CI,
IRBuilder<> &B,
LibFunc::Func Func) {
Function *Callee = CI->getCalledFunction();
StringRef Name = Callee->getName();
LibFunc::Func Func =
Name.startswith("str") ? LibFunc::strncpy_chk : LibFunc::stpncpy_chk;
if (!checkStringCopyLibFuncSignature(Callee, Func, DL))
return nullptr;
@ -2310,10 +2314,10 @@ Value *FortifiedLibCallSimplifier::optimizeCall(CallInst *CI) {
return optimizeMemSetChk(CI, Builder);
case LibFunc::stpcpy_chk:
case LibFunc::strcpy_chk:
return optimizeStrCpyChk(CI, Builder);
return optimizeStrpCpyChk(CI, Builder, Func);
case LibFunc::stpncpy_chk:
case LibFunc::strncpy_chk:
return optimizeStrNCpyChk(CI, Builder);
return optimizeStrpNCpyChk(CI, Builder, Func);
default:
break;
}

23
contrib/llvm/lib/Transforms/Utils/SymbolRewriter.cpp
View File

@ -79,6 +79,19 @@ static cl::list<std::string> RewriteMapFiles("rewrite-map-file",
namespace llvm {
namespace SymbolRewriter {
void rewriteComdat(Module &M, GlobalObject *GO, const std::string &Source,
const std::string &Target) {
if (Comdat *CD = GO->getComdat()) {
auto &Comdats = M.getComdatSymbolTable();
Comdat *C = M.getOrInsertComdat(Target);
C->setSelectionKind(CD->getSelectionKind());
GO->setComdat(C);
Comdats.erase(Comdats.find(Source));
}
}
template <RewriteDescriptor::Type DT, typename ValueType,
ValueType *(llvm::Module::*Get)(StringRef) const>
class ExplicitRewriteDescriptor : public RewriteDescriptor {
@ -102,10 +115,14 @@ template <RewriteDescriptor::Type DT, typename ValueType,
bool ExplicitRewriteDescriptor<DT, ValueType, Get>::performOnModule(Module &M) {
bool Changed = false;
if (ValueType *S = (M.*Get)(Source)) {
if (GlobalObject *GO = dyn_cast<GlobalObject>(S))
rewriteComdat(M, GO, Source, Target);
if (Value *T = (M.*Get)(Target))
S->setValueName(T->getValueName());
else
S->setName(Target);
Changed = true;
}
return Changed;
@ -145,6 +162,12 @@ performOnModule(Module &M) {
report_fatal_error("unable to transforn " + C.getName() + " in " +
M.getModuleIdentifier() + ": " + Error);
if (C.getName() == Name)
continue;
if (GlobalObject *GO = dyn_cast<GlobalObject>(&C))
rewriteComdat(M, GO, C.getName(), Name);
if (Value *V = (M.*Get)(Name))
C.setValueName(V->getValueName());
else

1
contrib/llvm/lib/Transforms/Utils/UnifyFunctionExitNodes.cpp
View File

@ -35,7 +35,6 @@ void UnifyFunctionExitNodes::getAnalysisUsage(AnalysisUsage &AU) const{
// We preserve the non-critical-edgeness property
AU.addPreservedID(BreakCriticalEdgesID);
// This is a cluster of orthogonal Transforms
AU.addPreserved("mem2reg");
AU.addPreservedID(LowerSwitchID);
}

10
contrib/llvm/patches/README.TXT
View File

@ -1,11 +1,11 @@
This is a set of individual patches, which contain all the customizations to
llvm/clang currently in the FreeBSD base system. These can be applied in
alphabetical order to a pristine llvm/clang 3.6.0 release source tree, for
example by doing:
alphabetical order to a pristine llvm/clang 3.6.0 RC2 source tree, for example
by doing:
svn co https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_351/rc1 llvm-3.6.0-rc1
svn co https://llvm.org/svn/llvm-project/cfe/tags/RELEASE_351/rc1 llvm-3.6.0-rc1/tools/clang
cd llvm-3.6.0-rc1
svn co https://llvm.org/svn/llvm-project/llvm/tags/RELEASE_351/rc2 llvm-3.6.0-rc2
svn co https://llvm.org/svn/llvm-project/cfe/tags/RELEASE_351/rc2 llvm-3.6.0-rc2/tools/clang
cd llvm-3.6.0-rc2
for p in /usr/src/contrib/llvm/patches/patch-*.diff; do
patch -p0 -f -F0 -E -i $p -s || break
done

83
contrib/llvm/patches/patch-06-llvm-r226664-aarch64-x18.diff
View File

@ -1,83 +0,0 @@
Pull in r226664 from upstream llvm trunk (by Tim Northover):
AArch64: add backend option to reserve x18 (platform register)
AAPCS64 says that it's up to the platform to specify whether x18 is
reserved, and a first step on that way is to add a flag controlling
it.
From: Andrew Turner <andrew@fubar.geek.nz>
Introduced here: http://svnweb.freebsd.org/changeset/base/277774
Index: lib/Target/AArch64/AArch64RegisterInfo.cpp
===================================================================
--- lib/Target/AArch64/AArch64RegisterInfo.cpp
+++ lib/Target/AArch64/AArch64RegisterInfo.cpp
@@ -33,6 +33,10 @@ using namespace llvm;
#define GET_REGINFO_TARGET_DESC
#include "AArch64GenRegisterInfo.inc"
+static cl::opt<bool>
+ReserveX18("aarch64-reserve-x18", cl::Hidden,
+ cl::desc("Reserve X18, making it unavailable as GPR"));
+
AArch64RegisterInfo::AArch64RegisterInfo(const AArch64InstrInfo *tii,
const AArch64Subtarget *sti)
: AArch64GenRegisterInfo(AArch64::LR), TII(tii), STI(sti) {}
@@ -90,7 +94,7 @@ AArch64RegisterInfo::getReservedRegs(const Machine
Reserved.set(AArch64::W29);
}
- if (STI->isTargetDarwin()) {
+ if (STI->isTargetDarwin() || ReserveX18) {
Reserved.set(AArch64::X18); // Platform register
Reserved.set(AArch64::W18);
}
@@ -117,7 +121,7 @@ bool AArch64RegisterInfo::isReservedReg(const Mach
return true;
case AArch64::X18:
case AArch64::W18:
- return STI->isTargetDarwin();
+ return STI->isTargetDarwin() || ReserveX18;
case AArch64::FP:
case AArch64::W29:
return TFI->hasFP(MF) || STI->isTargetDarwin();
@@ -379,7 +383,7 @@ unsigned AArch64RegisterInfo::getRegPressureLimit(
case AArch64::GPR64commonRegClassID:
return 32 - 1 // XZR/SP
- (TFI->hasFP(MF) || STI->isTargetDarwin()) // FP
- - STI->isTargetDarwin() // X18 reserved as platform register
+ - (STI->isTargetDarwin() || ReserveX18) // X18 reserved as platform register
- hasBasePointer(MF); // X19
case AArch64::FPR8RegClassID:
case AArch64::FPR16RegClassID:
Index: test/CodeGen/AArch64/arm64-platform-reg.ll
===================================================================
--- test/CodeGen/AArch64/arm64-platform-reg.ll
+++ test/CodeGen/AArch64/arm64-platform-reg.ll
@@ -1,4 +1,5 @@
-; RUN: llc -mtriple=arm64-apple-ios -o - %s | FileCheck %s --check-prefix=CHECK-DARWIN
+; RUN: llc -mtriple=arm64-apple-ios -o - %s | FileCheck %s --check-prefix=CHECK-RESERVE-X18
+; RUN: llc -mtriple=arm64-freebsd-gnu -aarch64-reserve-x18 -o - %s | FileCheck %s --check-prefix=CHECK-RESERVE-X18
; RUN: llc -mtriple=arm64-linux-gnu -o - %s | FileCheck %s
; x18 is reserved as a platform register on Darwin but not on other
@@ -16,11 +17,11 @@ define void @keep_live() {
; CHECK: ldr x18
; CHECK: str x18
-; CHECK-DARWIN-NOT: ldr fp
-; CHECK-DARWIN-NOT: ldr x18
-; CHECK-DARWIN: Spill
-; CHECK-DARWIN-NOT: ldr fp
-; CHECK-DARWIN-NOT: ldr x18
-; CHECK-DARWIN: ret
+; CHECK-RESERVE-X18-NOT: ldr fp
+; CHECK-RESERVE-X18-NOT: ldr x18
+; CHECK-RESERVE-X18: Spill
+; CHECK-RESERVE-X18-NOT: ldr fp
+; CHECK-RESERVE-X18-NOT: ldr x18
+; CHECK-RESERVE-X18: ret
ret void
}

53
contrib/llvm/patches/patch-07-clang-r227062-fixes-x18.diff
View File

@ -1,53 +0,0 @@
Pull in r227062 from upstream clang trunk (by Renato Golin):
Allows Clang to use LLVM's fixes-x18 option
This patch allows clang to have llvm reserve the x18
platform register on AArch64. FreeBSD will use this in the kernel for
per-cpu data but has no need to reserve this register in userland so
will need this flag to reserve it.
This uses llvm r226664 to allow this register to be reserved.
Patch by Andrew Turner.
Introduced here: http://svnweb.freebsd.org/changeset/base/277775
Index: tools/clang/include/clang/Driver/Options.td
===================================================================
--- tools/clang/include/clang/Driver/Options.td
+++ tools/clang/include/clang/Driver/Options.td
@@ -1209,6 +1209,8 @@ def mfix_cortex_a53_835769 : Flag<["-"], "mfix-cor
def mno_fix_cortex_a53_835769 : Flag<["-"], "mno-fix-cortex-a53-835769">,
Group<m_aarch64_Features_Group>,
HelpText<"Don't workaround Cortex-A53 erratum 835769 (AArch64 only)">;
+def ffixed_x18 : Flag<["-"], "ffixed-x18">, Group<m_aarch64_Features_Group>,
+ HelpText<"Reserve the x18 register (AArch64 only)">;
def mvsx : Flag<["-"], "mvsx">, Group<m_ppc_Features_Group>;
def mno_vsx : Flag<["-"], "mno-vsx">, Group<m_ppc_Features_Group>;
Index: tools/clang/lib/Driver/Tools.cpp
===================================================================
--- tools/clang/lib/Driver/Tools.cpp
+++ tools/clang/lib/Driver/Tools.cpp
@@ -958,6 +958,11 @@ void Clang::AddAArch64TargetArgs(const ArgList &Ar
if (A->getOption().matches(options::OPT_mno_global_merge))
CmdArgs.push_back("-mno-global-merge");
}
+
+ if (Args.hasArg(options::OPT_ffixed_x18)) {
+ CmdArgs.push_back("-backend-option");
+ CmdArgs.push_back("-aarch64-reserve-x18");
+ }
}
// Get CPU and ABI names. They are not independent
Index: tools/clang/test/Driver/aarch64-fixed-x18.c
===================================================================
--- tools/clang/test/Driver/aarch64-fixed-x18.c
+++ tools/clang/test/Driver/aarch64-fixed-x18.c
@@ -0,0 +1,4 @@
+// RUN: %clang -target aarch64-none-gnu -ffixed-x18 -### %s 2> %t
+// RUN: FileCheck --check-prefix=CHECK-FIXED-X18 < %t %s
+
+// CHECK-FIXED-X18: "-backend-option" "-aarch64-reserve-x18"

23
contrib/llvm/tools/clang/include/clang/Basic/Attr.td
View File

@ -224,14 +224,12 @@ class SubjectList<list<AttrSubject> subjects, SubjectDiag diag = WarnDiag,
string CustomDiag = customDiag;
}
class LangOpt<string name, bit negated = 0> {
class LangOpt<string name> {
string Name = name;
bit Negated = negated;
}
def MicrosoftExt : LangOpt<"MicrosoftExt">;
def Borland : LangOpt<"Borland">;
def CUDA : LangOpt<"CUDA">;
def COnly : LangOpt<"CPlusPlus", 1>;
// Defines targets for target-specific attributes. The list of strings should
// specify architectures for which the target applies, based off the ArchType
@ -709,25 +707,6 @@ def MinSize : InheritableAttr {
let Documentation = [Undocumented];
}
def FlagEnum : InheritableAttr {
let Spellings = [GNU<"flag_enum">];
let Subjects = SubjectList<[Enum]>;
let Documentation = [FlagEnumDocs];
let LangOpts = [COnly];
let AdditionalMembers = [{
private:
llvm::APInt FlagBits;
public:
llvm::APInt &getFlagBits() {
return FlagBits;
}
const llvm::APInt &getFlagBits() const {
return FlagBits;
}
}];
}
def Flatten : InheritableAttr {
let Spellings = [GCC<"flatten">];
let Subjects = SubjectList<[Function], ErrorDiag>;

10
contrib/llvm/tools/clang/include/clang/Basic/AttrDocs.td
View File

@ -1196,16 +1196,6 @@ behavior of the program is undefined.
}];
}
def FlagEnumDocs : Documentation {
let Category = DocCatType;
let Content = [{
This attribute can be added to an enumerator to signal to the compiler that it
is intended to be used as a flag type. This will cause the compiler to assume
that the range of the type includes all of the values that you can get by
manipulating bits of the enumerator when issuing warnings.
}];
}
def MSInheritanceDocs : Documentation {
let Category = DocCatType;
let Heading = "__single_inhertiance, __multiple_inheritance, __virtual_inheritance";

1
contrib/llvm/tools/clang/include/clang/Basic/DiagnosticGroups.td
View File

@ -190,7 +190,6 @@ def OverloadedShiftOpParentheses: DiagGroup<"overloaded-shift-op-parentheses">;
def DanglingElse: DiagGroup<"dangling-else">;
def DanglingField : DiagGroup<"dangling-field">;
def DistributedObjectModifiers : DiagGroup<"distributed-object-modifiers">;
def FlagEnum : DiagGroup<"flag-enum">;
def InfiniteRecursion : DiagGroup<"infinite-recursion">;
def GNUImaginaryConstant : DiagGroup<"gnu-imaginary-constant">;
def IgnoredQualifiers : DiagGroup<"ignored-qualifiers">;

7
contrib/llvm/tools/clang/include/clang/Basic/DiagnosticSemaKinds.td
View File

@ -2236,7 +2236,7 @@ def warn_attribute_wrong_decl_type : Warning<
"%0 attribute only applies to %select{functions|unions|"
"variables and functions|functions and methods|parameters|"
"functions, methods and blocks|functions, methods, and classes|"
"functions, methods, and parameters|classes|enums|variables|methods|"
"functions, methods, and parameters|classes|variables|methods|"
"variables, functions and labels|fields and global variables|structs|"
"variables and typedefs|thread-local variables|"
"variables and fields|variables, data members and tag types|"
@ -4059,9 +4059,6 @@ def ext_enum_too_large : ExtWarn<
def ext_enumerator_increment_too_large : ExtWarn<
"incremented enumerator value %0 is not representable in the "
"largest integer type">, InGroup<EnumTooLarge>;
def warn_flag_enum_constant_out_of_range : Warning<
"enumeration value %0 is out of range of flags in enumeration type %1">,
InGroup<FlagEnum>;
def warn_illegal_constant_array_size : Extension<
"size of static array must be an integer constant expression">;
@ -6162,6 +6159,8 @@ let CategoryName = "Inline Assembly Issue" in {
"invalid lvalue in asm input for constraint '%0'">;
def err_asm_invalid_input_constraint : Error<
"invalid input constraint '%0' in asm">;
def err_asm_immediate_expected : Error<"constraint '%0' expects "
"an integer constant expression">;
def err_asm_invalid_type_in_input : Error<
"invalid type %0 in asm input for constraint '%1'">;
def err_asm_tying_incompatible_types : Error<

1
contrib/llvm/tools/clang/include/clang/Sema/AttributeList.h
View File

@ -822,7 +822,6 @@ enum AttributeDeclKind {
ExpectedFunctionMethodOrClass,
ExpectedFunctionMethodOrParameter,
ExpectedClass,
ExpectedEnum,
ExpectedVariable,
ExpectedMethod,
ExpectedVariableFunctionOrLabel,

6
contrib/llvm/tools/clang/include/clang/Sema/Sema.h
View File

@ -7971,12 +7971,6 @@ public:
Expr *SrcExpr, AssignmentAction Action,
bool *Complained = nullptr);
/// IsValueInFlagEnum - Determine if a value is allowed as part of a flag
/// enum. If AllowMask is true, then we also allow the complement of a valid
/// value, to be used as a mask.
bool IsValueInFlagEnum(const EnumDecl *ED, const llvm::APInt &Val,
bool AllowMask) const;
/// DiagnoseAssignmentEnum - Warn if assignment to enum is a constant
/// integer not in the range of enum values.
void DiagnoseAssignmentEnum(QualType DstType, QualType SrcType,

14
contrib/llvm/tools/clang/include/clang/Sema/SemaInternal.h
View File

@ -101,7 +101,7 @@ public:
DeclContext *MemberContext,
bool EnteringContext)
: Typo(TypoName.getName().getAsIdentifierInfo()), CurrentTCIndex(0),
SemaRef(SemaRef), S(S),
SavedTCIndex(0), SemaRef(SemaRef), S(S),
SS(SS ? llvm::make_unique<CXXScopeSpec>(*SS) : nullptr),
CorrectionValidator(std::move(CCC)), MemberContext(MemberContext),
Result(SemaRef, TypoName, LookupKind),
@ -187,6 +187,17 @@ public:
CurrentTCIndex >= ValidatedCorrections.size();
}
/// \brief Save the current position in the correction stream (overwriting any
/// previously saved position).
void saveCurrentPosition() {
SavedTCIndex = CurrentTCIndex;
}
/// \brief Restore the saved position in the correction stream.
void restoreSavedPosition() {
CurrentTCIndex = SavedTCIndex;
}
ASTContext &getContext() const { return SemaRef.Context; }
const LookupResult &getLookupResult() const { return Result; }
@ -267,6 +278,7 @@ private:
SmallVector<TypoCorrection, 4> ValidatedCorrections;
size_t CurrentTCIndex;
size_t SavedTCIndex;
Sema &SemaRef;
Scope *S;

10
contrib/llvm/tools/clang/lib/Basic/Targets.cpp
View File

@ -5901,6 +5901,8 @@ public:
: MipsTargetInfoBase(Triple, "o32", "mips32r2") {
SizeType = UnsignedInt;
PtrDiffType = SignedInt;
Int64Type = SignedLongLong;
IntMaxType = Int64Type;
MaxAtomicPromoteWidth = MaxAtomicInlineWidth = 32;
}
bool setABI(const std::string &Name) override {
@ -5922,6 +5924,8 @@ public:
Builder.defineMacro("__mips_isa_rev", "1");
else if (CPUStr == "mips32r2")
Builder.defineMacro("__mips_isa_rev", "2");
else if (CPUStr == "mips32r6")
Builder.defineMacro("__mips_isa_rev", "6");
if (ABI == "o32") {
Builder.defineMacro("__mips_o32");
@ -6028,6 +6032,8 @@ public:
PointerWidth = PointerAlign = 64;
SizeType = UnsignedLong;
PtrDiffType = SignedLong;
Int64Type = SignedLong;
IntMaxType = Int64Type;
}
void setN32ABITypes() {
@ -6035,6 +6041,8 @@ public:
PointerWidth = PointerAlign = 32;
SizeType = UnsignedInt;
PtrDiffType = SignedInt;
Int64Type = SignedLongLong;
IntMaxType = Int64Type;
}
bool setABI(const std::string &Name) override {
@ -6065,6 +6073,8 @@ public:
Builder.defineMacro("__mips_isa_rev", "1");
else if (CPUStr == "mips64r2")
Builder.defineMacro("__mips_isa_rev", "2");
else if (CPUStr == "mips64r6")
Builder.defineMacro("__mips_isa_rev", "6");
if (ABI == "n32") {
Builder.defineMacro("__mips_n32");

3
contrib/llvm/tools/clang/lib/CodeGen/CGDecl.cpp
View File

@ -206,9 +206,6 @@ llvm::Constant *CodeGenModule::getOrCreateStaticVarDecl(
GV->setAlignment(getContext().getDeclAlign(&D).getQuantity());
setGlobalVisibility(GV, &D);
if (supportsCOMDAT() && GV->isWeakForLinker())
GV->setComdat(TheModule.getOrInsertComdat(GV->getName()));
if (D.getTLSKind())
setTLSMode(GV, D);

10
contrib/llvm/tools/clang/lib/CodeGen/CGDeclCXX.cpp
View File

@ -267,7 +267,15 @@ void CodeGenModule::EmitPointerToInitFunc(const VarDecl *D,
addUsedGlobal(PtrArray);
// If the GV is already in a comdat group, then we have to join it.
if (llvm::Comdat *C = GV->getComdat())
llvm::Comdat *C = GV->getComdat();
// LinkOnce and Weak linkage are lowered down to a single-member comdat group.
// Make an explicit group so we can join it.
if (!C && (GV->hasWeakLinkage() || GV->hasLinkOnceLinkage())) {
C = TheModule.getOrInsertComdat(GV->getName());
GV->setComdat(C);
}
if (C)
PtrArray->setComdat(C);
}

31
contrib/llvm/tools/clang/lib/CodeGen/CodeGenModule.cpp
View File

@ -1928,31 +1928,6 @@ void CodeGenModule::MaybeHandleStaticInExternC(const SomeDecl *D,
R.first->second = nullptr;
}
static bool shouldBeInCOMDAT(CodeGenModule &CGM, const Decl &D) {
if (!CGM.supportsCOMDAT())
return false;
if (D.hasAttr<SelectAnyAttr>())
return true;
GVALinkage Linkage;
if (auto *VD = dyn_cast<VarDecl>(&D))
Linkage = CGM.getContext().GetGVALinkageForVariable(VD);
else
Linkage = CGM.getContext().GetGVALinkageForFunction(cast<FunctionDecl>(&D));
switch (Linkage) {
case GVA_Internal:
case GVA_AvailableExternally:
case GVA_StrongExternal:
return false;
case GVA_DiscardableODR:
case GVA_StrongODR:
return true;
}
llvm_unreachable("No such linkage");
}
void CodeGenModule::EmitGlobalVarDefinition(const VarDecl *D) {
llvm::Constant *Init = nullptr;
QualType ASTTy = D->getType();
@ -2096,9 +2071,6 @@ void CodeGenModule::EmitGlobalVarDefinition(const VarDecl *D) {
setTLSMode(GV, *D);
}
if (shouldBeInCOMDAT(*this, *D))
GV->setComdat(TheModule.getOrInsertComdat(GV->getName()));
// Emit the initializer function if necessary.
if (NeedsGlobalCtor || NeedsGlobalDtor)
EmitCXXGlobalVarDeclInitFunc(D, GV, NeedsGlobalCtor);
@ -2433,9 +2405,6 @@ void CodeGenModule::EmitGlobalFunctionDefinition(GlobalDecl GD,
MaybeHandleStaticInExternC(D, Fn);
if (shouldBeInCOMDAT(*this, *D))
Fn->setComdat(TheModule.getOrInsertComdat(Fn->getName()));
CodeGenFunction(*this).GenerateCode(D, Fn, FI);
setFunctionDefinitionAttributes(D, Fn);

7
contrib/llvm/tools/clang/lib/CodeGen/ItaniumCXXABI.cpp
View File

@ -1711,12 +1711,11 @@ void ItaniumCXXABI::EmitGuardedInit(CodeGenFunction &CGF,
// The ABI says: It is suggested that it be emitted in the same COMDAT group
// as the associated data object
llvm::Comdat *C = var->getComdat();
if (!D.isLocalVarDecl() && C) {
if (!D.isLocalVarDecl() && var->isWeakForLinker() && CGM.supportsCOMDAT()) {
llvm::Comdat *C = CGM.getModule().getOrInsertComdat(var->getName());
guard->setComdat(C);
var->setComdat(C);
CGF.CurFn->setComdat(C);
} else if (CGM.supportsCOMDAT() && guard->isWeakForLinker()) {
guard->setComdat(CGM.getModule().getOrInsertComdat(guard->getName()));
}
CGM.setStaticLocalDeclGuardAddress(&D, guard);

18
contrib/llvm/tools/clang/lib/CodeGen/MicrosoftCXXABI.cpp
View File

@ -1534,6 +1534,12 @@ llvm::Function *MicrosoftCXXABI::EmitVirtualMemPtrThunk(
CGM.SetLLVMFunctionAttributes(MD, FnInfo, ThunkFn);
CGM.SetLLVMFunctionAttributesForDefinition(MD, ThunkFn);
// Add the "thunk" attribute so that LLVM knows that the return type is
// meaningless. These thunks can be used to call functions with differing
// return types, and the caller is required to cast the prototype
// appropriately to extract the correct value.
ThunkFn->addFnAttr("thunk");
// These thunks can be compared, so they are not unnamed.
ThunkFn->setUnnamedAddr(false);
@ -1829,10 +1835,18 @@ void MicrosoftCXXABI::EmitThreadLocalInitFuncs(
llvm::Function *F = CXXThreadLocalInits[I];
// If the GV is already in a comdat group, then we have to join it.
if (llvm::Comdat *C = GV->getComdat())
llvm::Comdat *C = GV->getComdat();
// LinkOnce and Weak linkage are lowered down to a single-member comdat
// group.
// Make an explicit group so we can join it.
if (!C && (GV->hasWeakLinkage() || GV->hasLinkOnceLinkage())) {
C = CGM.getModule().getOrInsertComdat(GV->getName());
GV->setComdat(C);
AddToXDU(F)->setComdat(C);
else
} else {
NonComdatInits.push_back(F);
}
}
if (!NonComdatInits.empty()) {

19
contrib/llvm/tools/clang/lib/Driver/Tools.cpp
View File

@ -5423,6 +5423,20 @@ const char *arm::getLLVMArchSuffixForARM(StringRef CPU) {
.Default("");
}
void arm::appendEBLinkFlags(const ArgList &Args, ArgStringList &CmdArgs, const llvm::Triple &Triple) {
if (Args.hasArg(options::OPT_r))
return;
StringRef Suffix = getLLVMArchSuffixForARM(getARMCPUForMArch(Args, Triple));
const char *LinkFlag = llvm::StringSwitch<const char *>(Suffix)
.Cases("v4", "v4t", "v5", "v5e", nullptr)
.Cases("v6", "v6t2", nullptr)
.Default("--be8");
if (LinkFlag)
CmdArgs.push_back(LinkFlag);
}
bool mips::hasMipsAbiArg(const ArgList &Args, const char *Value) {
Arg *A = Args.getLastArg(options::OPT_mabi_EQ);
return A && (A->getValue() == StringRef(Value));
@ -6897,6 +6911,7 @@ void netbsd::Link::ConstructJob(Compilation &C, const JobAction &JA,
break;
case llvm::Triple::armeb:
case llvm::Triple::thumbeb:
arm::appendEBLinkFlags(Args, CmdArgs, getToolChain().getTriple());
CmdArgs.push_back("-m");
switch (getToolChain().getTriple().getEnvironment()) {
case llvm::Triple::EABI:
@ -7447,6 +7462,10 @@ void gnutools::Link::ConstructJob(Compilation &C, const JobAction &JA,
if (Args.hasArg(options::OPT_s))
CmdArgs.push_back("-s");
if (ToolChain.getArch() == llvm::Triple::armeb ||
ToolChain.getArch() == llvm::Triple::thumbeb)
arm::appendEBLinkFlags(Args, CmdArgs, getToolChain().getTriple());
for (const auto &Opt : ToolChain.ExtraOpts)
CmdArgs.push_back(Opt.c_str());

2
contrib/llvm/tools/clang/lib/Driver/Tools.h
View File

@ -228,6 +228,8 @@ namespace arm {
const char* getARMCPUForMArch(const llvm::opt::ArgList &Args,
const llvm::Triple &Triple);
const char* getLLVMArchSuffixForARM(StringRef CPU);
void appendEBLinkFlags(const llvm::opt::ArgList &Args, ArgStringList &CmdArgs, const llvm::Triple &Triple);
}
namespace mips {

5
contrib/llvm/tools/clang/lib/Sema/SemaChecking.cpp
View File

@ -1340,6 +1340,11 @@ ExprResult Sema::SemaAtomicOpsOverloaded(ExprResult TheCallResult,
<< IsC11 << Ptr->getType() << Ptr->getSourceRange();
return ExprError();
}
if (IsC11 && ValType->isPointerType() &&
RequireCompleteType(Ptr->getLocStart(), ValType->getPointeeType(),
diag::err_incomplete_type)) {
return ExprError();
}
} else if (IsN && !ValType->isIntegerType() && !ValType->isPointerType()) {
// For __atomic_*_n operations, the value type must be a scalar integral or
// pointer type which is 1, 2, 4, 8 or 16 bytes in length.

93
contrib/llvm/tools/clang/lib/Sema/SemaDecl.cpp
View File

@ -8828,11 +8828,12 @@ void Sema::AddInitializerToDecl(Decl *RealDecl, Expr *Init,
});
if (Res.isInvalid()) {
VDecl->setInvalidDecl();
return;
}
if (Res.get() != Args[Idx])
} else if (Res.get() != Args[Idx]) {
Args[Idx] = Res.get();
}
}
if (VDecl->isInvalidDecl())
return;
InitializationSequence InitSeq(*this, Entity, Kind, Args);
ExprResult Result = InitSeq.Perform(*this, Entity, Kind, Args, &DclT);
@ -13519,49 +13520,6 @@ static void CheckForDuplicateEnumValues(Sema &S, ArrayRef<Decl *> Elements,
}
}
bool
Sema::IsValueInFlagEnum(const EnumDecl *ED, const llvm::APInt &Val,
bool AllowMask) const {
FlagEnumAttr *FEAttr = ED->getAttr<FlagEnumAttr>();
assert(FEAttr && "looking for value in non-flag enum");
llvm::APInt FlagMask = ~FEAttr->getFlagBits();
unsigned Width = FlagMask.getBitWidth();
// We will try a zero-extended value for the regular check first.
llvm::APInt ExtVal = Val.zextOrSelf(Width);
// A value is in a flag enum if either its bits are a subset of the enum's
// flag bits (the first condition) or we are allowing masks and the same is
// true of its complement (the second condition). When masks are allowed, we
// allow the common idiom of ~(enum1 | enum2) to be a valid enum value.
//
// While it's true that any value could be used as a mask, the assumption is
// that a mask will have all of the insignificant bits set. Anything else is
// likely a logic error.
if (!(FlagMask & ExtVal))
return true;
if (AllowMask) {
// Try a one-extended value instead. This can happen if the enum is wider
// than the constant used, in C with extensions to allow for wider enums.
// The mask will still have the correct behaviour, so we give the user the
// benefit of the doubt.
//
// FIXME: This heuristic can cause weird results if the enum was extended
// to a larger type and is signed, because then bit-masks of smaller types
// that get extended will fall out of range (e.g. ~0x1u). We currently don't
// detect that case and will get a false positive for it. In most cases,
// though, it can be fixed by making it a signed type (e.g. ~0x1), so it may
// be fine just to accept this as a warning.
ExtVal |= llvm::APInt::getHighBitsSet(Width, Width - Val.getBitWidth());
if (!(FlagMask & ~ExtVal))
return true;
}
return false;
}
void Sema::ActOnEnumBody(SourceLocation EnumLoc, SourceLocation LBraceLoc,
SourceLocation RBraceLoc, Decl *EnumDeclX,
ArrayRef<Decl *> Elements,
@ -13647,8 +13605,10 @@ void Sema::ActOnEnumBody(SourceLocation EnumLoc, SourceLocation LBraceLoc,
BestPromotionType = Context.getPromotedIntegerType(BestType);
else
BestPromotionType = BestType;
BestWidth = Context.getIntWidth(BestType);
// We don't need to set BestWidth, because BestType is going to be the type
// of the enumerators, but we do anyway because otherwise some compilers
// warn that it might be used uninitialized.
BestWidth = CharWidth;
}
else if (NumNegativeBits) {
// If there is a negative value, figure out the smallest integer type (of
@ -13713,15 +13673,10 @@ void Sema::ActOnEnumBody(SourceLocation EnumLoc, SourceLocation LBraceLoc,
}
}
FlagEnumAttr *FEAttr = Enum->getAttr<FlagEnumAttr>();
if (FEAttr)
FEAttr->getFlagBits() = llvm::APInt(BestWidth, 0);
// Loop over all of the enumerator constants, changing their types to match
// the type of the enum if needed. If we have a flag type, we also prepare the
// FlagBits cache.
for (auto *D : Elements) {
auto *ECD = cast_or_null<EnumConstantDecl>(D);
// the type of the enum if needed.
for (unsigned i = 0, e = Elements.size(); i != e; ++i) {
EnumConstantDecl *ECD = cast_or_null<EnumConstantDecl>(Elements[i]);
if (!ECD) continue; // Already issued a diagnostic.
// Standard C says the enumerators have int type, but we allow, as an
@ -13751,7 +13706,7 @@ void Sema::ActOnEnumBody(SourceLocation EnumLoc, SourceLocation LBraceLoc,
// enum-specifier, each enumerator has the type of its
// enumeration.
ECD->setType(EnumType);
goto flagbits;
continue;
} else {
NewTy = BestType;
NewWidth = BestWidth;
@ -13778,32 +13733,8 @@ void Sema::ActOnEnumBody(SourceLocation EnumLoc, SourceLocation LBraceLoc,
ECD->setType(EnumType);
else
ECD->setType(NewTy);
flagbits:
// Check to see if we have a constant with exactly one bit set. Note that x
// & (x - 1) will be nonzero if and only if x has more than one bit set.
if (FEAttr) {
llvm::APInt ExtVal = InitVal.zextOrSelf(BestWidth);
if (ExtVal != 0 && !(ExtVal & (ExtVal - 1))) {
FEAttr->getFlagBits() |= ExtVal;
}
}
}
if (FEAttr) {
for (Decl *D : Elements) {
EnumConstantDecl *ECD = cast_or_null<EnumConstantDecl>(D);
if (!ECD) continue; // Already issued a diagnostic.
llvm::APSInt InitVal = ECD->getInitVal();
if (InitVal != 0 && !IsValueInFlagEnum(Enum, InitVal, true))
Diag(ECD->getLocation(), diag::warn_flag_enum_constant_out_of_range)
<< ECD << Enum;
}
}
Enum->completeDefinition(BestType, BestPromotionType,
NumPositiveBits, NumNegativeBits);

3
contrib/llvm/tools/clang/lib/Sema/SemaDeclAttr.cpp
View File

@ -4396,9 +4396,6 @@ static void ProcessDeclAttribute(Sema &S, Scope *scope, Decl *D,
case AttributeList::AT_OptimizeNone:
handleOptimizeNoneAttr(S, D, Attr);
break;
case AttributeList::AT_FlagEnum:
handleSimpleAttribute<FlagEnumAttr>(S, D, Attr);
break;
case AttributeList::AT_Flatten:
handleSimpleAttribute<FlattenAttr>(S, D, Attr);
break;

46
contrib/llvm/tools/clang/lib/Sema/SemaExpr.cpp
View File

@ -4762,12 +4762,8 @@ Sema::BuildResolvedCallExpr(Expr *Fn, NamedDecl *NDecl,
VK_RValue, RParenLoc);
// Bail out early if calling a builtin with custom typechecking.
if (BuiltinID && Context.BuiltinInfo.hasCustomTypechecking(BuiltinID)) {
ExprResult Res = CorrectDelayedTyposInExpr(TheCall);
if (!Res.isUsable() || !isa<CallExpr>(Res.get()))
return Res;
return CheckBuiltinFunctionCall(FDecl, BuiltinID, cast<CallExpr>(Res.get()));
}
if (BuiltinID && Context.BuiltinInfo.hasCustomTypechecking(BuiltinID))
return CheckBuiltinFunctionCall(FDecl, BuiltinID, TheCall);
retry:
const FunctionType *FuncT;
@ -5785,15 +5781,6 @@ QualType Sema::CheckConditionalOperands(ExprResult &Cond, ExprResult &LHS,
ExprObjectKind &OK,
SourceLocation QuestionLoc) {
if (!getLangOpts().CPlusPlus) {
// C cannot handle TypoExpr nodes on either side of a binop because it
// doesn't handle dependent types properly, so make sure any TypoExprs have
// been dealt with before checking the operands.
ExprResult CondResult = CorrectDelayedTyposInExpr(Cond);
if (!CondResult.isUsable()) return QualType();
Cond = CondResult;
}
ExprResult LHSResult = CheckPlaceholderExpr(LHS.get());
if (!LHSResult.isUsable()) return QualType();
LHS = LHSResult;
@ -6173,6 +6160,15 @@ ExprResult Sema::ActOnConditionalOp(SourceLocation QuestionLoc,
SourceLocation ColonLoc,
Expr *CondExpr, Expr *LHSExpr,
Expr *RHSExpr) {
if (!getLangOpts().CPlusPlus) {
// C cannot handle TypoExpr nodes in the condition because it
// doesn't handle dependent types properly, so make sure any TypoExprs have
// been dealt with before checking the operands.
ExprResult CondResult = CorrectDelayedTyposInExpr(CondExpr);
if (!CondResult.isUsable()) return ExprError();
CondExpr = CondResult.get();
}
// If this is the gnu "x ?: y" extension, analyze the types as though the LHS
// was the condition.
OpaqueValueExpr *opaqueValue = nullptr;
@ -9457,6 +9453,18 @@ static void checkObjCPointerIntrospection(Sema &S, ExprResult &L, ExprResult &R,
}
}
static NamedDecl *getDeclFromExpr(Expr *E) {
if (!E)
return nullptr;
if (auto *DRE = dyn_cast<DeclRefExpr>(E))
return DRE->getDecl();
if (auto *ME = dyn_cast<MemberExpr>(E))
return ME->getMemberDecl();
if (auto *IRE = dyn_cast<ObjCIvarRefExpr>(E))
return IRE->getDecl();
return nullptr;
}
/// CreateBuiltinBinOp - Creates a new built-in binary operation with
/// operator @p Opc at location @c TokLoc. This routine only supports
/// built-in operations; ActOnBinOp handles overloaded operators.
@ -9494,7 +9502,13 @@ ExprResult Sema::CreateBuiltinBinOp(SourceLocation OpLoc,
// doesn't handle dependent types properly, so make sure any TypoExprs have
// been dealt with before checking the operands.
LHS = CorrectDelayedTyposInExpr(LHSExpr);
RHS = CorrectDelayedTyposInExpr(RHSExpr);
RHS = CorrectDelayedTyposInExpr(RHSExpr, [Opc, LHS](Expr *E) {
if (Opc != BO_Assign)
return ExprResult(E);
// Avoid correcting the RHS to the same Expr as the LHS.
Decl *D = getDeclFromExpr(E);
return (D && D == getDeclFromExpr(LHS.get())) ? ExprError() : E;
});
if (!LHS.isUsable() || !RHS.isUsable())
return ExprError();
}

21
contrib/llvm/tools/clang/lib/Sema/SemaExprCXX.cpp
View File

@ -6143,12 +6143,6 @@ public:
ExprResult TransformLambdaExpr(LambdaExpr *E) { return Owned(E); }
ExprResult TransformOpaqueValueExpr(OpaqueValueExpr *E) {
if (Expr *SE = E->getSourceExpr())
return TransformExpr(SE);
return BaseTransform::TransformOpaqueValueExpr(E);
}
ExprResult Transform(Expr *E) {
ExprResult Res;
while (true) {
@ -6168,15 +6162,18 @@ public:
while (!AmbiguousTypoExprs.empty()) {
auto TE = AmbiguousTypoExprs.back();
auto Cached = TransformCache[TE];
AmbiguousTypoExprs.pop_back();
auto &State = SemaRef.getTypoExprState(TE);
State.Consumer->saveCurrentPosition();
TransformCache.erase(TE);
if (!TryTransform(E).isInvalid()) {
SemaRef.getTypoExprState(TE).Consumer->resetCorrectionStream();
State.Consumer->resetCorrectionStream();
TransformCache.erase(TE);
Res = ExprError();
break;
} else
TransformCache[TE] = Cached;
}
AmbiguousTypoExprs.remove(TE);
State.Consumer->restoreSavedPosition();
TransformCache[TE] = Cached;
}
// Ensure that all of the TypoExprs within the current Expr have been found.
@ -6235,8 +6232,12 @@ ExprResult Sema::CorrectDelayedTyposInExpr(
if (E && !ExprEvalContexts.empty() && ExprEvalContexts.back().NumTypos &&
(E->isTypeDependent() || E->isValueDependent() ||
E->isInstantiationDependent())) {
auto TyposInContext = ExprEvalContexts.back().NumTypos;
assert(TyposInContext < ~0U && "Recursive call of CorrectDelayedTyposInExpr");
ExprEvalContexts.back().NumTypos = ~0U;
auto TyposResolved = DelayedTypos.size();
auto Result = TransformTypos(*this, Filter).Transform(E);
ExprEvalContexts.back().NumTypos = TyposInContext;
TyposResolved -= DelayedTypos.size();
if (Result.isInvalid() || Result.get() != E) {
ExprEvalContexts.back().NumTypos -= TyposResolved;

2
contrib/llvm/tools/clang/lib/Sema/SemaLookup.cpp
View File

@ -3587,7 +3587,7 @@ retry_lookup:
QualifiedResults.push_back(Candidate);
break;
}
Candidate.setCorrectionRange(TempSS, Result.getLookupNameInfo());
Candidate.setCorrectionRange(SS.get(), Result.getLookupNameInfo());
return true;
}
return false;

145
contrib/llvm/tools/clang/lib/Sema/SemaStmt.cpp
View File

@ -687,39 +687,26 @@ static void checkCaseValue(Sema &S, SourceLocation Loc, const llvm::APSInt &Val,
}
}
typedef SmallVector<std::pair<llvm::APSInt, EnumConstantDecl*>, 64> EnumValsTy;
/// Returns true if we should emit a diagnostic about this case expression not
/// being a part of the enum used in the switch controlling expression.
static bool ShouldDiagnoseSwitchCaseNotInEnum(const Sema &S,
static bool ShouldDiagnoseSwitchCaseNotInEnum(const ASTContext &Ctx,
const EnumDecl *ED,
const Expr *CaseExpr,
EnumValsTy::iterator &EI,
EnumValsTy::iterator &EIEnd,
const llvm::APSInt &Val) {
bool FlagType = ED->hasAttr<FlagEnumAttr>();
if (const DeclRefExpr *DRE =
dyn_cast<DeclRefExpr>(CaseExpr->IgnoreParenImpCasts())) {
const Expr *CaseExpr) {
// Don't warn if the 'case' expression refers to a static const variable of
// the enum type.
CaseExpr = CaseExpr->IgnoreParenImpCasts();
if (const DeclRefExpr *DRE = dyn_cast<DeclRefExpr>(CaseExpr)) {
if (const VarDecl *VD = dyn_cast<VarDecl>(DRE->getDecl())) {
if (!VD->hasGlobalStorage())
return true;
QualType VarType = VD->getType();
QualType EnumType = S.Context.getTypeDeclType(ED);
if (VD->hasGlobalStorage() && VarType.isConstQualified() &&
S.Context.hasSameUnqualifiedType(EnumType, VarType))
if (!VarType.isConstQualified())
return true;
QualType EnumType = Ctx.getTypeDeclType(ED);
if (Ctx.hasSameUnqualifiedType(EnumType, VarType))
return false;
}
}
if (FlagType) {
return !S.IsValueInFlagEnum(ED, Val, false);
} else {
while (EI != EIEnd && EI->first < Val)
EI++;
if (EI != EIEnd && EI->first == Val)
return false;
}
return true;
}
@ -1059,6 +1046,8 @@ Sema::ActOnFinishSwitchStmt(SourceLocation SwitchLoc, Stmt *Switch,
// If switch has default case, then ignore it.
if (!CaseListIsErroneous && !HasConstantCond && ET) {
const EnumDecl *ED = ET->getDecl();
typedef SmallVector<std::pair<llvm::APSInt, EnumConstantDecl*>, 64>
EnumValsTy;
EnumValsTy EnumVals;
// Gather all enum values, set their type and sort them,
@ -1069,48 +1058,57 @@ Sema::ActOnFinishSwitchStmt(SourceLocation SwitchLoc, Stmt *Switch,
EnumVals.push_back(std::make_pair(Val, EDI));
}
std::stable_sort(EnumVals.begin(), EnumVals.end(), CmpEnumVals);
auto EI = EnumVals.begin(), EIEnd =
EnumValsTy::iterator EIend =
std::unique(EnumVals.begin(), EnumVals.end(), EqEnumVals);
// See which case values aren't in enum.
EnumValsTy::const_iterator EI = EnumVals.begin();
for (CaseValsTy::const_iterator CI = CaseVals.begin();
CI != CaseVals.end(); CI++) {
Expr *CaseExpr = CI->second->getLHS();
if (ShouldDiagnoseSwitchCaseNotInEnum(*this, ED, CaseExpr, EI, EIEnd,
CI->first))
Diag(CaseExpr->getExprLoc(), diag::warn_not_in_enum)
<< CondTypeBeforePromotion;
CI != CaseVals.end(); CI++) {
while (EI != EIend && EI->first < CI->first)
EI++;
if (EI == EIend || EI->first > CI->first) {
Expr *CaseExpr = CI->second->getLHS();
if (ShouldDiagnoseSwitchCaseNotInEnum(Context, ED, CaseExpr))
Diag(CaseExpr->getExprLoc(), diag::warn_not_in_enum)
<< CondTypeBeforePromotion;
}
}
// See which of case ranges aren't in enum
EI = EnumVals.begin();
for (CaseRangesTy::const_iterator RI = CaseRanges.begin();
RI != CaseRanges.end(); RI++) {
Expr *CaseExpr = RI->second->getLHS();
if (ShouldDiagnoseSwitchCaseNotInEnum(*this, ED, CaseExpr, EI, EIEnd,
RI->first))
Diag(CaseExpr->getExprLoc(), diag::warn_not_in_enum)
<< CondTypeBeforePromotion;
RI != CaseRanges.end() && EI != EIend; RI++) {
while (EI != EIend && EI->first < RI->first)
EI++;
if (EI == EIend || EI->first != RI->first) {
Expr *CaseExpr = RI->second->getLHS();
if (ShouldDiagnoseSwitchCaseNotInEnum(Context, ED, CaseExpr))
Diag(CaseExpr->getExprLoc(), diag::warn_not_in_enum)
<< CondTypeBeforePromotion;
}
llvm::APSInt Hi =
RI->second->getRHS()->EvaluateKnownConstInt(Context);
AdjustAPSInt(Hi, CondWidth, CondIsSigned);
CaseExpr = RI->second->getRHS();
if (ShouldDiagnoseSwitchCaseNotInEnum(*this, ED, CaseExpr, EI, EIEnd,
Hi))
Diag(CaseExpr->getExprLoc(), diag::warn_not_in_enum)
<< CondTypeBeforePromotion;
while (EI != EIend && EI->first < Hi)
EI++;
if (EI == EIend || EI->first != Hi) {
Expr *CaseExpr = RI->second->getRHS();
if (ShouldDiagnoseSwitchCaseNotInEnum(Context, ED, CaseExpr))
Diag(CaseExpr->getExprLoc(), diag::warn_not_in_enum)
<< CondTypeBeforePromotion;
}
}
// Check which enum vals aren't in switch
auto CI = CaseVals.begin();
auto RI = CaseRanges.begin();
CaseValsTy::const_iterator CI = CaseVals.begin();
CaseRangesTy::const_iterator RI = CaseRanges.begin();
bool hasCasesNotInSwitch = false;
SmallVector<DeclarationName,8> UnhandledNames;
for (EI = EnumVals.begin(); EI != EIEnd; EI++){
for (EI = EnumVals.begin(); EI != EIend; EI++){
// Drop unneeded case values
while (CI != CaseVals.end() && CI->first < EI->first)
CI++;
@ -1197,37 +1195,30 @@ Sema::DiagnoseAssignmentEnum(QualType DstType, QualType SrcType,
llvm::APSInt RhsVal = SrcExpr->EvaluateKnownConstInt(Context);
AdjustAPSInt(RhsVal, DstWidth, DstIsSigned);
const EnumDecl *ED = ET->getDecl();
typedef SmallVector<std::pair<llvm::APSInt, EnumConstantDecl *>, 64>
EnumValsTy;
EnumValsTy EnumVals;
if (ED->hasAttr<FlagEnumAttr>()) {
if (!IsValueInFlagEnum(ED, RhsVal, true))
Diag(SrcExpr->getExprLoc(), diag::warn_not_in_enum_assignment)
// Gather all enum values, set their type and sort them,
// allowing easier comparison with rhs constant.
for (auto *EDI : ED->enumerators()) {
llvm::APSInt Val = EDI->getInitVal();
AdjustAPSInt(Val, DstWidth, DstIsSigned);
EnumVals.push_back(std::make_pair(Val, EDI));
}
if (EnumVals.empty())
return;
std::stable_sort(EnumVals.begin(), EnumVals.end(), CmpEnumVals);
EnumValsTy::iterator EIend =
std::unique(EnumVals.begin(), EnumVals.end(), EqEnumVals);
// See which values aren't in the enum.
EnumValsTy::const_iterator EI = EnumVals.begin();
while (EI != EIend && EI->first < RhsVal)
EI++;
if (EI == EIend || EI->first != RhsVal) {
Diag(SrcExpr->getExprLoc(), diag::warn_not_in_enum_assignment)
<< DstType.getUnqualifiedType();
} else {
typedef SmallVector<std::pair<llvm::APSInt, EnumConstantDecl *>, 64>
EnumValsTy;
EnumValsTy EnumVals;
// Gather all enum values, set their type and sort them,
// allowing easier comparison with rhs constant.
for (auto *EDI : ED->enumerators()) {
llvm::APSInt Val = EDI->getInitVal();
AdjustAPSInt(Val, DstWidth, DstIsSigned);
EnumVals.push_back(std::make_pair(Val, EDI));
}
if (EnumVals.empty())
return;
std::stable_sort(EnumVals.begin(), EnumVals.end(), CmpEnumVals);
EnumValsTy::iterator EIend =
std::unique(EnumVals.begin(), EnumVals.end(), EqEnumVals);
// See which values aren't in the enum.
EnumValsTy::const_iterator EI = EnumVals.begin();
while (EI != EIend && EI->first < RhsVal)
EI++;
if (EI == EIend || EI->first != RhsVal) {
Diag(SrcExpr->getExprLoc(), diag::warn_not_in_enum_assignment)
<< DstType.getUnqualifiedType();
}
}
}
}

5
contrib/llvm/tools/clang/lib/Sema/SemaStmtAsm.cpp
View File

@ -230,9 +230,8 @@ StmtResult Sema::ActOnGCCAsmStmt(SourceLocation AsmLoc, bool IsSimple,
llvm::APSInt Result;
if (!InputExpr->EvaluateAsInt(Result, Context))
return StmtError(
Diag(InputExpr->getLocStart(), diag::err_asm_invalid_type_in_input)
<< InputExpr->getType() << Info.getConstraintStr()
<< InputExpr->getSourceRange());
Diag(InputExpr->getLocStart(), diag::err_asm_immediate_expected)
<< Info.getConstraintStr() << InputExpr->getSourceRange());
if (Result.slt(Info.getImmConstantMin()) ||
Result.sgt(Info.getImmConstantMax()))
return StmtError(Diag(InputExpr->getLocStart(),

4
contrib/llvm/tools/clang/lib/Tooling/ArgumentsAdjusters.cpp
View File

@ -78,7 +78,9 @@ ArgumentsAdjuster getInsertArgumentAdjuster(const char *Extra,
ArgumentsAdjuster combineAdjusters(ArgumentsAdjuster First,
ArgumentsAdjuster Second) {
return std::bind(Second, std::bind(First, std::placeholders::_1));
return [First, Second](const CommandLineArguments &Args) {
return Second(First(Args));
};
}
} // end namespace tooling

7
contrib/llvm/tools/clang/utils/TableGen/ClangAttrEmitter.cpp
View File

@ -2185,8 +2185,7 @@ static std::string CalculateDiagnostic(const Record &S) {
Namespace = 1U << 11,
Field = 1U << 12,
CXXMethod = 1U << 13,
ObjCProtocol = 1U << 14,
Enum = 1U << 15
ObjCProtocol = 1U << 14
};
uint32_t SubMask = 0;
@ -2220,7 +2219,6 @@ static std::string CalculateDiagnostic(const Record &S) {
.Case("Namespace", Namespace)
.Case("Field", Field)
.Case("CXXMethod", CXXMethod)
.Case("Enum", Enum)
.Default(0);
if (!V) {
// Something wasn't in our mapping, so be helpful and let the developer
@ -2239,7 +2237,6 @@ static std::string CalculateDiagnostic(const Record &S) {
case Var: return "ExpectedVariable";
case Param: return "ExpectedParameter";
case Class: return "ExpectedClass";
case Enum: return "ExpectedEnum";
case CXXMethod:
// FIXME: Currently, this maps to ExpectedMethod based on existing code,
// but should map to something a bit more accurate at some point.
@ -2393,8 +2390,6 @@ static std::string GenerateLangOptRequirements(const Record &R,
std::string FnName = "check", Test;
for (auto I = LangOpts.begin(), E = LangOpts.end(); I != E; ++I) {
std::string Part = (*I)->getValueAsString("Name");
if ((*I)->getValueAsBit("Negated"))
Test += "!";
Test += "S.LangOpts." + Part;
if (I + 1 != E)
Test += " || ";

Some files were not shown because too many files have changed in this diff Show More