aboutsummaryrefslogtreecommitdiff
path: root/llvm/lib/CodeGen
diff options
context:
space:
mode:
Diffstat (limited to 'llvm/lib/CodeGen')
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/AIXException.cpp4
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp118
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp2
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp53
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp59
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/DwarfExpression.cpp10
-rw-r--r--llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp12
-rw-r--r--llvm/lib/CodeGen/AssignmentTrackingAnalysis.cpp4
-rw-r--r--llvm/lib/CodeGen/AtomicExpandPass.cpp2
-rw-r--r--llvm/lib/CodeGen/BasicBlockPathCloning.cpp4
-rw-r--r--llvm/lib/CodeGen/BranchRelaxation.cpp14
-rw-r--r--llvm/lib/CodeGen/BreakFalseDeps.cpp4
-rw-r--r--llvm/lib/CodeGen/CodeGenPrepare.cpp8
-rw-r--r--llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp4
-rw-r--r--llvm/lib/CodeGen/EdgeBundles.cpp11
-rw-r--r--llvm/lib/CodeGen/ExpandFp.cpp134
-rw-r--r--llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp1
-rw-r--r--llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp93
-rw-r--r--llvm/lib/CodeGen/GlobalISel/LegalizeMutations.cpp6
-rw-r--r--llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp6
-rw-r--r--llvm/lib/CodeGen/GlobalISel/Utils.cpp3
-rw-r--r--llvm/lib/CodeGen/GlobalMergeFunctions.cpp10
-rw-r--r--llvm/lib/CodeGen/LiveIntervals.cpp11
-rw-r--r--llvm/lib/CodeGen/MIR2Vec.cpp13
-rw-r--r--llvm/lib/CodeGen/MIRFSDiscriminator.cpp2
-rw-r--r--llvm/lib/CodeGen/MIRNamerPass.cpp17
-rw-r--r--llvm/lib/CodeGen/MIRPrinter.cpp36
-rw-r--r--llvm/lib/CodeGen/MLRegAllocEvictAdvisor.cpp9
-rw-r--r--llvm/lib/CodeGen/MachineBlockFrequencyInfo.cpp18
-rw-r--r--llvm/lib/CodeGen/MachineBranchProbabilityInfo.cpp9
-rw-r--r--llvm/lib/CodeGen/MachineCopyPropagation.cpp2
-rw-r--r--llvm/lib/CodeGen/MachineFunction.cpp62
-rw-r--r--llvm/lib/CodeGen/MachineFunctionPrinterPass.cpp8
-rw-r--r--llvm/lib/CodeGen/MachineOutliner.cpp5
-rw-r--r--llvm/lib/CodeGen/MachinePipeliner.cpp13
-rw-r--r--llvm/lib/CodeGen/MachineScheduler.cpp104
-rw-r--r--llvm/lib/CodeGen/MachineSink.cpp2
-rw-r--r--llvm/lib/CodeGen/MachineTraceMetrics.cpp7
-rw-r--r--llvm/lib/CodeGen/NonRelocatableStringpool.cpp4
-rw-r--r--llvm/lib/CodeGen/RegisterCoalescer.cpp6
-rw-r--r--llvm/lib/CodeGen/RegisterPressure.cpp2
-rw-r--r--llvm/lib/CodeGen/SafeStack.cpp4
-rw-r--r--llvm/lib/CodeGen/ScheduleDAGInstrs.cpp8
-rw-r--r--llvm/lib/CodeGen/ScheduleDAGPrinter.cpp80
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp204
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp26
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h2
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp70
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp17
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp1
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp32
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp1
-rw-r--r--llvm/lib/CodeGen/StackFrameLayoutAnalysisPass.cpp5
-rw-r--r--llvm/lib/CodeGen/StaticDataAnnotator.cpp15
-rw-r--r--llvm/lib/CodeGen/StaticDataSplitter.cpp6
-rw-r--r--llvm/lib/CodeGen/TargetLoweringBase.cpp3
-rw-r--r--llvm/lib/CodeGen/TargetRegisterInfo.cpp17
57 files changed, 754 insertions, 629 deletions
diff --git a/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp b/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp
index 6356d71..873ac8f 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp
@@ -20,7 +20,7 @@
#include "llvm/Target/TargetLoweringObjectFile.h"
#include "llvm/Target/TargetMachine.h"
-namespace llvm {
+using namespace llvm;
AIXException::AIXException(AsmPrinter *A) : EHStreamer(A) {}
@@ -90,5 +90,3 @@ void AIXException::endFunction(const MachineFunction *MF) {
emitExceptionInfoTable(LSDALabel, PerSym);
}
-
-} // End of namespace llvm
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
index 10df9c1..e2af0c5 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
@@ -20,6 +20,7 @@
#include "WinException.h"
#include "llvm/ADT/APFloat.h"
#include "llvm/ADT/APInt.h"
+#include "llvm/ADT/BitmaskEnum.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
@@ -118,6 +119,7 @@
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/Path.h"
#include "llvm/Support/VCSRevision.h"
+#include "llvm/Support/VirtualFileSystem.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetLoweringObjectFile.h"
#include "llvm/Target/TargetMachine.h"
@@ -205,6 +207,17 @@ public:
};
} // namespace
+namespace callgraph {
+LLVM_ENABLE_BITMASK_ENUMS_IN_NAMESPACE();
+enum Flags : uint8_t {
+ None = 0,
+ IsIndirectTarget = 1u << 0,
+ HasDirectCallees = 1u << 1,
+ HasIndirectCallees = 1u << 2,
+ LLVM_MARK_AS_BITMASK_ENUM(/*LargestValue*/ HasIndirectCallees)
+};
+} // namespace callgraph
+
class llvm::AddrLabelMap {
MCContext &Context;
struct AddrLabelSymEntry {
@@ -464,6 +477,7 @@ void AsmPrinter::getAnalysisUsage(AnalysisUsage &AU) const {
}
bool AsmPrinter::doInitialization(Module &M) {
+ VFS = vfs::getRealFileSystem();
auto *MMIWP = getAnalysisIfAvailable<MachineModuleInfoWrapperPass>();
MMI = MMIWP ? &MMIWP->getMMI() : nullptr;
HasSplitStack = false;
@@ -1425,7 +1439,8 @@ getBBAddrMapFeature(const MachineFunction &MF, int NumMBBSectionRanges,
BrProbEnabled,
MF.hasBBSections() && NumMBBSectionRanges > 1,
static_cast<bool>(BBAddrMapSkipEmitBBEntries),
- HasCalls};
+ HasCalls,
+ false};
}
void AsmPrinter::emitBBAddrMapSection(const MachineFunction &MF) {
@@ -1670,7 +1685,7 @@ static ConstantInt *extractNumericCGTypeId(const Function &F) {
return nullptr;
}
-/// Emits .callgraph section.
+/// Emits .llvm.callgraph section.
void AsmPrinter::emitCallGraphSection(const MachineFunction &MF,
FunctionCallGraphInfo &FuncCGInfo) {
if (!MF.getTarget().Options.EmitCallGraphSection)
@@ -1683,63 +1698,65 @@ void AsmPrinter::emitCallGraphSection(const MachineFunction &MF,
OutStreamer->pushSection();
OutStreamer->switchSection(FuncCGSection);
- // Emit format version number.
- OutStreamer->emitInt64(CallGraphSectionFormatVersion::V_0);
-
- // Emit function's self information, which is composed of:
- // 1) FunctionEntryPc
- // 2) FunctionKind: Whether the function is indirect target, and if so,
- // whether its type id is known.
- // 3) FunctionTypeId: Emit only when the function is an indirect target
- // and its type id is known.
-
- // Emit function entry pc.
const MCSymbol *FunctionSymbol = getFunctionBegin();
- OutStreamer->emitSymbolValue(FunctionSymbol, TM.getProgramPointerSize());
-
+ const Function &F = MF.getFunction();
// If this function has external linkage or has its address taken and
// it is not a callback, then anything could call it.
- const Function &F = MF.getFunction();
bool IsIndirectTarget =
!F.hasLocalLinkage() || F.hasAddressTaken(nullptr,
/*IgnoreCallbackUses=*/true,
/*IgnoreAssumeLikeCalls=*/true,
/*IgnoreLLVMUsed=*/false);
- // FIXME: FunctionKind takes a few values but emitted as a 64-bit value.
- // Can be optimized to occupy 2 bits instead.
- // Emit function kind, and type id if available.
- if (!IsIndirectTarget) {
- OutStreamer->emitInt64(
- static_cast<uint64_t>(FunctionKind::NOT_INDIRECT_TARGET));
- } else {
- if (const auto *TypeId = extractNumericCGTypeId(F)) {
- OutStreamer->emitInt64(
- static_cast<uint64_t>(FunctionKind::INDIRECT_TARGET_KNOWN_TID));
- OutStreamer->emitInt64(TypeId->getZExtValue());
- } else {
- OutStreamer->emitInt64(
- static_cast<uint64_t>(FunctionKind::INDIRECT_TARGET_UNKNOWN_TID));
- }
- }
+ const auto &DirectCallees = FuncCGInfo.DirectCallees;
+ const auto &IndirectCalleeTypeIDs = FuncCGInfo.IndirectCalleeTypeIDs;
+
+ using namespace callgraph;
+ Flags CGFlags = Flags::None;
+ if (IsIndirectTarget)
+ CGFlags |= Flags::IsIndirectTarget;
+ if (DirectCallees.size() > 0)
+ CGFlags |= Flags::HasDirectCallees;
+ if (IndirectCalleeTypeIDs.size() > 0)
+ CGFlags |= Flags::HasIndirectCallees;
+
+ // Emit function's call graph information.
+ // 1) CallGraphSectionFormatVersion
+ // 2) Flags
+ // a. LSB bit 0 is set to 1 if the function is a potential indirect
+ // target.
+ // b. LSB bit 1 is set to 1 if there are direct callees.
+ // c. LSB bit 2 is set to 1 if there are indirect callees.
+ // d. Rest of the 5 bits in Flags are reserved for any future use.
+ // 3) Function entry PC.
+ // 4) FunctionTypeID if the function is indirect target and its type id
+ // is known, otherwise it is set to 0.
+ // 5) Number of unique direct callees, if at least one exists.
+ // 6) For each unique direct callee, the callee's PC.
+ // 7) Number of unique indirect target type IDs, if at least one exists.
+ // 8) Each unique indirect target type id.
+ OutStreamer->emitInt8(CallGraphSectionFormatVersion::V_0);
+ OutStreamer->emitInt8(static_cast<uint8_t>(CGFlags));
+ OutStreamer->emitSymbolValue(FunctionSymbol, TM.getProgramPointerSize());
+ const auto *TypeId = extractNumericCGTypeId(F);
+ if (IsIndirectTarget && TypeId)
+ OutStreamer->emitInt64(TypeId->getZExtValue());
+ else
+ OutStreamer->emitInt64(0);
- // Emit callsite labels, where each element is a pair of type id and
- // indirect callsite pc.
- const auto &CallSiteLabels = FuncCGInfo.CallSiteLabels;
- OutStreamer->emitInt64(CallSiteLabels.size());
- for (const auto &[TypeId, Label] : CallSiteLabels) {
- OutStreamer->emitInt64(TypeId);
- OutStreamer->emitSymbolValue(Label, TM.getProgramPointerSize());
+ if (DirectCallees.size() > 0) {
+ OutStreamer->emitULEB128IntValue(DirectCallees.size());
+ for (const auto &CalleeSymbol : DirectCallees)
+ OutStreamer->emitSymbolValue(CalleeSymbol, TM.getProgramPointerSize());
+ FuncCGInfo.DirectCallees.clear();
}
- FuncCGInfo.CallSiteLabels.clear();
-
- const auto &DirectCallees = FuncCGInfo.DirectCallees;
- OutStreamer->emitInt64(DirectCallees.size());
- for (const auto &CalleeSymbol : DirectCallees) {
- OutStreamer->emitSymbolValue(CalleeSymbol, TM.getProgramPointerSize());
+ if (IndirectCalleeTypeIDs.size() > 0) {
+ OutStreamer->emitULEB128IntValue(IndirectCalleeTypeIDs.size());
+ for (const auto &CalleeTypeId : IndirectCalleeTypeIDs)
+ OutStreamer->emitInt64(CalleeTypeId);
+ FuncCGInfo.IndirectCalleeTypeIDs.clear();
}
- FuncCGInfo.DirectCallees.clear();
-
+ // End of emitting call graph section contents.
OutStreamer->popSection();
}
@@ -1877,8 +1894,7 @@ void AsmPrinter::handleCallsiteForCallgraph(
FunctionCallGraphInfo &FuncCGInfo,
const MachineFunction::CallSiteInfoMap &CallSitesInfoMap,
const MachineInstr &MI) {
- assert(MI.isCall() &&
- "Callsite labels are meant for call instructions only.");
+ assert(MI.isCall() && "This method is meant for call instructions only.");
const MachineOperand &CalleeOperand = MI.getOperand(0);
if (CalleeOperand.isGlobal() || CalleeOperand.isSymbol()) {
// Handle direct calls.
@@ -1903,10 +1919,8 @@ void AsmPrinter::handleCallsiteForCallgraph(
// Handle indirect callsite info.
// Only indirect calls have type identifiers set.
for (ConstantInt *CalleeTypeId : CallSiteInfo->second.CalleeTypeIds) {
- MCSymbol *S = MF->getContext().createTempSymbol();
- OutStreamer->emitLabel(S);
uint64_t CalleeTypeIdVal = CalleeTypeId->getZExtValue();
- FuncCGInfo.CallSiteLabels.emplace_back(CalleeTypeIdVal, S);
+ FuncCGInfo.IndirectCalleeTypeIDs.insert(CalleeTypeIdVal);
}
}
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp
index c364ffc..8dd8b9da 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinterInlineAsm.cpp
@@ -36,6 +36,7 @@
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MemoryBuffer.h"
#include "llvm/Support/SourceMgr.h"
+#include "llvm/Support/VirtualFileSystem.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetMachine.h"
using namespace llvm;
@@ -98,6 +99,7 @@ void AsmPrinter::emitInlineAsm(StringRef Str, const MCSubtargetInfo &STI,
unsigned BufNum = addInlineAsmDiagBuffer(Str, LocMDNode);
SourceMgr &SrcMgr = *MMI->getContext().getInlineSourceManager();
SrcMgr.setIncludeDirs(MCOptions.IASSearchPaths);
+ SrcMgr.setVirtualFileSystem(VFS);
std::unique_ptr<MCAsmParser> Parser(
createMCAsmParser(SrcMgr, OutContext, *OutStreamer, *MAI, BufNum));
diff --git a/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
index 12d749c..e57ed24 100644
--- a/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
@@ -569,40 +569,30 @@ void CodeViewDebug::emitCodeViewMagicVersion() {
OS.emitInt32(COFF::DEBUG_SECTION_MAGIC);
}
-static SourceLanguage MapDWLangToCVLang(unsigned DWLang) {
- switch (DWLang) {
- case dwarf::DW_LANG_C:
- case dwarf::DW_LANG_C89:
- case dwarf::DW_LANG_C99:
- case dwarf::DW_LANG_C11:
+static SourceLanguage
+MapDWARFLanguageToCVLang(dwarf::SourceLanguageName DWLName) {
+ switch (DWLName) {
+ case dwarf::DW_LNAME_C:
return SourceLanguage::C;
- case dwarf::DW_LANG_C_plus_plus:
- case dwarf::DW_LANG_C_plus_plus_03:
- case dwarf::DW_LANG_C_plus_plus_11:
- case dwarf::DW_LANG_C_plus_plus_14:
+ case dwarf::DW_LNAME_C_plus_plus:
return SourceLanguage::Cpp;
- case dwarf::DW_LANG_Fortran77:
- case dwarf::DW_LANG_Fortran90:
- case dwarf::DW_LANG_Fortran95:
- case dwarf::DW_LANG_Fortran03:
- case dwarf::DW_LANG_Fortran08:
+ case dwarf::DW_LNAME_Fortran:
return SourceLanguage::Fortran;
- case dwarf::DW_LANG_Pascal83:
+ case dwarf::DW_LNAME_Pascal:
return SourceLanguage::Pascal;
- case dwarf::DW_LANG_Cobol74:
- case dwarf::DW_LANG_Cobol85:
+ case dwarf::DW_LNAME_Cobol:
return SourceLanguage::Cobol;
- case dwarf::DW_LANG_Java:
+ case dwarf::DW_LNAME_Java:
return SourceLanguage::Java;
- case dwarf::DW_LANG_D:
+ case dwarf::DW_LNAME_D:
return SourceLanguage::D;
- case dwarf::DW_LANG_Swift:
+ case dwarf::DW_LNAME_Swift:
return SourceLanguage::Swift;
- case dwarf::DW_LANG_Rust:
+ case dwarf::DW_LNAME_Rust:
return SourceLanguage::Rust;
- case dwarf::DW_LANG_ObjC:
+ case dwarf::DW_LNAME_ObjC:
return SourceLanguage::ObjC;
- case dwarf::DW_LANG_ObjC_plus_plus:
+ case dwarf::DW_LNAME_ObjC_plus_plus:
return SourceLanguage::ObjCpp;
default:
// There's no CodeView representation for this language, and CV doesn't
@@ -612,6 +602,14 @@ static SourceLanguage MapDWLangToCVLang(unsigned DWLang) {
}
}
+static SourceLanguage MapDWARFLanguageToCVLang(dwarf::SourceLanguage DWLang) {
+ auto MaybeLName = dwarf::toDW_LNAME(DWLang);
+ if (!MaybeLName)
+ return MapDWARFLanguageToCVLang(static_cast<dwarf::SourceLanguageName>(0));
+
+ return MapDWARFLanguageToCVLang(MaybeLName->first);
+}
+
void CodeViewDebug::beginModule(Module *M) {
// If COFF debug section is not available, skip any debug info related stuff.
if (!Asm->getObjFileLowering().getCOFFDebugSymbolsSection()) {
@@ -633,8 +631,13 @@ void CodeViewDebug::beginModule(Module *M) {
Node = *CUs->operands().begin();
}
const auto *CU = cast<DICompileUnit>(Node);
+ DISourceLanguageName Lang = CU->getSourceLanguage();
CurrentSourceLanguage =
- MapDWLangToCVLang(CU->getSourceLanguage().getUnversionedName());
+ Lang.hasVersionedName()
+ ? MapDWARFLanguageToCVLang(
+ static_cast<dwarf::SourceLanguageName>(Lang.getName()))
+ : MapDWARFLanguageToCVLang(
+ static_cast<dwarf::SourceLanguage>(Lang.getName()));
if (!M->getCodeViewFlag() ||
CU->getEmissionKind() == DICompileUnit::NoDebug) {
Asm = nullptr;
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
index d751a7f..567acf7 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -1039,8 +1039,17 @@ void DwarfDebug::finishUnitAttributes(const DICompileUnit *DIUnit,
} else
NewCU.addString(Die, dwarf::DW_AT_producer, Producer);
- NewCU.addUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data2,
- DIUnit->getSourceLanguage().getUnversionedName());
+ if (auto Lang = DIUnit->getSourceLanguage(); Lang.hasVersionedName()) {
+ NewCU.addUInt(Die, dwarf::DW_AT_language_name, dwarf::DW_FORM_data2,
+ Lang.getName());
+
+ if (uint32_t LangVersion = Lang.getVersion(); LangVersion != 0)
+ NewCU.addUInt(Die, dwarf::DW_AT_language_version, /*Form=*/std::nullopt,
+ LangVersion);
+ } else {
+ NewCU.addUInt(Die, dwarf::DW_AT_language, dwarf::DW_FORM_data2,
+ Lang.getName());
+ }
NewCU.addString(Die, dwarf::DW_AT_name, FN);
StringRef SysRoot = DIUnit->getSysRoot();
@@ -2062,11 +2071,36 @@ void DwarfDebug::beginInstruction(const MachineInstr *MI) {
if (NoDebug)
return;
+ auto RecordLineZero = [&]() {
+ // Preserve the file and column numbers, if we can, to save space in
+ // the encoded line table.
+ // Do not update PrevInstLoc, it remembers the last non-0 line.
+ const MDNode *Scope = nullptr;
+ unsigned Column = 0;
+ if (PrevInstLoc) {
+ Scope = PrevInstLoc.getScope();
+ Column = PrevInstLoc.getCol();
+ }
+ recordSourceLine(/*Line=*/0, Column, Scope, /*Flags=*/0);
+ };
+
+ // When we emit a line-0 record, we don't update PrevInstLoc; so look at
+ // the last line number actually emitted, to see if it was line 0.
+ unsigned LastAsmLine =
+ Asm->OutStreamer->getContext().getCurrentDwarfLoc().getLine();
+
// Check if source location changes, but ignore DBG_VALUE and CFI locations.
// If the instruction is part of the function frame setup code, do not emit
// any line record, as there is no correspondence with any user code.
- if (MI->isMetaInstruction() || MI->getFlag(MachineInstr::FrameSetup))
+ if (MI->isMetaInstruction())
return;
+ if (MI->getFlag(MachineInstr::FrameSetup)) {
+ // Prevent a loc from the previous block leaking into frame setup instrs.
+ if (LastAsmLine && PrevInstBB && PrevInstBB != MI->getParent())
+ RecordLineZero();
+ return;
+ }
+
const DebugLoc &DL = MI->getDebugLoc();
unsigned Flags = 0;
@@ -2089,11 +2123,6 @@ void DwarfDebug::beginInstruction(const MachineInstr *MI) {
LocationString);
};
- // When we emit a line-0 record, we don't update PrevInstLoc; so look at
- // the last line number actually emitted, to see if it was line 0.
- unsigned LastAsmLine =
- Asm->OutStreamer->getContext().getCurrentDwarfLoc().getLine();
-
// There may be a mixture of scopes using and not using Key Instructions.
// Not-Key-Instructions functions inlined into Key Instructions functions
// should use not-key is_stmt handling. Key Instructions functions inlined
@@ -2159,18 +2188,8 @@ void DwarfDebug::beginInstruction(const MachineInstr *MI) {
// - Instruction is at the top of a block; we don't want to inherit the
// location from the physically previous (maybe unrelated) block.
if (UnknownLocations == Enable || PrevLabel ||
- (PrevInstBB && PrevInstBB != MI->getParent())) {
- // Preserve the file and column numbers, if we can, to save space in
- // the encoded line table.
- // Do not update PrevInstLoc, it remembers the last non-0 line.
- const MDNode *Scope = nullptr;
- unsigned Column = 0;
- if (PrevInstLoc) {
- Scope = PrevInstLoc.getScope();
- Column = PrevInstLoc.getCol();
- }
- recordSourceLine(/*Line=*/0, Column, Scope, /*Flags=*/0);
- }
+ (PrevInstBB && PrevInstBB != MI->getParent()))
+ RecordLineZero();
return;
}
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfExpression.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfExpression.cpp
index bc0bb34..f0f0861 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfExpression.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfExpression.cpp
@@ -587,10 +587,12 @@ bool DwarfExpression::addExpression(
emitUnsigned(LeftShift);
emitOp(dwarf::DW_OP_shl);
}
- emitOp(dwarf::DW_OP_constu);
- emitUnsigned(RightShift);
- emitOp(OpNum == dwarf::DW_OP_LLVM_extract_bits_sext ? dwarf::DW_OP_shra
- : dwarf::DW_OP_shr);
+ if (RightShift) {
+ emitOp(dwarf::DW_OP_constu);
+ emitUnsigned(RightShift);
+ emitOp(OpNum == dwarf::DW_OP_LLVM_extract_bits_sext ? dwarf::DW_OP_shra
+ : dwarf::DW_OP_shr);
+ }
// The value is now at the top of the stack, so set the location to
// implicit so that we get a stack_value at the end.
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
index aa078f3..e40fb76 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfUnit.cpp
@@ -704,9 +704,17 @@ void DwarfUnit::addType(DIE &Entity, const DIType *Ty,
addDIEEntry(Entity, Attribute, DIEEntry(*getOrCreateTypeDIE(Ty)));
}
+// FIXME: change callsites to use the new DW_LNAME_ language codes.
llvm::dwarf::SourceLanguage DwarfUnit::getSourceLanguage() const {
- return static_cast<llvm::dwarf::SourceLanguage>(
- getLanguage().getUnversionedName());
+ const auto &Lang = getLanguage();
+
+ if (!Lang.hasVersionedName())
+ return static_cast<llvm::dwarf::SourceLanguage>(Lang.getName());
+
+ return llvm::dwarf::toDW_LANG(
+ static_cast<llvm::dwarf::SourceLanguageName>(Lang.getName()),
+ Lang.getVersion())
+ .value_or(llvm::dwarf::DW_LANG_hi_user);
}
std::string DwarfUnit::getParentContextString(const DIScope *Context) const {
diff --git a/llvm/lib/CodeGen/AssignmentTrackingAnalysis.cpp b/llvm/lib/CodeGen/AssignmentTrackingAnalysis.cpp
index 260ce8f..93ae548 100644
--- a/llvm/lib/CodeGen/AssignmentTrackingAnalysis.cpp
+++ b/llvm/lib/CodeGen/AssignmentTrackingAnalysis.cpp
@@ -85,8 +85,7 @@ template <> struct llvm::DenseMapInfo<VariableID> {
using VarLocInsertPt = PointerUnion<const Instruction *, const DbgRecord *>;
-namespace std {
-template <> struct hash<VarLocInsertPt> {
+template <> struct std::hash<VarLocInsertPt> {
using argument_type = VarLocInsertPt;
using result_type = std::size_t;
@@ -94,7 +93,6 @@ template <> struct hash<VarLocInsertPt> {
return std::hash<void *>()(Arg.getOpaqueValue());
}
};
-} // namespace std
/// Helper class to build FunctionVarLocs, since that class isn't easy to
/// modify. TODO: There's not a great deal of value in the split, it could be
diff --git a/llvm/lib/CodeGen/AtomicExpandPass.cpp b/llvm/lib/CodeGen/AtomicExpandPass.cpp
index 4931403..53f1cfe2 100644
--- a/llvm/lib/CodeGen/AtomicExpandPass.cpp
+++ b/llvm/lib/CodeGen/AtomicExpandPass.cpp
@@ -770,7 +770,7 @@ struct PartwordMaskValues {
Value *Inv_Mask = nullptr;
};
-LLVM_ATTRIBUTE_UNUSED
+[[maybe_unused]]
raw_ostream &operator<<(raw_ostream &O, const PartwordMaskValues &PMV) {
auto PrintObj = [&O](auto *V) {
if (V)
diff --git a/llvm/lib/CodeGen/BasicBlockPathCloning.cpp b/llvm/lib/CodeGen/BasicBlockPathCloning.cpp
index fd7df6b..47b7a88 100644
--- a/llvm/lib/CodeGen/BasicBlockPathCloning.cpp
+++ b/llvm/lib/CodeGen/BasicBlockPathCloning.cpp
@@ -207,9 +207,7 @@ bool ApplyCloning(MachineFunction &MF,
}
return AnyPathsCloned;
}
-} // end anonymous namespace
-namespace llvm {
class BasicBlockPathCloning : public MachineFunctionPass {
public:
static char ID;
@@ -229,7 +227,7 @@ public:
bool runOnMachineFunction(MachineFunction &MF) override;
};
-} // namespace llvm
+} // namespace
char BasicBlockPathCloning::ID = 0;
INITIALIZE_PASS_BEGIN(
diff --git a/llvm/lib/CodeGen/BranchRelaxation.cpp b/llvm/lib/CodeGen/BranchRelaxation.cpp
index 2d50167..fae952e 100644
--- a/llvm/lib/CodeGen/BranchRelaxation.cpp
+++ b/llvm/lib/CodeGen/BranchRelaxation.cpp
@@ -491,6 +491,20 @@ bool BranchRelaxation::fixupConditionalBranch(MachineInstr &MI) {
return true;
}
if (FBB) {
+ // If we get here with a MBB which ends like this:
+ //
+ // bb.1:
+ // successors: %bb.2;
+ // ...
+ // BNE $x1, $x0, %bb.2
+ // PseudoBR %bb.2
+ //
+ // Just remove conditional branch.
+ if (TBB == FBB) {
+ removeBranch(MBB);
+ insertUncondBranch(MBB, TBB);
+ return true;
+ }
// We need to split the basic block here to obtain two long-range
// unconditional branches.
NewBB = createNewBlockAfter(*MBB);
diff --git a/llvm/lib/CodeGen/BreakFalseDeps.cpp b/llvm/lib/CodeGen/BreakFalseDeps.cpp
index 28e6728..1846880 100644
--- a/llvm/lib/CodeGen/BreakFalseDeps.cpp
+++ b/llvm/lib/CodeGen/BreakFalseDeps.cpp
@@ -31,7 +31,7 @@
using namespace llvm;
-namespace llvm {
+namespace {
class BreakFalseDeps : public MachineFunctionPass {
private:
@@ -95,7 +95,7 @@ private:
void processUndefReads(MachineBasicBlock *);
};
-} // namespace llvm
+} // namespace
#define DEBUG_TYPE "break-false-deps"
diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index eb73d01b..9e78ec9 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -819,7 +819,7 @@ void CodeGenPrepare::removeAllAssertingVHReferences(Value *V) {
}
// Verify BFI has been updated correctly by recomputing BFI and comparing them.
-void LLVM_ATTRIBUTE_UNUSED CodeGenPrepare::verifyBFIUpdates(Function &F) {
+[[maybe_unused]] void CodeGenPrepare::verifyBFIUpdates(Function &F) {
DominatorTree NewDT(F);
LoopInfo NewLI(NewDT);
BranchProbabilityInfo NewBPI(F, NewLI, TLInfo);
@@ -3194,7 +3194,7 @@ struct ExtAddrMode : public TargetLowering::AddrMode {
case ScaledRegField:
return ScaledReg;
case BaseOffsField:
- return ConstantInt::get(IntPtrTy, BaseOffs);
+ return ConstantInt::getSigned(IntPtrTy, BaseOffs);
}
}
@@ -6100,7 +6100,7 @@ bool CodeGenPrepare::optimizeMemoryInst(Instruction *MemoryInst, Value *Addr,
// Add in the Base Offset if present.
if (AddrMode.BaseOffs) {
- Value *V = ConstantInt::get(IntPtrTy, AddrMode.BaseOffs);
+ Value *V = ConstantInt::getSigned(IntPtrTy, AddrMode.BaseOffs);
if (ResultIndex) {
// We need to add this separately from the scale above to help with
// SDAG consecutive load/store merging.
@@ -6226,7 +6226,7 @@ bool CodeGenPrepare::optimizeMemoryInst(Instruction *MemoryInst, Value *Addr,
// Add in the Base Offset if present.
if (AddrMode.BaseOffs) {
- Value *V = ConstantInt::get(IntPtrTy, AddrMode.BaseOffs);
+ Value *V = ConstantInt::getSigned(IntPtrTy, AddrMode.BaseOffs);
if (Result)
Result = Builder.CreateAdd(Result, V, "sunkaddr");
else
diff --git a/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp b/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp
index 6c2a5a7..87ada87 100644
--- a/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp
+++ b/llvm/lib/CodeGen/ComplexDeinterleavingPass.cpp
@@ -126,8 +126,7 @@ hash_code hash_value(const ComplexValue &Arg) {
} // end namespace
typedef SmallVector<struct ComplexValue, 2> ComplexValues;
-namespace llvm {
-template <> struct DenseMapInfo<ComplexValue> {
+template <> struct llvm::DenseMapInfo<ComplexValue> {
static inline ComplexValue getEmptyKey() {
return {DenseMapInfo<Value *>::getEmptyKey(),
DenseMapInfo<Value *>::getEmptyKey()};
@@ -144,7 +143,6 @@ template <> struct DenseMapInfo<ComplexValue> {
return LHS.Real == RHS.Real && LHS.Imag == RHS.Imag;
}
};
-} // end namespace llvm
namespace {
template <typename T, typename IterT>
diff --git a/llvm/lib/CodeGen/EdgeBundles.cpp b/llvm/lib/CodeGen/EdgeBundles.cpp
index f4335396..50dd66f 100644
--- a/llvm/lib/CodeGen/EdgeBundles.cpp
+++ b/llvm/lib/CodeGen/EdgeBundles.cpp
@@ -81,13 +81,10 @@ void EdgeBundles::init() {
}
}
-namespace llvm {
-
/// Specialize WriteGraph, the standard implementation won't work.
-template<>
-raw_ostream &WriteGraph<>(raw_ostream &O, const EdgeBundles &G,
- bool ShortNames,
- const Twine &Title) {
+template <>
+raw_ostream &llvm::WriteGraph<>(raw_ostream &O, const EdgeBundles &G,
+ bool ShortNames, const Twine &Title) {
const MachineFunction *MF = G.getMachineFunction();
O << "digraph {\n";
@@ -107,8 +104,6 @@ raw_ostream &WriteGraph<>(raw_ostream &O, const EdgeBundles &G,
return O;
}
-} // end namespace llvm
-
/// view - Visualize the annotated bipartite CFG with Graphviz.
void EdgeBundles::view() const {
ViewGraph(*this, "EdgeBundles");
diff --git a/llvm/lib/CodeGen/ExpandFp.cpp b/llvm/lib/CodeGen/ExpandFp.cpp
index 9cc6c6a..04c7008 100644
--- a/llvm/lib/CodeGen/ExpandFp.cpp
+++ b/llvm/lib/CodeGen/ExpandFp.cpp
@@ -82,7 +82,7 @@ public:
}
static FRemExpander create(IRBuilder<> &B, Type *Ty) {
- assert(canExpandType(Ty));
+ assert(canExpandType(Ty) && "Expected supported floating point type");
// The type to use for the computation of the remainder. This may be
// wider than the input/result type which affects the ...
@@ -356,8 +356,9 @@ Value *FRemExpander::buildFRem(Value *X, Value *Y,
static bool expandFRem(BinaryOperator &I, std::optional<SimplifyQuery> &SQ) {
LLVM_DEBUG(dbgs() << "Expanding instruction: " << I << '\n');
- Type *ReturnTy = I.getType();
- assert(FRemExpander::canExpandType(ReturnTy->getScalarType()));
+ Type *Ty = I.getType();
+ assert(FRemExpander::canExpandType(Ty) &&
+ "Expected supported floating point type");
FastMathFlags FMF = I.getFastMathFlags();
// TODO Make use of those flags for optimization?
@@ -368,32 +369,10 @@ static bool expandFRem(BinaryOperator &I, std::optional<SimplifyQuery> &SQ) {
B.setFastMathFlags(FMF);
B.SetCurrentDebugLocation(I.getDebugLoc());
- Type *ElemTy = ReturnTy->getScalarType();
- const FRemExpander Expander = FRemExpander::create(B, ElemTy);
-
- Value *Ret;
- if (ReturnTy->isFloatingPointTy())
- Ret = FMF.approxFunc()
- ? Expander.buildApproxFRem(I.getOperand(0), I.getOperand(1))
- : Expander.buildFRem(I.getOperand(0), I.getOperand(1), SQ);
- else {
- auto *VecTy = cast<FixedVectorType>(ReturnTy);
-
- // This could use SplitBlockAndInsertForEachLane but the interface
- // is a bit awkward for a constant number of elements and it will
- // boil down to the same code.
- // TODO Expand the FRem instruction only once and reuse the code.
- Value *Nums = I.getOperand(0);
- Value *Denums = I.getOperand(1);
- Ret = PoisonValue::get(I.getType());
- for (int I = 0, E = VecTy->getNumElements(); I != E; ++I) {
- Value *Num = B.CreateExtractElement(Nums, I);
- Value *Denum = B.CreateExtractElement(Denums, I);
- Value *Rem = FMF.approxFunc() ? Expander.buildApproxFRem(Num, Denum)
- : Expander.buildFRem(Num, Denum, SQ);
- Ret = B.CreateInsertElement(Ret, Rem, I);
- }
- }
+ const FRemExpander Expander = FRemExpander::create(B, Ty);
+ Value *Ret = FMF.approxFunc()
+ ? Expander.buildApproxFRem(I.getOperand(0), I.getOperand(1))
+ : Expander.buildFRem(I.getOperand(0), I.getOperand(1), SQ);
I.replaceAllUsesWith(Ret);
Ret->takeName(&I);
@@ -939,7 +918,8 @@ static void expandIToFP(Instruction *IToFP) {
IToFP->eraseFromParent();
}
-static void scalarize(Instruction *I, SmallVectorImpl<Instruction *> &Replace) {
+static void scalarize(Instruction *I,
+ SmallVectorImpl<Instruction *> &Worklist) {
VectorType *VTy = cast<FixedVectorType>(I->getType());
IRBuilder<> Builder(I);
@@ -948,12 +928,25 @@ static void scalarize(Instruction *I, SmallVectorImpl<Instruction *> &Replace) {
Value *Result = PoisonValue::get(VTy);
for (unsigned Idx = 0; Idx < NumElements; ++Idx) {
Value *Ext = Builder.CreateExtractElement(I->getOperand(0), Idx);
- Value *Cast = Builder.CreateCast(cast<CastInst>(I)->getOpcode(), Ext,
- I->getType()->getScalarType());
- Result = Builder.CreateInsertElement(Result, Cast, Idx);
- if (isa<Instruction>(Cast))
- Replace.push_back(cast<Instruction>(Cast));
+
+ Value *NewOp = nullptr;
+ if (auto *BinOp = dyn_cast<BinaryOperator>(I))
+ NewOp = Builder.CreateBinOp(
+ BinOp->getOpcode(), Ext,
+ Builder.CreateExtractElement(I->getOperand(1), Idx));
+ else if (auto *CastI = dyn_cast<CastInst>(I))
+ NewOp = Builder.CreateCast(CastI->getOpcode(), Ext,
+ I->getType()->getScalarType());
+ else
+ llvm_unreachable("Unsupported instruction type");
+
+ Result = Builder.CreateInsertElement(Result, NewOp, Idx);
+ if (auto *ScalarizedI = dyn_cast<Instruction>(NewOp)) {
+ ScalarizedI->copyIRFlags(I, true);
+ Worklist.push_back(ScalarizedI);
+ }
}
+
I->replaceAllUsesWith(Result);
I->dropAllReferences();
I->eraseFromParent();
@@ -989,10 +982,17 @@ static bool targetSupportsFrem(const TargetLowering &TLI, Type *Ty) {
return TLI.getLibcallName(fremToLibcall(Ty->getScalarType()));
}
+static void addToWorklist(Instruction &I,
+ SmallVector<Instruction *, 4> &Worklist) {
+ if (I.getOperand(0)->getType()->isVectorTy())
+ scalarize(&I, Worklist);
+ else
+ Worklist.push_back(&I);
+}
+
static bool runImpl(Function &F, const TargetLowering &TLI,
AssumptionCache *AC) {
- SmallVector<Instruction *, 4> Replace;
- SmallVector<Instruction *, 4> ReplaceVector;
+ SmallVector<Instruction *, 4> Worklist;
bool Modified = false;
unsigned MaxLegalFpConvertBitWidth =
@@ -1003,55 +1003,39 @@ static bool runImpl(Function &F, const TargetLowering &TLI,
if (MaxLegalFpConvertBitWidth >= llvm::IntegerType::MAX_INT_BITS)
return false;
- for (auto &I : instructions(F)) {
- switch (I.getOpcode()) {
- case Instruction::FRem: {
- Type *Ty = I.getType();
- // TODO: This pass doesn't handle scalable vectors.
- if (Ty->isScalableTy())
- continue;
-
- if (targetSupportsFrem(TLI, Ty) ||
- !FRemExpander::canExpandType(Ty->getScalarType()))
- continue;
-
- Replace.push_back(&I);
- Modified = true;
+ for (auto It = inst_begin(&F), End = inst_end(F); It != End;) {
+ Instruction &I = *It++;
+ Type *Ty = I.getType();
+ // TODO: This pass doesn't handle scalable vectors.
+ if (Ty->isScalableTy())
+ continue;
+ switch (I.getOpcode()) {
+ case Instruction::FRem:
+ if (!targetSupportsFrem(TLI, Ty) &&
+ FRemExpander::canExpandType(Ty->getScalarType())) {
+ addToWorklist(I, Worklist);
+ Modified = true;
+ }
break;
- }
case Instruction::FPToUI:
case Instruction::FPToSI: {
- // TODO: This pass doesn't handle scalable vectors.
- if (I.getOperand(0)->getType()->isScalableTy())
- continue;
-
- auto *IntTy = cast<IntegerType>(I.getType()->getScalarType());
+ auto *IntTy = cast<IntegerType>(Ty->getScalarType());
if (IntTy->getIntegerBitWidth() <= MaxLegalFpConvertBitWidth)
continue;
- if (I.getOperand(0)->getType()->isVectorTy())
- ReplaceVector.push_back(&I);
- else
- Replace.push_back(&I);
+ addToWorklist(I, Worklist);
Modified = true;
break;
}
case Instruction::UIToFP:
case Instruction::SIToFP: {
- // TODO: This pass doesn't handle scalable vectors.
- if (I.getOperand(0)->getType()->isScalableTy())
- continue;
-
auto *IntTy =
cast<IntegerType>(I.getOperand(0)->getType()->getScalarType());
if (IntTy->getIntegerBitWidth() <= MaxLegalFpConvertBitWidth)
continue;
- if (I.getOperand(0)->getType()->isVectorTy())
- ReplaceVector.push_back(&I);
- else
- Replace.push_back(&I);
+ addToWorklist(I, Worklist);
Modified = true;
break;
}
@@ -1060,16 +1044,8 @@ static bool runImpl(Function &F, const TargetLowering &TLI,
}
}
- while (!ReplaceVector.empty()) {
- Instruction *I = ReplaceVector.pop_back_val();
- scalarize(I, Replace);
- }
-
- if (Replace.empty())
- return false;
-
- while (!Replace.empty()) {
- Instruction *I = Replace.pop_back_val();
+ while (!Worklist.empty()) {
+ Instruction *I = Worklist.pop_back_val();
if (I->getOpcode() == Instruction::FRem) {
auto SQ = [&]() -> std::optional<SimplifyQuery> {
if (AC) {
diff --git a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
index 906d62a3..b425b95 100644
--- a/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp
@@ -1728,6 +1728,7 @@ static APFloat constantFoldFpUnary(const MachineInstr &MI,
Result.clearSign();
return Result;
}
+ case TargetOpcode::G_FPEXT:
case TargetOpcode::G_FPTRUNC: {
bool Unused;
LLT DstTy = MRI.getType(MI.getOperand(0).getReg());
diff --git a/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp b/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp
index 3f6813e..04d9309 100644
--- a/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp
@@ -112,7 +112,7 @@ APInt GISelValueTracking::getKnownOnes(Register R) {
return getKnownBits(R).One;
}
-LLVM_ATTRIBUTE_UNUSED static void
+[[maybe_unused]] static void
dumpResult(const MachineInstr &MI, const KnownBits &Known, unsigned Depth) {
dbgs() << "[" << Depth << "] Compute known bits: " << MI << "[" << Depth
<< "] Computed for: " << MI << "[" << Depth << "] Known: 0x"
@@ -344,6 +344,22 @@ void GISelValueTracking::computeKnownBitsImpl(Register R, KnownBits &Known,
Known = KnownBits::mul(Known, Known2);
break;
}
+ case TargetOpcode::G_UMULH: {
+ computeKnownBitsImpl(MI.getOperand(2).getReg(), Known, DemandedElts,
+ Depth + 1);
+ computeKnownBitsImpl(MI.getOperand(1).getReg(), Known2, DemandedElts,
+ Depth + 1);
+ Known = KnownBits::mulhu(Known, Known2);
+ break;
+ }
+ case TargetOpcode::G_SMULH: {
+ computeKnownBitsImpl(MI.getOperand(2).getReg(), Known, DemandedElts,
+ Depth + 1);
+ computeKnownBitsImpl(MI.getOperand(1).getReg(), Known2, DemandedElts,
+ Depth + 1);
+ Known = KnownBits::mulhs(Known, Known2);
+ break;
+ }
case TargetOpcode::G_SELECT: {
computeKnownBitsMin(MI.getOperand(2).getReg(), MI.getOperand(3).getReg(),
Known, DemandedElts, Depth + 1);
@@ -1959,6 +1975,81 @@ unsigned GISelValueTracking::computeNumSignBits(Register R,
break;
}
+ case TargetOpcode::G_SUB: {
+ Register Src2 = MI.getOperand(2).getReg();
+ unsigned Src2NumSignBits =
+ computeNumSignBits(Src2, DemandedElts, Depth + 1);
+ if (Src2NumSignBits == 1)
+ return 1; // Early out.
+
+ // Handle NEG.
+ Register Src1 = MI.getOperand(1).getReg();
+ KnownBits Known1 = getKnownBits(Src1, DemandedElts, Depth);
+ if (Known1.isZero()) {
+ KnownBits Known2 = getKnownBits(Src2, DemandedElts, Depth);
+ // If the input is known to be 0 or 1, the output is 0/-1, which is all
+ // sign bits set.
+ if ((Known2.Zero | 1).isAllOnes())
+ return TyBits;
+
+ // If the input is known to be positive (the sign bit is known clear),
+ // the output of the NEG has, at worst, the same number of sign bits as
+ // the input.
+ if (Known2.isNonNegative()) {
+ FirstAnswer = Src2NumSignBits;
+ break;
+ }
+
+ // Otherwise, we treat this like a SUB.
+ }
+
+ unsigned Src1NumSignBits =
+ computeNumSignBits(Src1, DemandedElts, Depth + 1);
+ if (Src1NumSignBits == 1)
+ return 1; // Early Out.
+
+ // Sub can have at most one carry bit. Thus we know that the output
+ // is, at worst, one more bit than the inputs.
+ FirstAnswer = std::min(Src1NumSignBits, Src2NumSignBits) - 1;
+ break;
+ }
+ case TargetOpcode::G_ADD: {
+ Register Src2 = MI.getOperand(2).getReg();
+ unsigned Src2NumSignBits =
+ computeNumSignBits(Src2, DemandedElts, Depth + 1);
+ if (Src2NumSignBits <= 2)
+ return 1; // Early out.
+
+ Register Src1 = MI.getOperand(1).getReg();
+ unsigned Src1NumSignBits =
+ computeNumSignBits(Src1, DemandedElts, Depth + 1);
+ if (Src1NumSignBits == 1)
+ return 1; // Early Out.
+
+ // Special case decrementing a value (ADD X, -1):
+ KnownBits Known2 = getKnownBits(Src2, DemandedElts, Depth);
+ if (Known2.isAllOnes()) {
+ KnownBits Known1 = getKnownBits(Src1, DemandedElts, Depth);
+ // If the input is known to be 0 or 1, the output is 0/-1, which is all
+ // sign bits set.
+ if ((Known1.Zero | 1).isAllOnes())
+ return TyBits;
+
+ // If we are subtracting one from a positive number, there is no carry
+ // out of the result.
+ if (Known1.isNonNegative()) {
+ FirstAnswer = Src1NumSignBits;
+ break;
+ }
+
+ // Otherwise, we treat this like an ADD.
+ }
+
+ // Add can have at most one carry bit. Thus we know that the output
+ // is, at worst, one more bit than the inputs.
+ FirstAnswer = std::min(Src1NumSignBits, Src2NumSignBits) - 1;
+ break;
+ }
case TargetOpcode::G_FCMP:
case TargetOpcode::G_ICMP: {
bool IsFP = Opcode == TargetOpcode::G_FCMP;
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizeMutations.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizeMutations.cpp
index 25c1db9..ded4df4 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizeMutations.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizeMutations.cpp
@@ -55,12 +55,10 @@ LegalizeMutation LegalizeMutations::changeElementCountTo(unsigned TypeIdx,
}
LegalizeMutation LegalizeMutations::changeElementCountTo(unsigned TypeIdx,
- LLT NewEltTy) {
+ ElementCount EC) {
return [=](const LegalityQuery &Query) {
const LLT OldTy = Query.Types[TypeIdx];
- ElementCount NewEltCount = NewEltTy.isVector() ? NewEltTy.getElementCount()
- : ElementCount::getFixed(1);
- return std::make_pair(TypeIdx, OldTy.changeElementCount(NewEltCount));
+ return std::make_pair(TypeIdx, OldTy.changeElementCount(EC));
};
}
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index cffaf7c..38ec83f 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -3292,8 +3292,7 @@ LegalizerHelper::widenScalar(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) {
if (TypeIdx != 2)
return UnableToLegalize;
Observer.changingInstr(MI);
- // TODO: Probably should be zext
- widenScalarSrc(MI, WideTy, 2, TargetOpcode::G_SEXT);
+ widenScalarSrc(MI, WideTy, 2, TargetOpcode::G_ZEXT);
Observer.changedInstr(MI);
return Legalized;
}
@@ -3325,8 +3324,7 @@ LegalizerHelper::widenScalar(MachineInstr &MI, unsigned TypeIdx, LLT WideTy) {
if (TypeIdx == 2) {
Observer.changingInstr(MI);
- // TODO: Probably should be zext
- widenScalarSrc(MI, WideTy, 3, TargetOpcode::G_SEXT);
+ widenScalarSrc(MI, WideTy, 3, TargetOpcode::G_ZEXT);
Observer.changedInstr(MI);
return Legalized;
}
diff --git a/llvm/lib/CodeGen/GlobalISel/Utils.cpp b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
index 055fdc6..ca82857 100644
--- a/llvm/lib/CodeGen/GlobalISel/Utils.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
@@ -818,8 +818,7 @@ bool llvm::isKnownNeverNaN(Register Val, const MachineRegisterInfo &MRI,
if (!DefMI)
return false;
- const TargetMachine& TM = DefMI->getMF()->getTarget();
- if (DefMI->getFlag(MachineInstr::FmNoNans) || TM.Options.NoNaNsFPMath)
+ if (DefMI->getFlag(MachineInstr::FmNoNans))
return true;
// If the value is a constant, we can obviously see if it is a NaN or not.
diff --git a/llvm/lib/CodeGen/GlobalMergeFunctions.cpp b/llvm/lib/CodeGen/GlobalMergeFunctions.cpp
index 47640c4a..81ab317 100644
--- a/llvm/lib/CodeGen/GlobalMergeFunctions.cpp
+++ b/llvm/lib/CodeGen/GlobalMergeFunctions.cpp
@@ -587,16 +587,12 @@ public:
} // namespace
char GlobalMergeFuncPassWrapper::ID = 0;
-INITIALIZE_PASS_BEGIN(GlobalMergeFuncPassWrapper, "global-merge-func",
- "Global merge function pass", false, false)
-INITIALIZE_PASS_END(GlobalMergeFuncPassWrapper, "global-merge-func",
- "Global merge function pass", false, false)
+INITIALIZE_PASS(GlobalMergeFuncPassWrapper, "global-merge-func",
+ "Global merge function pass", false, false)
-namespace llvm {
-ModulePass *createGlobalMergeFuncPass() {
+ModulePass *llvm::createGlobalMergeFuncPass() {
return new GlobalMergeFuncPassWrapper();
}
-} // namespace llvm
GlobalMergeFuncPassWrapper::GlobalMergeFuncPassWrapper() : ModulePass(ID) {
initializeGlobalMergeFuncPassWrapperPass(
diff --git a/llvm/lib/CodeGen/LiveIntervals.cpp b/llvm/lib/CodeGen/LiveIntervals.cpp
index 3485a27..d2f2c3e 100644
--- a/llvm/lib/CodeGen/LiveIntervals.cpp
+++ b/llvm/lib/CodeGen/LiveIntervals.cpp
@@ -101,15 +101,11 @@ static cl::opt<bool> EnablePrecomputePhysRegs(
static bool EnablePrecomputePhysRegs = false;
#endif // NDEBUG
-namespace llvm {
-
-cl::opt<bool> UseSegmentSetForPhysRegs(
+cl::opt<bool> llvm::UseSegmentSetForPhysRegs(
"use-segment-set-for-physregs", cl::Hidden, cl::init(true),
cl::desc(
"Use segment set for the computation of the live ranges of physregs."));
-} // end namespace llvm
-
void LiveIntervalsWrapperPass::getAnalysisUsage(AnalysisUsage &AU) const {
AU.setPreservesCFG();
AU.addPreserved<LiveVariablesWrapperPass>();
@@ -665,7 +661,10 @@ void LiveIntervals::extendToIndices(LiveRange &LR,
void LiveIntervals::pruneValue(LiveRange &LR, SlotIndex Kill,
SmallVectorImpl<SlotIndex> *EndPoints) {
LiveQueryResult LRQ = LR.Query(Kill);
- VNInfo *VNI = LRQ.valueOutOrDead();
+ // LR may have liveness reachable from early clobber slot, which may be
+ // only live-in instead of live-out of the instruction.
+ // For example, LR =[1r, 3r), Kill = 3e, we have to prune [3e, 3r) of LR.
+ VNInfo *VNI = LRQ.valueOutOrDead() ? LRQ.valueOutOrDead() : LRQ.valueIn();
if (!VNI)
return;
diff --git a/llvm/lib/CodeGen/MIR2Vec.cpp b/llvm/lib/CodeGen/MIR2Vec.cpp
index e859765..5c78d98 100644
--- a/llvm/lib/CodeGen/MIR2Vec.cpp
+++ b/llvm/lib/CodeGen/MIR2Vec.cpp
@@ -29,20 +29,17 @@ using namespace mir2vec;
STATISTIC(MIRVocabMissCounter,
"Number of lookups to MIR entities not present in the vocabulary");
-namespace llvm {
-namespace mir2vec {
-cl::OptionCategory MIR2VecCategory("MIR2Vec Options");
+cl::OptionCategory llvm::mir2vec::MIR2VecCategory("MIR2Vec Options");
// FIXME: Use a default vocab when not specified
static cl::opt<std::string>
VocabFile("mir2vec-vocab-path", cl::Optional,
cl::desc("Path to the vocabulary file for MIR2Vec"), cl::init(""),
cl::cat(MIR2VecCategory));
-cl::opt<float> OpcWeight("mir2vec-opc-weight", cl::Optional, cl::init(1.0),
- cl::desc("Weight for machine opcode embeddings"),
- cl::cat(MIR2VecCategory));
-} // namespace mir2vec
-} // namespace llvm
+cl::opt<float>
+ llvm::mir2vec::OpcWeight("mir2vec-opc-weight", cl::Optional, cl::init(1.0),
+ cl::desc("Weight for machine opcode embeddings"),
+ cl::cat(MIR2VecCategory));
//===----------------------------------------------------------------------===//
// Vocabulary Implementation
diff --git a/llvm/lib/CodeGen/MIRFSDiscriminator.cpp b/llvm/lib/CodeGen/MIRFSDiscriminator.cpp
index f5146f5..d988a2a 100644
--- a/llvm/lib/CodeGen/MIRFSDiscriminator.cpp
+++ b/llvm/lib/CodeGen/MIRFSDiscriminator.cpp
@@ -40,7 +40,7 @@ cl::opt<bool> ImprovedFSDiscriminator(
"improved-fs-discriminator", cl::Hidden, cl::init(false),
cl::desc("New FS discriminators encoding (incompatible with the original "
"encoding)"));
-}
+} // namespace llvm
char MIRAddFSDiscriminators::ID = 0;
diff --git a/llvm/lib/CodeGen/MIRNamerPass.cpp b/llvm/lib/CodeGen/MIRNamerPass.cpp
index bc65700..cbf8867 100644
--- a/llvm/lib/CodeGen/MIRNamerPass.cpp
+++ b/llvm/lib/CodeGen/MIRNamerPass.cpp
@@ -23,10 +23,6 @@
using namespace llvm;
-namespace llvm {
-extern char &MIRNamerID;
-} // namespace llvm
-
#define DEBUG_TYPE "mir-namer"
namespace {
@@ -53,10 +49,9 @@ public:
VRegRenamer Renamer(MF.getRegInfo());
- unsigned BBIndex = 0;
ReversePostOrderTraversal<MachineBasicBlock *> RPOT(&*MF.begin());
- for (auto &MBB : RPOT)
- Changed |= Renamer.renameVRegs(MBB, BBIndex++);
+ for (const auto &[BBIndex, MBB] : enumerate(RPOT))
+ Changed |= Renamer.renameVRegs(MBB, BBIndex);
return Changed;
}
@@ -66,10 +61,4 @@ public:
char MIRNamer::ID;
-char &llvm::MIRNamerID = MIRNamer::ID;
-
-INITIALIZE_PASS_BEGIN(MIRNamer, "mir-namer", "Rename Register Operands", false,
- false)
-
-INITIALIZE_PASS_END(MIRNamer, "mir-namer", "Rename Register Operands", false,
- false)
+INITIALIZE_PASS(MIRNamer, "mir-namer", "Rename Register Operands", false, false)
diff --git a/llvm/lib/CodeGen/MIRPrinter.cpp b/llvm/lib/CodeGen/MIRPrinter.cpp
index bf8a6cd..1d54d72 100644
--- a/llvm/lib/CodeGen/MIRPrinter.cpp
+++ b/llvm/lib/CodeGen/MIRPrinter.cpp
@@ -107,10 +107,8 @@ struct MFPrintState {
} // end anonymous namespace
-namespace llvm::yaml {
-
/// This struct serializes the LLVM IR module.
-template <> struct BlockScalarTraits<Module> {
+template <> struct yaml::BlockScalarTraits<Module> {
static void output(const Module &Mod, void *Ctxt, raw_ostream &OS) {
Mod.print(OS, nullptr);
}
@@ -121,8 +119,6 @@ template <> struct BlockScalarTraits<Module> {
}
};
-} // end namespace llvm::yaml
-
static void printRegMIR(Register Reg, yaml::StringValue &Dest,
const TargetRegisterInfo *TRI) {
raw_string_ostream OS(Dest.Value);
@@ -866,48 +862,46 @@ static void printMI(raw_ostream &OS, MFPrintState &State,
OS << TII->getName(MI.getOpcode());
- LS = ListSeparator();
+ // Print a space after the opcode if any additional tokens are printed.
+ LS = ListSeparator(", ", " ");
- if (I < E) {
- OS << ' ';
- for (; I < E; ++I) {
- OS << LS;
- printMIOperand(OS, State, MI, I, TRI, TII, ShouldPrintRegisterTies,
- PrintedTypes, MRI, /*PrintDef=*/true);
- }
+ for (; I < E; ++I) {
+ OS << LS;
+ printMIOperand(OS, State, MI, I, TRI, TII, ShouldPrintRegisterTies,
+ PrintedTypes, MRI, /*PrintDef=*/true);
}
// Print any optional symbols attached to this instruction as-if they were
// operands.
if (MCSymbol *PreInstrSymbol = MI.getPreInstrSymbol()) {
- OS << LS << " pre-instr-symbol ";
+ OS << LS << "pre-instr-symbol ";
MachineOperand::printSymbol(OS, *PreInstrSymbol);
}
if (MCSymbol *PostInstrSymbol = MI.getPostInstrSymbol()) {
- OS << LS << " post-instr-symbol ";
+ OS << LS << "post-instr-symbol ";
MachineOperand::printSymbol(OS, *PostInstrSymbol);
}
if (MDNode *HeapAllocMarker = MI.getHeapAllocMarker()) {
- OS << LS << " heap-alloc-marker ";
+ OS << LS << "heap-alloc-marker ";
HeapAllocMarker->printAsOperand(OS, State.MST);
}
if (MDNode *PCSections = MI.getPCSections()) {
- OS << LS << " pcsections ";
+ OS << LS << "pcsections ";
PCSections->printAsOperand(OS, State.MST);
}
if (MDNode *MMRA = MI.getMMRAMetadata()) {
- OS << LS << " mmra ";
+ OS << LS << "mmra ";
MMRA->printAsOperand(OS, State.MST);
}
if (uint32_t CFIType = MI.getCFIType())
- OS << LS << " cfi-type " << CFIType;
+ OS << LS << "cfi-type " << CFIType;
if (auto Num = MI.peekDebugInstrNum())
- OS << LS << " debug-instr-number " << Num;
+ OS << LS << "debug-instr-number " << Num;
if (PrintLocations) {
if (const DebugLoc &DL = MI.getDebugLoc()) {
- OS << LS << " debug-location ";
+ OS << LS << "debug-location ";
DL->printAsOperand(OS, State.MST);
}
}
diff --git a/llvm/lib/CodeGen/MLRegAllocEvictAdvisor.cpp b/llvm/lib/CodeGen/MLRegAllocEvictAdvisor.cpp
index b2731b69..a72c2c4 100644
--- a/llvm/lib/CodeGen/MLRegAllocEvictAdvisor.cpp
+++ b/llvm/lib/CodeGen/MLRegAllocEvictAdvisor.cpp
@@ -97,7 +97,9 @@ static const bool EnableDevelopmentFeatures = false;
/// this happens only in development mode. It's a no-op otherwise.
namespace llvm {
extern cl::opt<unsigned> EvictInterferenceCutoff;
+} // namespace llvm
+namespace {
class RegAllocScoring : public MachineFunctionPass {
public:
static char ID;
@@ -124,11 +126,12 @@ public:
/// Performs this pass
bool runOnMachineFunction(MachineFunction &) override;
};
+} // namespace
char RegAllocScoring::ID = 0;
-FunctionPass *createRegAllocScoringPass() { return new RegAllocScoring(); }
-
-} // namespace llvm
+FunctionPass *llvm::createRegAllocScoringPass() {
+ return new RegAllocScoring();
+}
INITIALIZE_PASS(RegAllocScoring, "regallocscoringpass",
"Register Allocation Scoring Pass", false, false)
diff --git a/llvm/lib/CodeGen/MachineBlockFrequencyInfo.cpp b/llvm/lib/CodeGen/MachineBlockFrequencyInfo.cpp
index e7fa082..26eb10f 100644
--- a/llvm/lib/CodeGen/MachineBlockFrequencyInfo.cpp
+++ b/llvm/lib/CodeGen/MachineBlockFrequencyInfo.cpp
@@ -29,7 +29,6 @@ using namespace llvm;
#define DEBUG_TYPE "machine-block-freq"
-namespace llvm {
static cl::opt<GVDAGType> ViewMachineBlockFreqPropagationDAG(
"view-machine-block-freq-propagation-dags", cl::Hidden,
cl::desc("Pop up a window to show a dag displaying how machine block "
@@ -44,6 +43,7 @@ static cl::opt<GVDAGType> ViewMachineBlockFreqPropagationDAG(
clEnumValN(GVDT_Count, "count", "display a graph using the real "
"profile count if available.")));
+namespace llvm {
// Similar option above, but used to control BFI display only after MBP pass
cl::opt<GVDAGType> ViewBlockLayoutWithBFI(
"view-block-layout-with-bfi", cl::Hidden,
@@ -69,15 +69,15 @@ extern cl::opt<std::string> ViewBlockFreqFuncName;
// Defined in Analysis/BlockFrequencyInfo.cpp: -view-hot-freq-perc=
extern cl::opt<unsigned> ViewHotFreqPercent;
-static cl::opt<bool> PrintMachineBlockFreq(
- "print-machine-bfi", cl::init(false), cl::Hidden,
- cl::desc("Print the machine block frequency info."));
-
// Command line option to specify the name of the function for block frequency
// dump. Defined in Analysis/BlockFrequencyInfo.cpp.
extern cl::opt<std::string> PrintBFIFuncName;
} // namespace llvm
+static cl::opt<bool>
+ PrintMachineBlockFreq("print-machine-bfi", cl::init(false), cl::Hidden,
+ cl::desc("Print the machine block frequency info."));
+
static GVDAGType getGVDT() {
if (ViewBlockLayoutWithBFI != GVDT_None)
return ViewBlockLayoutWithBFI;
@@ -85,9 +85,7 @@ static GVDAGType getGVDT() {
return ViewMachineBlockFreqPropagationDAG;
}
-namespace llvm {
-
-template <> struct GraphTraits<MachineBlockFrequencyInfo *> {
+template <> struct llvm::GraphTraits<MachineBlockFrequencyInfo *> {
using NodeRef = const MachineBasicBlock *;
using ChildIteratorType = MachineBasicBlock::const_succ_iterator;
using nodes_iterator = pointer_iterator<MachineFunction::const_iterator>;
@@ -116,7 +114,7 @@ using MBFIDOTGraphTraitsBase =
MachineBranchProbabilityInfo>;
template <>
-struct DOTGraphTraits<MachineBlockFrequencyInfo *>
+struct llvm::DOTGraphTraits<MachineBlockFrequencyInfo *>
: public MBFIDOTGraphTraitsBase {
const MachineFunction *CurFunc = nullptr;
DenseMap<const MachineBasicBlock *, int> LayoutOrderMap;
@@ -159,8 +157,6 @@ struct DOTGraphTraits<MachineBlockFrequencyInfo *>
}
};
-} // end namespace llvm
-
AnalysisKey MachineBlockFrequencyAnalysis::Key;
MachineBlockFrequencyAnalysis::Result
diff --git a/llvm/lib/CodeGen/MachineBranchProbabilityInfo.cpp b/llvm/lib/CodeGen/MachineBranchProbabilityInfo.cpp
index 2e92dd8..7ca4582 100644
--- a/llvm/lib/CodeGen/MachineBranchProbabilityInfo.cpp
+++ b/llvm/lib/CodeGen/MachineBranchProbabilityInfo.cpp
@@ -18,13 +18,8 @@
using namespace llvm;
-INITIALIZE_PASS_BEGIN(MachineBranchProbabilityInfoWrapperPass,
- "machine-branch-prob",
- "Machine Branch Probability Analysis", false, true)
-INITIALIZE_PASS_END(MachineBranchProbabilityInfoWrapperPass,
- "machine-branch-prob",
- "Machine Branch Probability Analysis", false, true)
-
+INITIALIZE_PASS(MachineBranchProbabilityInfoWrapperPass, "machine-branch-prob",
+ "Machine Branch Probability Analysis", false, true)
namespace llvm {
cl::opt<unsigned>
StaticLikelyProb("static-likely-prob",
diff --git a/llvm/lib/CodeGen/MachineCopyPropagation.cpp b/llvm/lib/CodeGen/MachineCopyPropagation.cpp
index e359831..ea08365 100644
--- a/llvm/lib/CodeGen/MachineCopyPropagation.cpp
+++ b/llvm/lib/CodeGen/MachineCopyPropagation.cpp
@@ -1257,7 +1257,7 @@ void MachineCopyPropagation::BackwardCopyPropagateBlock(
Tracker.clear();
}
-static void LLVM_ATTRIBUTE_UNUSED printSpillReloadChain(
+[[maybe_unused]] static void printSpillReloadChain(
DenseMap<MachineInstr *, SmallVector<MachineInstr *>> &SpillChain,
DenseMap<MachineInstr *, SmallVector<MachineInstr *>> &ReloadChain,
MachineInstr *Leader) {
diff --git a/llvm/lib/CodeGen/MachineFunction.cpp b/llvm/lib/CodeGen/MachineFunction.cpp
index 224231c..bfa5ab2 100644
--- a/llvm/lib/CodeGen/MachineFunction.cpp
+++ b/llvm/lib/CodeGen/MachineFunction.cpp
@@ -719,43 +719,41 @@ MachineFunction::CallSiteInfo::CallSiteInfo(const CallBase &CB) {
}
}
-namespace llvm {
+template <>
+struct llvm::DOTGraphTraits<const MachineFunction *>
+ : public DefaultDOTGraphTraits {
+ DOTGraphTraits(bool isSimple = false) : DefaultDOTGraphTraits(isSimple) {}
- template<>
- struct DOTGraphTraits<const MachineFunction*> : public DefaultDOTGraphTraits {
- DOTGraphTraits(bool isSimple = false) : DefaultDOTGraphTraits(isSimple) {}
+ static std::string getGraphName(const MachineFunction *F) {
+ return ("CFG for '" + F->getName() + "' function").str();
+ }
- static std::string getGraphName(const MachineFunction *F) {
- return ("CFG for '" + F->getName() + "' function").str();
+ std::string getNodeLabel(const MachineBasicBlock *Node,
+ const MachineFunction *Graph) {
+ std::string OutStr;
+ {
+ raw_string_ostream OSS(OutStr);
+
+ if (isSimple()) {
+ OSS << printMBBReference(*Node);
+ if (const BasicBlock *BB = Node->getBasicBlock())
+ OSS << ": " << BB->getName();
+ } else
+ Node->print(OSS);
}
- std::string getNodeLabel(const MachineBasicBlock *Node,
- const MachineFunction *Graph) {
- std::string OutStr;
- {
- raw_string_ostream OSS(OutStr);
-
- if (isSimple()) {
- OSS << printMBBReference(*Node);
- if (const BasicBlock *BB = Node->getBasicBlock())
- OSS << ": " << BB->getName();
- } else
- Node->print(OSS);
- }
-
- if (OutStr[0] == '\n') OutStr.erase(OutStr.begin());
-
- // Process string output to make it nicer...
- for (unsigned i = 0; i != OutStr.length(); ++i)
- if (OutStr[i] == '\n') { // Left justify
- OutStr[i] = '\\';
- OutStr.insert(OutStr.begin()+i+1, 'l');
- }
- return OutStr;
- }
- };
+ if (OutStr[0] == '\n')
+ OutStr.erase(OutStr.begin());
-} // end namespace llvm
+ // Process string output to make it nicer...
+ for (unsigned i = 0; i != OutStr.length(); ++i)
+ if (OutStr[i] == '\n') { // Left justify
+ OutStr[i] = '\\';
+ OutStr.insert(OutStr.begin() + i + 1, 'l');
+ }
+ return OutStr;
+ }
+};
void MachineFunction::viewCFG() const
{
diff --git a/llvm/lib/CodeGen/MachineFunctionPrinterPass.cpp b/llvm/lib/CodeGen/MachineFunctionPrinterPass.cpp
index 0f88a7b..5111322 100644
--- a/llvm/lib/CodeGen/MachineFunctionPrinterPass.cpp
+++ b/llvm/lib/CodeGen/MachineFunctionPrinterPass.cpp
@@ -60,13 +60,11 @@ char &llvm::MachineFunctionPrinterPassID = MachineFunctionPrinterPass::ID;
INITIALIZE_PASS(MachineFunctionPrinterPass, "machineinstr-printer",
"Machine Function Printer", false, false)
-namespace llvm {
/// Returns a newly-created MachineFunction Printer pass. The
/// default banner is empty.
///
-MachineFunctionPass *createMachineFunctionPrinterPass(raw_ostream &OS,
- const std::string &Banner){
+MachineFunctionPass *
+llvm::createMachineFunctionPrinterPass(raw_ostream &OS,
+ const std::string &Banner) {
return new MachineFunctionPrinterPass(OS, Banner);
}
-
-}
diff --git a/llvm/lib/CodeGen/MachineOutliner.cpp b/llvm/lib/CodeGen/MachineOutliner.cpp
index fdae3b4..9feb974 100644
--- a/llvm/lib/CodeGen/MachineOutliner.cpp
+++ b/llvm/lib/CodeGen/MachineOutliner.cpp
@@ -593,15 +593,12 @@ struct MachineOutliner : public ModulePass {
char MachineOutliner::ID = 0;
-namespace llvm {
-ModulePass *createMachineOutlinerPass(RunOutliner RunOutlinerMode) {
+ModulePass *llvm::createMachineOutlinerPass(RunOutliner RunOutlinerMode) {
MachineOutliner *OL = new MachineOutliner();
OL->RunOutlinerMode = RunOutlinerMode;
return OL;
}
-} // namespace llvm
-
INITIALIZE_PASS(MachineOutliner, DEBUG_TYPE, "Machine Function Outliner", false,
false)
diff --git a/llvm/lib/CodeGen/MachinePipeliner.cpp b/llvm/lib/CodeGen/MachinePipeliner.cpp
index 89ed4da..a717d9e 100644
--- a/llvm/lib/CodeGen/MachinePipeliner.cpp
+++ b/llvm/lib/CodeGen/MachinePipeliner.cpp
@@ -201,16 +201,15 @@ static cl::opt<unsigned> SwpMaxNumStores(
cl::desc("Maximum number of stores allwed in the target loop."), cl::Hidden,
cl::init(200));
-namespace llvm {
-
// A command line option to enable the CopyToPhi DAG mutation.
-cl::opt<bool> SwpEnableCopyToPhi("pipeliner-enable-copytophi", cl::ReallyHidden,
- cl::init(true),
- cl::desc("Enable CopyToPhi DAG Mutation"));
+cl::opt<bool>
+ llvm::SwpEnableCopyToPhi("pipeliner-enable-copytophi", cl::ReallyHidden,
+ cl::init(true),
+ cl::desc("Enable CopyToPhi DAG Mutation"));
/// A command line argument to force pipeliner to use specified issue
/// width.
-cl::opt<int> SwpForceIssueWidth(
+cl::opt<int> llvm::SwpForceIssueWidth(
"pipeliner-force-issue-width",
cl::desc("Force pipeliner to use specified issue width."), cl::Hidden,
cl::init(-1));
@@ -226,8 +225,6 @@ static cl::opt<WindowSchedulingFlag> WindowSchedulingOption(
clEnumValN(WindowSchedulingFlag::WS_Force, "force",
"Use window algorithm instead of SMS algorithm.")));
-} // end namespace llvm
-
unsigned SwingSchedulerDAG::Circuits::MaxPaths = 5;
char MachinePipeliner::ID = 0;
#ifndef NDEBUG
diff --git a/llvm/lib/CodeGen/MachineScheduler.cpp b/llvm/lib/CodeGen/MachineScheduler.cpp
index 299bcc4..3ed1045 100644
--- a/llvm/lib/CodeGen/MachineScheduler.cpp
+++ b/llvm/lib/CodeGen/MachineScheduler.cpp
@@ -176,9 +176,7 @@ STATISTIC(NumNodeOrderPostRA,
STATISTIC(NumFirstValidPostRA,
"Number of scheduling units chosen for FirstValid heuristic post-RA");
-namespace llvm {
-
-cl::opt<MISched::Direction> PreRADirection(
+cl::opt<MISched::Direction> llvm::PreRADirection(
"misched-prera-direction", cl::Hidden,
cl::desc("Pre reg-alloc list scheduling direction"),
cl::init(MISched::Unspecified),
@@ -206,33 +204,31 @@ static cl::opt<bool>
DumpCriticalPathLength("misched-dcpl", cl::Hidden,
cl::desc("Print critical path length to stdout"));
-cl::opt<bool> VerifyScheduling(
+cl::opt<bool> llvm::VerifyScheduling(
"verify-misched", cl::Hidden,
cl::desc("Verify machine instrs before and after machine scheduling"));
#ifndef NDEBUG
-cl::opt<bool> ViewMISchedDAGs(
+cl::opt<bool> llvm::ViewMISchedDAGs(
"view-misched-dags", cl::Hidden,
cl::desc("Pop up a window to show MISched dags after they are processed"));
-cl::opt<bool> PrintDAGs("misched-print-dags", cl::Hidden,
- cl::desc("Print schedule DAGs"));
-cl::opt<bool> MISchedDumpReservedCycles(
+cl::opt<bool> llvm::PrintDAGs("misched-print-dags", cl::Hidden,
+ cl::desc("Print schedule DAGs"));
+static cl::opt<bool> MISchedDumpReservedCycles(
"misched-dump-reserved-cycles", cl::Hidden, cl::init(false),
cl::desc("Dump resource usage at schedule boundary."));
-cl::opt<bool> MischedDetailResourceBooking(
+static cl::opt<bool> MischedDetailResourceBooking(
"misched-detail-resource-booking", cl::Hidden, cl::init(false),
cl::desc("Show details of invoking getNextResoufceCycle."));
#else
-const bool ViewMISchedDAGs = false;
-const bool PrintDAGs = false;
-const bool MischedDetailResourceBooking = false;
+const bool llvm::ViewMISchedDAGs = false;
+const bool llvm::PrintDAGs = false;
+static const bool MischedDetailResourceBooking = false;
#ifdef LLVM_ENABLE_DUMP
-const bool MISchedDumpReservedCycles = false;
+static const bool MISchedDumpReservedCycles = false;
#endif // LLVM_ENABLE_DUMP
#endif // NDEBUG
-} // end namespace llvm
-
#ifndef NDEBUG
/// In some situations a few uninteresting nodes depend on nearly all other
/// nodes in the graph, provide a cutoff to hide them.
@@ -2053,28 +2049,24 @@ public:
} // end anonymous namespace
-namespace llvm {
-
std::unique_ptr<ScheduleDAGMutation>
-createLoadClusterDAGMutation(const TargetInstrInfo *TII,
- const TargetRegisterInfo *TRI,
- bool ReorderWhileClustering) {
+llvm::createLoadClusterDAGMutation(const TargetInstrInfo *TII,
+ const TargetRegisterInfo *TRI,
+ bool ReorderWhileClustering) {
return EnableMemOpCluster ? std::make_unique<LoadClusterMutation>(
TII, TRI, ReorderWhileClustering)
: nullptr;
}
std::unique_ptr<ScheduleDAGMutation>
-createStoreClusterDAGMutation(const TargetInstrInfo *TII,
- const TargetRegisterInfo *TRI,
- bool ReorderWhileClustering) {
+llvm::createStoreClusterDAGMutation(const TargetInstrInfo *TII,
+ const TargetRegisterInfo *TRI,
+ bool ReorderWhileClustering) {
return EnableMemOpCluster ? std::make_unique<StoreClusterMutation>(
TII, TRI, ReorderWhileClustering)
: nullptr;
}
-} // end namespace llvm
-
// Sorting all the loads/stores first, then for each load/store, checking the
// following load/store one by one, until reach the first non-dependent one and
// call target hook to see if they can cluster.
@@ -2304,16 +2296,12 @@ protected:
} // end anonymous namespace
-namespace llvm {
-
std::unique_ptr<ScheduleDAGMutation>
-createCopyConstrainDAGMutation(const TargetInstrInfo *TII,
- const TargetRegisterInfo *TRI) {
+llvm::createCopyConstrainDAGMutation(const TargetInstrInfo *TII,
+ const TargetRegisterInfo *TRI) {
return std::make_unique<CopyConstrain>(TII, TRI);
}
-} // end namespace llvm
-
/// constrainLocalCopy handles two possibilities:
/// 1) Local src:
/// I0: = dst
@@ -3445,14 +3433,13 @@ void GenericSchedulerBase::traceCandidate(const SchedCandidate &Cand) {
}
#endif
-namespace llvm {
/// Return true if this heuristic determines order.
/// TODO: Consider refactor return type of these functions as integer or enum,
/// as we may need to differentiate whether TryCand is better than Cand.
-bool tryLess(int TryVal, int CandVal,
- GenericSchedulerBase::SchedCandidate &TryCand,
- GenericSchedulerBase::SchedCandidate &Cand,
- GenericSchedulerBase::CandReason Reason) {
+bool llvm::tryLess(int TryVal, int CandVal,
+ GenericSchedulerBase::SchedCandidate &TryCand,
+ GenericSchedulerBase::SchedCandidate &Cand,
+ GenericSchedulerBase::CandReason Reason) {
if (TryVal < CandVal) {
TryCand.Reason = Reason;
return true;
@@ -3465,10 +3452,10 @@ bool tryLess(int TryVal, int CandVal,
return false;
}
-bool tryGreater(int TryVal, int CandVal,
- GenericSchedulerBase::SchedCandidate &TryCand,
- GenericSchedulerBase::SchedCandidate &Cand,
- GenericSchedulerBase::CandReason Reason) {
+bool llvm::tryGreater(int TryVal, int CandVal,
+ GenericSchedulerBase::SchedCandidate &TryCand,
+ GenericSchedulerBase::SchedCandidate &Cand,
+ GenericSchedulerBase::CandReason Reason) {
if (TryVal > CandVal) {
TryCand.Reason = Reason;
return true;
@@ -3481,9 +3468,9 @@ bool tryGreater(int TryVal, int CandVal,
return false;
}
-bool tryLatency(GenericSchedulerBase::SchedCandidate &TryCand,
- GenericSchedulerBase::SchedCandidate &Cand,
- SchedBoundary &Zone) {
+bool llvm::tryLatency(GenericSchedulerBase::SchedCandidate &TryCand,
+ GenericSchedulerBase::SchedCandidate &Cand,
+ SchedBoundary &Zone) {
if (Zone.isTop()) {
// Prefer the candidate with the lesser depth, but only if one of them has
// depth greater than the total latency scheduled so far, otherwise either
@@ -3513,7 +3500,6 @@ bool tryLatency(GenericSchedulerBase::SchedCandidate &TryCand,
}
return false;
}
-} // end namespace llvm
static void tracePick(GenericSchedulerBase::CandReason Reason, bool IsTop,
bool IsPostRA = false) {
@@ -3798,14 +3784,12 @@ void GenericScheduler::registerRoots() {
}
}
-namespace llvm {
-bool tryPressure(const PressureChange &TryP,
- const PressureChange &CandP,
- GenericSchedulerBase::SchedCandidate &TryCand,
- GenericSchedulerBase::SchedCandidate &Cand,
- GenericSchedulerBase::CandReason Reason,
- const TargetRegisterInfo *TRI,
- const MachineFunction &MF) {
+bool llvm::tryPressure(const PressureChange &TryP, const PressureChange &CandP,
+ GenericSchedulerBase::SchedCandidate &TryCand,
+ GenericSchedulerBase::SchedCandidate &Cand,
+ GenericSchedulerBase::CandReason Reason,
+ const TargetRegisterInfo *TRI,
+ const MachineFunction &MF) {
// If one candidate decreases and the other increases, go with it.
// Invalid candidates have UnitInc==0.
if (tryGreater(TryP.getUnitInc() < 0, CandP.getUnitInc() < 0, TryCand, Cand,
@@ -3838,7 +3822,7 @@ bool tryPressure(const PressureChange &TryP,
return tryGreater(TryRank, CandRank, TryCand, Cand, Reason);
}
-unsigned getWeakLeft(const SUnit *SU, bool isTop) {
+unsigned llvm::getWeakLeft(const SUnit *SU, bool isTop) {
return (isTop) ? SU->WeakPredsLeft : SU->WeakSuccsLeft;
}
@@ -3849,7 +3833,7 @@ unsigned getWeakLeft(const SUnit *SU, bool isTop) {
/// copies which can be prescheduled. The rest (e.g. x86 MUL) could be bundled
/// with the operation that produces or consumes the physreg. We'll do this when
/// regalloc has support for parallel copies.
-int biasPhysReg(const SUnit *SU, bool isTop) {
+int llvm::biasPhysReg(const SUnit *SU, bool isTop) {
const MachineInstr *MI = SU->getInstr();
if (MI->isCopy()) {
@@ -3884,7 +3868,6 @@ int biasPhysReg(const SUnit *SU, bool isTop) {
return 0;
}
-} // end namespace llvm
void GenericScheduler::initCandidate(SchedCandidate &Cand, SUnit *SU,
bool AtTop,
@@ -4812,13 +4795,13 @@ static MachineSchedRegistry ShufflerRegistry(
//===----------------------------------------------------------------------===//
#ifndef NDEBUG
-namespace llvm {
-template<> struct GraphTraits<
- ScheduleDAGMI*> : public GraphTraits<ScheduleDAG*> {};
+template <>
+struct llvm::GraphTraits<ScheduleDAGMI *> : public GraphTraits<ScheduleDAG *> {
+};
-template<>
-struct DOTGraphTraits<ScheduleDAGMI*> : public DefaultDOTGraphTraits {
+template <>
+struct llvm::DOTGraphTraits<ScheduleDAGMI *> : public DefaultDOTGraphTraits {
DOTGraphTraits(bool isSimple = false) : DefaultDOTGraphTraits(isSimple) {}
static std::string getGraphName(const ScheduleDAG *G) {
@@ -4878,7 +4861,6 @@ struct DOTGraphTraits<ScheduleDAGMI*> : public DefaultDOTGraphTraits {
}
};
-} // end namespace llvm
#endif // NDEBUG
/// viewGraph - Pop up a ghostview window with the reachable parts of the DAG
diff --git a/llvm/lib/CodeGen/MachineSink.cpp b/llvm/lib/CodeGen/MachineSink.cpp
index d5153b7..cdcb29d9 100644
--- a/llvm/lib/CodeGen/MachineSink.cpp
+++ b/llvm/lib/CodeGen/MachineSink.cpp
@@ -1209,7 +1209,7 @@ MachineSinking::getBBRegisterPressure(const MachineBasicBlock &MBB,
MIE = MBB.instr_begin();
MII != MIE; --MII) {
const MachineInstr &MI = *std::prev(MII);
- if (MI.isDebugInstr() || MI.isPseudoProbe())
+ if (MI.isDebugOrPseudoInstr())
continue;
RegisterOperands RegOpers;
RegOpers.collect(MI, *TRI, *MRI, false, false);
diff --git a/llvm/lib/CodeGen/MachineTraceMetrics.cpp b/llvm/lib/CodeGen/MachineTraceMetrics.cpp
index c2d4aa0..9ac3f741 100644
--- a/llvm/lib/CodeGen/MachineTraceMetrics.cpp
+++ b/llvm/lib/CodeGen/MachineTraceMetrics.cpp
@@ -485,10 +485,7 @@ struct LoopBounds {
// Specialize po_iterator_storage in order to prune the post-order traversal so
// it is limited to the current loop and doesn't traverse the loop back edges.
-namespace llvm {
-
-template<>
-class po_iterator_storage<LoopBounds, true> {
+template <> class llvm::po_iterator_storage<LoopBounds, true> {
LoopBounds &LB;
public:
@@ -519,8 +516,6 @@ public:
}
};
-} // end namespace llvm
-
/// Compute the trace through MBB.
void MachineTraceMetrics::Ensemble::computeTrace(const MachineBasicBlock *MBB) {
LLVM_DEBUG(dbgs() << "Computing " << getName() << " trace through "
diff --git a/llvm/lib/CodeGen/NonRelocatableStringpool.cpp b/llvm/lib/CodeGen/NonRelocatableStringpool.cpp
index 087ac62..59c587c 100644
--- a/llvm/lib/CodeGen/NonRelocatableStringpool.cpp
+++ b/llvm/lib/CodeGen/NonRelocatableStringpool.cpp
@@ -9,7 +9,7 @@
#include "llvm/CodeGen/NonRelocatableStringpool.h"
#include "llvm/ADT/STLExtras.h"
-namespace llvm {
+using namespace llvm;
DwarfStringPoolEntryRef NonRelocatableStringpool::getEntry(StringRef S) {
auto I = Strings.try_emplace(S);
@@ -43,5 +43,3 @@ NonRelocatableStringpool::getEntriesForEmission() const {
});
return Result;
}
-
-} // namespace llvm
diff --git a/llvm/lib/CodeGen/RegisterCoalescer.cpp b/llvm/lib/CodeGen/RegisterCoalescer.cpp
index ebfea8e..e17a214b 100644
--- a/llvm/lib/CodeGen/RegisterCoalescer.cpp
+++ b/llvm/lib/CodeGen/RegisterCoalescer.cpp
@@ -2051,6 +2051,12 @@ bool RegisterCoalescer::joinCopy(
}
if (CP.getNewRC()) {
+ if (RegClassInfo.getNumAllocatableRegs(CP.getNewRC()) == 0) {
+ LLVM_DEBUG(dbgs() << "\tNo " << TRI->getRegClassName(CP.getNewRC())
+ << "are available for allocation\n");
+ return false;
+ }
+
auto SrcRC = MRI->getRegClass(CP.getSrcReg());
auto DstRC = MRI->getRegClass(CP.getDstReg());
unsigned SrcIdx = CP.getSrcIdx();
diff --git a/llvm/lib/CodeGen/RegisterPressure.cpp b/llvm/lib/CodeGen/RegisterPressure.cpp
index 5f37890..7d4674b 100644
--- a/llvm/lib/CodeGen/RegisterPressure.cpp
+++ b/llvm/lib/CodeGen/RegisterPressure.cpp
@@ -858,7 +858,7 @@ void RegPressureTracker::recedeSkipDebugValues() {
void RegPressureTracker::recede(SmallVectorImpl<VRegMaskOrUnit> *LiveUses) {
recedeSkipDebugValues();
- if (CurrPos->isDebugInstr() || CurrPos->isPseudoProbe()) {
+ if (CurrPos->isDebugOrPseudoInstr()) {
// It's possible to only have debug_value and pseudo probe instructions and
// hit the start of the block.
assert(CurrPos == MBB->begin());
diff --git a/llvm/lib/CodeGen/SafeStack.cpp b/llvm/lib/CodeGen/SafeStack.cpp
index 6f373a5..e9ffa85 100644
--- a/llvm/lib/CodeGen/SafeStack.cpp
+++ b/llvm/lib/CodeGen/SafeStack.cpp
@@ -76,8 +76,6 @@ using namespace llvm::safestack;
#define DEBUG_TYPE "safe-stack"
-namespace llvm {
-
STATISTIC(NumFunctions, "Total number of functions");
STATISTIC(NumUnsafeStackFunctions, "Number of functions with unsafe stack");
STATISTIC(NumUnsafeStackRestorePointsFunctions,
@@ -89,8 +87,6 @@ STATISTIC(NumUnsafeDynamicAllocas, "Number of unsafe dynamic allocas");
STATISTIC(NumUnsafeByValArguments, "Number of unsafe byval arguments");
STATISTIC(NumUnsafeStackRestorePoints, "Number of setjmps and landingpads");
-} // namespace llvm
-
/// Use __safestack_pointer_address even if the platform has a faster way of
/// access safe stack pointer.
static cl::opt<bool>
diff --git a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
index eae2e8c..9662511 100644
--- a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
+++ b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
@@ -1551,14 +1551,10 @@ LLVM_DUMP_METHOD void ILPValue::dump() const {
dbgs() << *this << '\n';
}
-namespace llvm {
-
-LLVM_ATTRIBUTE_UNUSED
-raw_ostream &operator<<(raw_ostream &OS, const ILPValue &Val) {
+[[maybe_unused]]
+raw_ostream &llvm::operator<<(raw_ostream &OS, const ILPValue &Val) {
Val.print(OS);
return OS;
}
-} // end namespace llvm
-
#endif
diff --git a/llvm/lib/CodeGen/ScheduleDAGPrinter.cpp b/llvm/lib/CodeGen/ScheduleDAGPrinter.cpp
index e7b1494..c80eade 100644
--- a/llvm/lib/CodeGen/ScheduleDAGPrinter.cpp
+++ b/llvm/lib/CodeGen/ScheduleDAGPrinter.cpp
@@ -16,57 +16,51 @@
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
-namespace llvm {
- template<>
- struct DOTGraphTraits<ScheduleDAG*> : public DefaultDOTGraphTraits {
+template <>
+struct llvm::DOTGraphTraits<ScheduleDAG *> : public DefaultDOTGraphTraits {
- DOTGraphTraits (bool isSimple=false) : DefaultDOTGraphTraits(isSimple) {}
+ DOTGraphTraits(bool isSimple = false) : DefaultDOTGraphTraits(isSimple) {}
- static std::string getGraphName(const ScheduleDAG *G) {
- return std::string(G->MF.getName());
- }
+ static std::string getGraphName(const ScheduleDAG *G) {
+ return std::string(G->MF.getName());
+ }
- static bool renderGraphFromBottomUp() {
- return true;
- }
+ static bool renderGraphFromBottomUp() { return true; }
- static bool isNodeHidden(const SUnit *Node, const ScheduleDAG *G) {
- return (Node->NumPreds > 10 || Node->NumSuccs > 10);
- }
+ static bool isNodeHidden(const SUnit *Node, const ScheduleDAG *G) {
+ return (Node->NumPreds > 10 || Node->NumSuccs > 10);
+ }
- static std::string getNodeIdentifierLabel(const SUnit *Node,
- const ScheduleDAG *Graph) {
- std::string R;
- raw_string_ostream OS(R);
- OS << static_cast<const void *>(Node);
- return R;
- }
+ static std::string getNodeIdentifierLabel(const SUnit *Node,
+ const ScheduleDAG *Graph) {
+ std::string R;
+ raw_string_ostream OS(R);
+ OS << static_cast<const void *>(Node);
+ return R;
+ }
- /// If you want to override the dot attributes printed for a particular
- /// edge, override this method.
- static std::string getEdgeAttributes(const SUnit *Node,
- SUnitIterator EI,
- const ScheduleDAG *Graph) {
- if (EI.isArtificialDep())
- return "color=cyan,style=dashed";
- if (EI.isCtrlDep())
- return "color=blue,style=dashed";
- return "";
- }
+ /// If you want to override the dot attributes printed for a particular
+ /// edge, override this method.
+ static std::string getEdgeAttributes(const SUnit *Node, SUnitIterator EI,
+ const ScheduleDAG *Graph) {
+ if (EI.isArtificialDep())
+ return "color=cyan,style=dashed";
+ if (EI.isCtrlDep())
+ return "color=blue,style=dashed";
+ return "";
+ }
+ std::string getNodeLabel(const SUnit *SU, const ScheduleDAG *Graph);
+ static std::string getNodeAttributes(const SUnit *N,
+ const ScheduleDAG *Graph) {
+ return "shape=Mrecord";
+ }
- std::string getNodeLabel(const SUnit *SU, const ScheduleDAG *Graph);
- static std::string getNodeAttributes(const SUnit *N,
- const ScheduleDAG *Graph) {
- return "shape=Mrecord";
- }
-
- static void addCustomGraphFeatures(ScheduleDAG *G,
- GraphWriter<ScheduleDAG*> &GW) {
- return G->addCustomGraphFeatures(GW);
- }
- };
-}
+ static void addCustomGraphFeatures(ScheduleDAG *G,
+ GraphWriter<ScheduleDAG *> &GW) {
+ return G->addCustomGraphFeatures(GW);
+ }
+};
std::string DOTGraphTraits<ScheduleDAG*>::getNodeLabel(const SUnit *SU,
const ScheduleDAG *G) {
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index c5c3866..c97300d 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -509,6 +509,7 @@ namespace {
SDValue visitFMUL(SDNode *N);
template <class MatchContextClass> SDValue visitFMA(SDNode *N);
SDValue visitFMAD(SDNode *N);
+ SDValue visitFMULADD(SDNode *N);
SDValue visitFDIV(SDNode *N);
SDValue visitFREM(SDNode *N);
SDValue visitFSQRT(SDNode *N);
@@ -657,13 +658,13 @@ namespace {
bool InexpensiveOnly = false,
std::optional<EVT> OutVT = std::nullopt);
SDValue BuildDivEstimate(SDValue N, SDValue Op, SDNodeFlags Flags);
- SDValue buildRsqrtEstimate(SDValue Op, SDNodeFlags Flags);
- SDValue buildSqrtEstimate(SDValue Op, SDNodeFlags Flags);
- SDValue buildSqrtEstimateImpl(SDValue Op, SDNodeFlags Flags, bool Recip);
+ SDValue buildRsqrtEstimate(SDValue Op);
+ SDValue buildSqrtEstimate(SDValue Op);
+ SDValue buildSqrtEstimateImpl(SDValue Op, bool Recip);
SDValue buildSqrtNROneConst(SDValue Arg, SDValue Est, unsigned Iterations,
- SDNodeFlags Flags, bool Reciprocal);
+ bool Reciprocal);
SDValue buildSqrtNRTwoConst(SDValue Arg, SDValue Est, unsigned Iterations,
- SDNodeFlags Flags, bool Reciprocal);
+ bool Reciprocal);
SDValue MatchBSwapHWordLow(SDNode *N, SDValue N0, SDValue N1,
bool DemandHighBits = true);
SDValue MatchBSwapHWord(SDNode *N, SDValue N0, SDValue N1);
@@ -1991,6 +1992,7 @@ SDValue DAGCombiner::visit(SDNode *N) {
case ISD::FMUL: return visitFMUL(N);
case ISD::FMA: return visitFMA<EmptyMatchContext>(N);
case ISD::FMAD: return visitFMAD(N);
+ case ISD::FMULADD: return visitFMULADD(N);
case ISD::FDIV: return visitFDIV(N);
case ISD::FREM: return visitFREM(N);
case ISD::FSQRT: return visitFSQRT(N);
@@ -5042,7 +5044,6 @@ static SDValue simplifyDivRem(SDNode *N, SelectionDAG &DAG) {
unsigned Opc = N->getOpcode();
bool IsDiv = (ISD::SDIV == Opc) || (ISD::UDIV == Opc);
- ConstantSDNode *N1C = isConstOrConstSplat(N1);
// X / undef -> undef
// X % undef -> undef
@@ -5074,7 +5075,7 @@ static SDValue simplifyDivRem(SDNode *N, SelectionDAG &DAG) {
// division-by-zero or remainder-by-zero, so assume the divisor is 1.
// TODO: Similarly, if we're zero-extending a boolean divisor, then assume
// it's a 1.
- if ((N1C && N1C->isOne()) || (VT.getScalarType() == MVT::i1))
+ if (isOneOrOneSplat(N1) || (VT.getScalarType() == MVT::i1))
return IsDiv ? N0 : DAG.getConstant(0, DL, VT);
return SDValue();
@@ -10628,7 +10629,7 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
// folding this will increase the total number of instructions.
if (N0.getOpcode() == ISD::SRL &&
(N0.getOperand(1) == N1 || N0.hasOneUse()) &&
- TLI.shouldFoldConstantShiftPairToMask(N, Level)) {
+ TLI.shouldFoldConstantShiftPairToMask(N)) {
if (ISD::matchBinaryPredicate(N1, N0.getOperand(1), MatchShiftAmount,
/*AllowUndefs*/ false,
/*AllowTypeMismatch*/ true)) {
@@ -11207,7 +11208,7 @@ SDValue DAGCombiner::visitSRL(SDNode *N) {
// fold (srl (shl x, c1), c2) -> (and (shl x, (sub c1, c2), MASK) or
// (and (srl x, (sub c2, c1), MASK)
if ((N0.getOperand(1) == N1 || N0->hasOneUse()) &&
- TLI.shouldFoldConstantShiftPairToMask(N, Level)) {
+ TLI.shouldFoldConstantShiftPairToMask(N)) {
auto MatchShiftAmount = [OpSizeInBits](ConstantSDNode *LHS,
ConstantSDNode *RHS) {
const APInt &LHSC = LHS->getAPIntValue();
@@ -17086,11 +17087,6 @@ static bool isContractableFMUL(const TargetOptions &Options, SDValue N) {
N->getFlags().hasAllowContract();
}
-// Returns true if `N` can assume no infinities involved in its computation.
-static bool hasNoInfs(const TargetOptions &Options, SDValue N) {
- return Options.NoInfsFPMath || N->getFlags().hasNoInfs();
-}
-
/// Try to perform FMA combining on a given FADD node.
template <class MatchContextClass>
SDValue DAGCombiner::visitFADDForFMACombine(SDNode *N) {
@@ -17666,7 +17662,7 @@ SDValue DAGCombiner::visitFMULForFMADistributiveCombine(SDNode *N) {
// The transforms below are incorrect when x == 0 and y == inf, because the
// intermediate multiplication produces a nan.
SDValue FAdd = N0.getOpcode() == ISD::FADD ? N0 : N1;
- if (!hasNoInfs(Options, FAdd))
+ if (!FAdd->getFlags().hasNoInfs())
return SDValue();
// Floating-point multiply-add without intermediate rounding.
@@ -17763,7 +17759,6 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
bool N1CFP = DAG.isConstantFPBuildVectorOrConstantFP(N1);
EVT VT = N->getValueType(0);
SDLoc DL(N);
- const TargetOptions &Options = DAG.getTarget().Options;
SDNodeFlags Flags = N->getFlags();
SelectionDAG::FlagInserter FlagsInserter(DAG, N);
@@ -17829,7 +17824,7 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
bool AllowNewConst = (Level < AfterLegalizeDAG);
// If nnan is enabled, fold lots of things.
- if ((Options.NoNaNsFPMath || Flags.hasNoNaNs()) && AllowNewConst) {
+ if (Flags.hasNoNaNs() && AllowNewConst) {
// If allowed, fold (fadd (fneg x), x) -> 0.0
if (N0.getOpcode() == ISD::FNEG && N0.getOperand(0) == N1)
return DAG.getConstantFP(0.0, DL, VT);
@@ -17978,7 +17973,6 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
ConstantFPSDNode *N1CFP = isConstOrConstSplatFP(N1, true);
EVT VT = N->getValueType(0);
SDLoc DL(N);
- const TargetOptions &Options = DAG.getTarget().Options;
const SDNodeFlags Flags = N->getFlags();
SelectionDAG::FlagInserter FlagsInserter(DAG, N);
@@ -18006,7 +18000,7 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
if (N0 == N1) {
// (fsub x, x) -> 0.0
- if (Options.NoNaNsFPMath || Flags.hasNoNaNs())
+ if (Flags.hasNoNaNs())
return DAG.getConstantFP(0.0f, DL, VT);
}
@@ -18317,7 +18311,6 @@ template <class MatchContextClass> SDValue DAGCombiner::visitFMA(SDNode *N) {
ConstantFPSDNode *N2CFP = dyn_cast<ConstantFPSDNode>(N2);
EVT VT = N->getValueType(0);
SDLoc DL(N);
- const TargetOptions &Options = DAG.getTarget().Options;
// FMA nodes have flags that propagate to the created nodes.
SelectionDAG::FlagInserter FlagsInserter(DAG, N);
MatchContextClass matcher(DAG, TLI, N);
@@ -18343,8 +18336,7 @@ template <class MatchContextClass> SDValue DAGCombiner::visitFMA(SDNode *N) {
return matcher.getNode(ISD::FMA, DL, VT, NegN0, NegN1, N2);
}
- if ((Options.NoNaNsFPMath && Options.NoInfsFPMath) ||
- (N->getFlags().hasNoNaNs() && N->getFlags().hasNoInfs())) {
+ if (N->getFlags().hasNoNaNs() && N->getFlags().hasNoInfs()) {
if (N->getFlags().hasNoSignedZeros() ||
(N2CFP && !N2CFP->isExactlyValue(-0.0))) {
if (N0CFP && N0CFP->isZero())
@@ -18449,6 +18441,21 @@ SDValue DAGCombiner::visitFMAD(SDNode *N) {
return SDValue();
}
+SDValue DAGCombiner::visitFMULADD(SDNode *N) {
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ SDValue N2 = N->getOperand(2);
+ EVT VT = N->getValueType(0);
+ SDLoc DL(N);
+
+ // Constant fold FMULADD.
+ if (SDValue C =
+ DAG.FoldConstantArithmetic(ISD::FMULADD, DL, VT, {N0, N1, N2}))
+ return C;
+
+ return SDValue();
+}
+
// Combine multiple FDIVs with the same divisor into multiple FMULs by the
// reciprocal.
// E.g., (a / D; b / D;) -> (recip = 1.0 / D; a * recip; b * recip)
@@ -18533,7 +18540,6 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
SDValue N1 = N->getOperand(1);
EVT VT = N->getValueType(0);
SDLoc DL(N);
- const TargetOptions &Options = DAG.getTarget().Options;
SDNodeFlags Flags = N->getFlags();
SelectionDAG::FlagInserter FlagsInserter(DAG, N);
@@ -18580,20 +18586,18 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
// If this FDIV is part of a reciprocal square root, it may be folded
// into a target-specific square root estimate instruction.
if (N1.getOpcode() == ISD::FSQRT) {
- if (SDValue RV = buildRsqrtEstimate(N1.getOperand(0), Flags))
+ if (SDValue RV = buildRsqrtEstimate(N1.getOperand(0)))
return DAG.getNode(ISD::FMUL, DL, VT, N0, RV);
} else if (N1.getOpcode() == ISD::FP_EXTEND &&
N1.getOperand(0).getOpcode() == ISD::FSQRT) {
- if (SDValue RV =
- buildRsqrtEstimate(N1.getOperand(0).getOperand(0), Flags)) {
+ if (SDValue RV = buildRsqrtEstimate(N1.getOperand(0).getOperand(0))) {
RV = DAG.getNode(ISD::FP_EXTEND, SDLoc(N1), VT, RV);
AddToWorklist(RV.getNode());
return DAG.getNode(ISD::FMUL, DL, VT, N0, RV);
}
} else if (N1.getOpcode() == ISD::FP_ROUND &&
N1.getOperand(0).getOpcode() == ISD::FSQRT) {
- if (SDValue RV =
- buildRsqrtEstimate(N1.getOperand(0).getOperand(0), Flags)) {
+ if (SDValue RV = buildRsqrtEstimate(N1.getOperand(0).getOperand(0))) {
RV = DAG.getNode(ISD::FP_ROUND, SDLoc(N1), VT, RV, N1.getOperand(1));
AddToWorklist(RV.getNode());
return DAG.getNode(ISD::FMUL, DL, VT, N0, RV);
@@ -18625,7 +18629,7 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
SDValue AA = DAG.getNode(ISD::FMUL, DL, VT, A, A);
SDValue AAZ =
DAG.getNode(ISD::FMUL, DL, VT, AA, Sqrt.getOperand(0));
- if (SDValue Rsqrt = buildRsqrtEstimate(AAZ, Flags))
+ if (SDValue Rsqrt = buildRsqrtEstimate(AAZ))
return DAG.getNode(ISD::FMUL, DL, VT, N0, Rsqrt);
// Estimate creation failed. Clean up speculatively created nodes.
@@ -18635,7 +18639,7 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
// We found a FSQRT, so try to make this fold:
// X / (Y * sqrt(Z)) -> X * (rsqrt(Z) / Y)
- if (SDValue Rsqrt = buildRsqrtEstimate(Sqrt.getOperand(0), Flags)) {
+ if (SDValue Rsqrt = buildRsqrtEstimate(Sqrt.getOperand(0))) {
SDValue Div = DAG.getNode(ISD::FDIV, SDLoc(N1), VT, Rsqrt, Y);
AddToWorklist(Div.getNode());
return DAG.getNode(ISD::FMUL, DL, VT, N0, Div);
@@ -18644,7 +18648,7 @@ SDValue DAGCombiner::visitFDIV(SDNode *N) {
}
// Fold into a reciprocal estimate and multiply instead of a real divide.
- if (Options.NoInfsFPMath || Flags.hasNoInfs())
+ if (Flags.hasNoInfs())
if (SDValue RV = BuildDivEstimate(N0, N1, Flags))
return RV;
}
@@ -18721,12 +18725,10 @@ SDValue DAGCombiner::visitFREM(SDNode *N) {
SDValue DAGCombiner::visitFSQRT(SDNode *N) {
SDNodeFlags Flags = N->getFlags();
- const TargetOptions &Options = DAG.getTarget().Options;
// Require 'ninf' flag since sqrt(+Inf) = +Inf, but the estimation goes as:
// sqrt(+Inf) == rsqrt(+Inf) * +Inf = 0 * +Inf = NaN
- if (!Flags.hasApproximateFuncs() ||
- (!Options.NoInfsFPMath && !Flags.hasNoInfs()))
+ if (!Flags.hasApproximateFuncs() || !Flags.hasNoInfs())
return SDValue();
SDValue N0 = N->getOperand(0);
@@ -18734,11 +18736,12 @@ SDValue DAGCombiner::visitFSQRT(SDNode *N) {
return SDValue();
// FSQRT nodes have flags that propagate to the created nodes.
+ SelectionDAG::FlagInserter FlagInserter(DAG, Flags);
// TODO: If this is N0/sqrt(N0), and we reach this node before trying to
// transform the fdiv, we may produce a sub-optimal estimate sequence
// because the reciprocal calculation may not have to filter out a
// 0.0 input.
- return buildSqrtEstimate(N0, Flags);
+ return buildSqrtEstimate(N0);
}
/// copysign(x, fp_extend(y)) -> copysign(x, y)
@@ -18870,27 +18873,38 @@ SDValue DAGCombiner::visitFPOW(SDNode *N) {
static SDValue foldFPToIntToFP(SDNode *N, const SDLoc &DL, SelectionDAG &DAG,
const TargetLowering &TLI) {
- // We only do this if the target has legal ftrunc. Otherwise, we'd likely be
- // replacing casts with a libcall. We also must be allowed to ignore -0.0
- // because FTRUNC will return -0.0 for (-1.0, -0.0), but using integer
- // conversions would return +0.0.
+ // We can fold the fpto[us]i -> [us]itofp pattern into a single ftrunc.
+ // If NoSignedZerosFPMath is enabled, this is a direct replacement.
+ // Otherwise, for strict math, we must handle edge cases:
+ // 1. For unsigned conversions, use FABS to handle negative cases. Take -0.0
+ // as example, it first becomes integer 0, and is converted back to +0.0.
+ // FTRUNC on its own could produce -0.0.
+
// FIXME: We should be able to use node-level FMF here.
- // TODO: If strict math, should we use FABS (+ range check for signed cast)?
EVT VT = N->getValueType(0);
- if (!TLI.isOperationLegal(ISD::FTRUNC, VT) ||
- !DAG.getTarget().Options.NoSignedZerosFPMath)
+ if (!TLI.isOperationLegal(ISD::FTRUNC, VT))
return SDValue();
// fptosi/fptoui round towards zero, so converting from FP to integer and
// back is the same as an 'ftrunc': [us]itofp (fpto[us]i X) --> ftrunc X
SDValue N0 = N->getOperand(0);
if (N->getOpcode() == ISD::SINT_TO_FP && N0.getOpcode() == ISD::FP_TO_SINT &&
- N0.getOperand(0).getValueType() == VT)
- return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+ N0.getOperand(0).getValueType() == VT) {
+ if (DAG.getTarget().Options.NoSignedZerosFPMath)
+ return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+ }
if (N->getOpcode() == ISD::UINT_TO_FP && N0.getOpcode() == ISD::FP_TO_UINT &&
- N0.getOperand(0).getValueType() == VT)
- return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+ N0.getOperand(0).getValueType() == VT) {
+ if (DAG.getTarget().Options.NoSignedZerosFPMath)
+ return DAG.getNode(ISD::FTRUNC, DL, VT, N0.getOperand(0));
+
+ // Strict math: use FABS to handle negative inputs correctly.
+ if (TLI.isFAbsFree(VT)) {
+ SDValue Abs = DAG.getNode(ISD::FABS, DL, VT, N0.getOperand(0));
+ return DAG.getNode(ISD::FTRUNC, DL, VT, Abs);
+ }
+ }
return SDValue();
}
@@ -19340,8 +19354,10 @@ SDValue DAGCombiner::visitFMinMax(SDNode *N) {
EVT VT = N->getValueType(0);
const SDNodeFlags Flags = N->getFlags();
unsigned Opc = N->getOpcode();
- bool PropagatesNaN = Opc == ISD::FMINIMUM || Opc == ISD::FMAXIMUM;
- bool IsMin = Opc == ISD::FMINNUM || Opc == ISD::FMINIMUM;
+ bool PropAllNaNsToQNaNs = Opc == ISD::FMINIMUM || Opc == ISD::FMAXIMUM;
+ bool PropOnlySNaNsToQNaNs = Opc == ISD::FMINNUM || Opc == ISD::FMAXNUM;
+ bool IsMin =
+ Opc == ISD::FMINNUM || Opc == ISD::FMINIMUM || Opc == ISD::FMINIMUMNUM;
SelectionDAG::FlagInserter FlagsInserter(DAG, N);
// Constant fold.
@@ -19356,34 +19372,53 @@ SDValue DAGCombiner::visitFMinMax(SDNode *N) {
if (const ConstantFPSDNode *N1CFP = isConstOrConstSplatFP(N1)) {
const APFloat &AF = N1CFP->getValueAPF();
- // minnum(X, nan) -> X
- // maxnum(X, nan) -> X
- // minimum(X, nan) -> nan
- // maximum(X, nan) -> nan
- if (AF.isNaN())
- return PropagatesNaN ? N->getOperand(1) : N->getOperand(0);
+ // minnum(X, qnan) -> X
+ // maxnum(X, qnan) -> X
+ // minnum(X, snan) -> qnan
+ // maxnum(X, snan) -> qnan
+ // minimum(X, nan) -> qnan
+ // maximum(X, nan) -> qnan
+ // minimumnum(X, nan) -> X
+ // maximumnum(X, nan) -> X
+ if (AF.isNaN()) {
+ if (PropAllNaNsToQNaNs || (AF.isSignaling() && PropOnlySNaNsToQNaNs)) {
+ if (AF.isSignaling())
+ return DAG.getConstantFP(AF.makeQuiet(), SDLoc(N), VT);
+ return N->getOperand(1);
+ }
+ return N->getOperand(0);
+ }
// In the following folds, inf can be replaced with the largest finite
// float, if the ninf flag is set.
if (AF.isInfinity() || (Flags.hasNoInfs() && AF.isLargest())) {
- // minnum(X, -inf) -> -inf
- // maxnum(X, +inf) -> +inf
+ // minnum(X, -inf) -> -inf (ignoring sNaN -> qNaN propagation)
+ // maxnum(X, +inf) -> +inf (ignoring sNaN -> qNaN propagation)
// minimum(X, -inf) -> -inf if nnan
// maximum(X, +inf) -> +inf if nnan
- if (IsMin == AF.isNegative() && (!PropagatesNaN || Flags.hasNoNaNs()))
+ // minimumnum(X, -inf) -> -inf
+ // maximumnum(X, +inf) -> +inf
+ if (IsMin == AF.isNegative() &&
+ (!PropAllNaNsToQNaNs || Flags.hasNoNaNs()))
return N->getOperand(1);
// minnum(X, +inf) -> X if nnan
// maxnum(X, -inf) -> X if nnan
- // minimum(X, +inf) -> X
- // maximum(X, -inf) -> X
- if (IsMin != AF.isNegative() && (PropagatesNaN || Flags.hasNoNaNs()))
+ // minimum(X, +inf) -> X (ignoring quieting of sNaNs)
+ // maximum(X, -inf) -> X (ignoring quieting of sNaNs)
+ // minimumnum(X, +inf) -> X if nnan
+ // maximumnum(X, -inf) -> X if nnan
+ if (IsMin != AF.isNegative() && (PropAllNaNsToQNaNs || Flags.hasNoNaNs()))
return N->getOperand(0);
}
}
+ // There are no VECREDUCE variants of FMINIMUMNUM or FMAXIMUMNUM
+ if (Opc == ISD::FMINIMUMNUM || Opc == ISD::FMAXIMUMNUM)
+ return SDValue();
+
if (SDValue SD = reassociateReduction(
- PropagatesNaN
+ PropAllNaNsToQNaNs
? (IsMin ? ISD::VECREDUCE_FMINIMUM : ISD::VECREDUCE_FMAXIMUM)
: (IsMin ? ISD::VECREDUCE_FMIN : ISD::VECREDUCE_FMAX),
Opc, SDLoc(N), VT, N0, N1, Flags))
@@ -29703,28 +29738,27 @@ SDValue DAGCombiner::BuildDivEstimate(SDValue N, SDValue Op,
/// X_{i+1} = X_i (1.5 - A X_i^2 / 2)
/// As a result, we precompute A/2 prior to the iteration loop.
SDValue DAGCombiner::buildSqrtNROneConst(SDValue Arg, SDValue Est,
- unsigned Iterations,
- SDNodeFlags Flags, bool Reciprocal) {
+ unsigned Iterations, bool Reciprocal) {
EVT VT = Arg.getValueType();
SDLoc DL(Arg);
SDValue ThreeHalves = DAG.getConstantFP(1.5, DL, VT);
// We now need 0.5 * Arg which we can write as (1.5 * Arg - Arg) so that
// this entire sequence requires only one FP constant.
- SDValue HalfArg = DAG.getNode(ISD::FMUL, DL, VT, ThreeHalves, Arg, Flags);
- HalfArg = DAG.getNode(ISD::FSUB, DL, VT, HalfArg, Arg, Flags);
+ SDValue HalfArg = DAG.getNode(ISD::FMUL, DL, VT, ThreeHalves, Arg);
+ HalfArg = DAG.getNode(ISD::FSUB, DL, VT, HalfArg, Arg);
// Newton iterations: Est = Est * (1.5 - HalfArg * Est * Est)
for (unsigned i = 0; i < Iterations; ++i) {
- SDValue NewEst = DAG.getNode(ISD::FMUL, DL, VT, Est, Est, Flags);
- NewEst = DAG.getNode(ISD::FMUL, DL, VT, HalfArg, NewEst, Flags);
- NewEst = DAG.getNode(ISD::FSUB, DL, VT, ThreeHalves, NewEst, Flags);
- Est = DAG.getNode(ISD::FMUL, DL, VT, Est, NewEst, Flags);
+ SDValue NewEst = DAG.getNode(ISD::FMUL, DL, VT, Est, Est);
+ NewEst = DAG.getNode(ISD::FMUL, DL, VT, HalfArg, NewEst);
+ NewEst = DAG.getNode(ISD::FSUB, DL, VT, ThreeHalves, NewEst);
+ Est = DAG.getNode(ISD::FMUL, DL, VT, Est, NewEst);
}
// If non-reciprocal square root is requested, multiply the result by Arg.
if (!Reciprocal)
- Est = DAG.getNode(ISD::FMUL, DL, VT, Est, Arg, Flags);
+ Est = DAG.getNode(ISD::FMUL, DL, VT, Est, Arg);
return Est;
}
@@ -29735,8 +29769,7 @@ SDValue DAGCombiner::buildSqrtNROneConst(SDValue Arg, SDValue Est,
/// =>
/// X_{i+1} = (-0.5 * X_i) * (A * X_i * X_i + (-3.0))
SDValue DAGCombiner::buildSqrtNRTwoConst(SDValue Arg, SDValue Est,
- unsigned Iterations,
- SDNodeFlags Flags, bool Reciprocal) {
+ unsigned Iterations, bool Reciprocal) {
EVT VT = Arg.getValueType();
SDLoc DL(Arg);
SDValue MinusThree = DAG.getConstantFP(-3.0, DL, VT);
@@ -29749,9 +29782,9 @@ SDValue DAGCombiner::buildSqrtNRTwoConst(SDValue Arg, SDValue Est,
// Newton iterations for reciprocal square root:
// E = (E * -0.5) * ((A * E) * E + -3.0)
for (unsigned i = 0; i < Iterations; ++i) {
- SDValue AE = DAG.getNode(ISD::FMUL, DL, VT, Arg, Est, Flags);
- SDValue AEE = DAG.getNode(ISD::FMUL, DL, VT, AE, Est, Flags);
- SDValue RHS = DAG.getNode(ISD::FADD, DL, VT, AEE, MinusThree, Flags);
+ SDValue AE = DAG.getNode(ISD::FMUL, DL, VT, Arg, Est);
+ SDValue AEE = DAG.getNode(ISD::FMUL, DL, VT, AE, Est);
+ SDValue RHS = DAG.getNode(ISD::FADD, DL, VT, AEE, MinusThree);
// When calculating a square root at the last iteration build:
// S = ((A * E) * -0.5) * ((A * E) * E + -3.0)
@@ -29759,13 +29792,13 @@ SDValue DAGCombiner::buildSqrtNRTwoConst(SDValue Arg, SDValue Est,
SDValue LHS;
if (Reciprocal || (i + 1) < Iterations) {
// RSQRT: LHS = (E * -0.5)
- LHS = DAG.getNode(ISD::FMUL, DL, VT, Est, MinusHalf, Flags);
+ LHS = DAG.getNode(ISD::FMUL, DL, VT, Est, MinusHalf);
} else {
// SQRT: LHS = (A * E) * -0.5
- LHS = DAG.getNode(ISD::FMUL, DL, VT, AE, MinusHalf, Flags);
+ LHS = DAG.getNode(ISD::FMUL, DL, VT, AE, MinusHalf);
}
- Est = DAG.getNode(ISD::FMUL, DL, VT, LHS, RHS, Flags);
+ Est = DAG.getNode(ISD::FMUL, DL, VT, LHS, RHS);
}
return Est;
@@ -29774,8 +29807,7 @@ SDValue DAGCombiner::buildSqrtNRTwoConst(SDValue Arg, SDValue Est,
/// Build code to calculate either rsqrt(Op) or sqrt(Op). In the latter case
/// Op*rsqrt(Op) is actually computed, so additional postprocessing is needed if
/// Op can be zero.
-SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, SDNodeFlags Flags,
- bool Reciprocal) {
+SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, bool Reciprocal) {
if (LegalDAG)
return SDValue();
@@ -29803,8 +29835,8 @@ SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, SDNodeFlags Flags,
if (Iterations > 0)
Est = UseOneConstNR
- ? buildSqrtNROneConst(Op, Est, Iterations, Flags, Reciprocal)
- : buildSqrtNRTwoConst(Op, Est, Iterations, Flags, Reciprocal);
+ ? buildSqrtNROneConst(Op, Est, Iterations, Reciprocal)
+ : buildSqrtNRTwoConst(Op, Est, Iterations, Reciprocal);
if (!Reciprocal) {
SDLoc DL(Op);
// Try the target specific test first.
@@ -29822,12 +29854,12 @@ SDValue DAGCombiner::buildSqrtEstimateImpl(SDValue Op, SDNodeFlags Flags,
return SDValue();
}
-SDValue DAGCombiner::buildRsqrtEstimate(SDValue Op, SDNodeFlags Flags) {
- return buildSqrtEstimateImpl(Op, Flags, true);
+SDValue DAGCombiner::buildRsqrtEstimate(SDValue Op) {
+ return buildSqrtEstimateImpl(Op, true);
}
-SDValue DAGCombiner::buildSqrtEstimate(SDValue Op, SDNodeFlags Flags) {
- return buildSqrtEstimateImpl(Op, Flags, false);
+SDValue DAGCombiner::buildSqrtEstimate(SDValue Op) {
+ return buildSqrtEstimateImpl(Op, false);
}
/// Return true if there is any possibility that the two addresses overlap.
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp
index b5f8a61..437d0f4 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp
@@ -3313,7 +3313,6 @@ void DAGTypeLegalizer::SoftPromoteHalfResult(SDNode *N, unsigned ResNo) {
case ISD::FP_ROUND: R = SoftPromoteHalfRes_FP_ROUND(N); break;
// Unary FP Operations
- case ISD::FABS:
case ISD::FACOS:
case ISD::FASIN:
case ISD::FATAN:
@@ -3329,7 +3328,6 @@ void DAGTypeLegalizer::SoftPromoteHalfResult(SDNode *N, unsigned ResNo) {
case ISD::FLOG2:
case ISD::FLOG10:
case ISD::FNEARBYINT:
- case ISD::FNEG:
case ISD::FREEZE:
case ISD::FRINT:
case ISD::FROUND:
@@ -3341,6 +3339,12 @@ void DAGTypeLegalizer::SoftPromoteHalfResult(SDNode *N, unsigned ResNo) {
case ISD::FTAN:
case ISD::FTANH:
case ISD::FCANONICALIZE: R = SoftPromoteHalfRes_UnaryOp(N); break;
+ case ISD::FABS:
+ R = SoftPromoteHalfRes_FABS(N);
+ break;
+ case ISD::FNEG:
+ R = SoftPromoteHalfRes_FNEG(N);
+ break;
case ISD::AssertNoFPClass:
R = SoftPromoteHalfRes_AssertNoFPClass(N);
break;
@@ -3670,6 +3674,24 @@ SDValue DAGTypeLegalizer::SoftPromoteHalfRes_UnaryOp(SDNode *N) {
return DAG.getNode(GetPromotionOpcode(NVT, OVT), dl, MVT::i16, Res);
}
+SDValue DAGTypeLegalizer::SoftPromoteHalfRes_FABS(SDNode *N) {
+ SDValue Op = GetSoftPromotedHalf(N->getOperand(0));
+ SDLoc dl(N);
+
+ // Clear the sign bit.
+ return DAG.getNode(ISD::AND, dl, MVT::i16, Op,
+ DAG.getConstant(0x7fff, dl, MVT::i16));
+}
+
+SDValue DAGTypeLegalizer::SoftPromoteHalfRes_FNEG(SDNode *N) {
+ SDValue Op = GetSoftPromotedHalf(N->getOperand(0));
+ SDLoc dl(N);
+
+ // Invert the sign bit.
+ return DAG.getNode(ISD::XOR, dl, MVT::i16, Op,
+ DAG.getConstant(0x8000, dl, MVT::i16));
+}
+
SDValue DAGTypeLegalizer::SoftPromoteHalfRes_AssertNoFPClass(SDNode *N) {
return GetSoftPromotedHalf(N->getOperand(0));
}
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h
index d580ce0..603dc34 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h
@@ -832,6 +832,8 @@ private:
SDValue SoftPromoteHalfRes_SELECT(SDNode *N);
SDValue SoftPromoteHalfRes_SELECT_CC(SDNode *N);
SDValue SoftPromoteHalfRes_UnaryOp(SDNode *N);
+ SDValue SoftPromoteHalfRes_FABS(SDNode *N);
+ SDValue SoftPromoteHalfRes_FNEG(SDNode *N);
SDValue SoftPromoteHalfRes_AssertNoFPClass(SDNode *N);
SDValue SoftPromoteHalfRes_XINT_TO_FP(SDNode *N);
SDValue SoftPromoteHalfRes_UNDEF(SDNode *N);
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 6ea2e27..90edaf3 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -5063,8 +5063,7 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
break;
case ISD::ADD:
case ISD::ADDC:
- // Add can have at most one carry bit. Thus we know that the output
- // is, at worst, one more bit than the inputs.
+ // TODO: Move Operand 1 check before Operand 0 check
Tmp = ComputeNumSignBits(Op.getOperand(0), DemandedElts, Depth + 1);
if (Tmp == 1) return 1; // Early out.
@@ -5088,6 +5087,9 @@ unsigned SelectionDAG::ComputeNumSignBits(SDValue Op, const APInt &DemandedElts,
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth + 1);
if (Tmp2 == 1) return 1; // Early out.
+
+ // Add can have at most one carry bit. Thus we know that the output
+ // is, at worst, one more bit than the inputs.
return std::min(Tmp, Tmp2) - 1;
case ISD::SUB:
Tmp2 = ComputeNumSignBits(Op.getOperand(1), DemandedElts, Depth + 1);
@@ -5767,11 +5769,7 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
// even if the nonan flag is dropped somewhere.
unsigned CCOp = Opcode == ISD::SETCC ? 2 : 4;
ISD::CondCode CCCode = cast<CondCodeSDNode>(Op.getOperand(CCOp))->get();
- if (((unsigned)CCCode & 0x10U))
- return true;
-
- const TargetOptions &Options = getTarget().Options;
- return Options.NoNaNsFPMath || Options.NoInfsFPMath;
+ return (unsigned)CCCode & 0x10U;
}
case ISD::OR:
@@ -5790,6 +5788,7 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
case ISD::FCOPYSIGN:
case ISD::FMA:
case ISD::FMAD:
+ case ISD::FMULADD:
case ISD::FP_EXTEND:
case ISD::FP_TO_SINT_SAT:
case ISD::FP_TO_UINT_SAT:
@@ -5908,6 +5907,7 @@ bool SelectionDAG::isKnownNeverNaN(SDValue Op, const APInt &DemandedElts,
case ISD::FCOSH:
case ISD::FTANH:
case ISD::FMA:
+ case ISD::FMULADD:
case ISD::FMAD: {
if (SNaN)
return true;
@@ -6405,8 +6405,9 @@ static SDValue foldCONCAT_VECTORS(const SDLoc &DL, EVT VT,
if (VT.isScalableVector())
return SDValue();
- // A CONCAT_VECTOR with all UNDEF/BUILD_VECTOR operands can be
- // simplified to one big BUILD_VECTOR.
+ // A CONCAT_VECTOR of scalar sources, such as UNDEF, BUILD_VECTOR and
+ // single-element INSERT_VECTOR_ELT operands can be simplified to one big
+ // BUILD_VECTOR.
// FIXME: Add support for SCALAR_TO_VECTOR as well.
EVT SVT = VT.getScalarType();
SmallVector<SDValue, 16> Elts;
@@ -6416,6 +6417,10 @@ static SDValue foldCONCAT_VECTORS(const SDLoc &DL, EVT VT,
Elts.append(OpVT.getVectorNumElements(), DAG.getUNDEF(SVT));
else if (Op.getOpcode() == ISD::BUILD_VECTOR)
Elts.append(Op->op_begin(), Op->op_end());
+ else if (Op.getOpcode() == ISD::INSERT_VECTOR_ELT &&
+ OpVT.getVectorNumElements() == 1 &&
+ isNullConstant(Op.getOperand(2)))
+ Elts.push_back(Op.getOperand(1));
else
return SDValue();
}
@@ -7235,7 +7240,7 @@ SDValue SelectionDAG::FoldConstantArithmetic(unsigned Opcode, const SDLoc &DL,
}
// Handle fma/fmad special cases.
- if (Opcode == ISD::FMA || Opcode == ISD::FMAD) {
+ if (Opcode == ISD::FMA || Opcode == ISD::FMAD || Opcode == ISD::FMULADD) {
assert(VT.isFloatingPoint() && "This operator only applies to FP types!");
assert(Ops[0].getValueType() == VT && Ops[1].getValueType() == VT &&
Ops[2].getValueType() == VT && "FMA types must match!");
@@ -7246,7 +7251,7 @@ SDValue SelectionDAG::FoldConstantArithmetic(unsigned Opcode, const SDLoc &DL,
APFloat V1 = C1->getValueAPF();
const APFloat &V2 = C2->getValueAPF();
const APFloat &V3 = C3->getValueAPF();
- if (Opcode == ISD::FMAD) {
+ if (Opcode == ISD::FMAD || Opcode == ISD::FMULADD) {
V1.multiply(V2, APFloat::rmNearestTiesToEven);
V1.add(V3, APFloat::rmNearestTiesToEven);
} else
@@ -8785,7 +8790,7 @@ static SDValue getMemcpyLoadsAndStores(
if (Value.getNode()) {
Store = DAG.getStore(
Chain, dl, Value,
- DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
+ DAG.getObjectPtrOffset(dl, Dst, TypeSize::getFixed(DstOff)),
DstPtrInfo.getWithOffset(DstOff), Alignment, MMOFlags, NewAAInfo);
OutChains.push_back(Store);
}
@@ -8801,7 +8806,7 @@ static SDValue getMemcpyLoadsAndStores(
assert(NVT.bitsGE(VT));
bool isDereferenceable =
- SrcPtrInfo.getWithOffset(SrcOff).isDereferenceable(VTSize, C, DL);
+ SrcPtrInfo.getWithOffset(SrcOff).isDereferenceable(VTSize, C, DL);
MachineMemOperand::Flags SrcMMOFlags = MMOFlags;
if (isDereferenceable)
SrcMMOFlags |= MachineMemOperand::MODereferenceable;
@@ -8810,14 +8815,14 @@ static SDValue getMemcpyLoadsAndStores(
Value = DAG.getExtLoad(
ISD::EXTLOAD, dl, NVT, Chain,
- DAG.getMemBasePlusOffset(Src, TypeSize::getFixed(SrcOff), dl),
+ DAG.getObjectPtrOffset(dl, Src, TypeSize::getFixed(SrcOff)),
SrcPtrInfo.getWithOffset(SrcOff), VT,
commonAlignment(*SrcAlign, SrcOff), SrcMMOFlags, NewAAInfo);
OutLoadChains.push_back(Value.getValue(1));
Store = DAG.getTruncStore(
Chain, dl, Value,
- DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
+ DAG.getObjectPtrOffset(dl, Dst, TypeSize::getFixed(DstOff)),
DstPtrInfo.getWithOffset(DstOff), VT, Alignment, MMOFlags, NewAAInfo);
OutStoreChains.push_back(Store);
}
@@ -8947,14 +8952,14 @@ static SDValue getMemmoveLoadsAndStores(SelectionDAG &DAG, const SDLoc &dl,
SDValue Value;
bool isDereferenceable =
- SrcPtrInfo.getWithOffset(SrcOff).isDereferenceable(VTSize, C, DL);
+ SrcPtrInfo.getWithOffset(SrcOff).isDereferenceable(VTSize, C, DL);
MachineMemOperand::Flags SrcMMOFlags = MMOFlags;
if (isDereferenceable)
SrcMMOFlags |= MachineMemOperand::MODereferenceable;
Value = DAG.getLoad(
VT, dl, Chain,
- DAG.getMemBasePlusOffset(Src, TypeSize::getFixed(SrcOff), dl),
+ DAG.getObjectPtrOffset(dl, Src, TypeSize::getFixed(SrcOff)),
SrcPtrInfo.getWithOffset(SrcOff), *SrcAlign, SrcMMOFlags, NewAAInfo);
LoadValues.push_back(Value);
LoadChains.push_back(Value.getValue(1));
@@ -8969,7 +8974,7 @@ static SDValue getMemmoveLoadsAndStores(SelectionDAG &DAG, const SDLoc &dl,
Store = DAG.getStore(
Chain, dl, LoadValues[i],
- DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
+ DAG.getObjectPtrOffset(dl, Dst, TypeSize::getFixed(DstOff)),
DstPtrInfo.getWithOffset(DstOff), Alignment, MMOFlags, NewAAInfo);
OutChains.push_back(Store);
DstOff += VTSize;
@@ -9101,7 +9106,7 @@ static SDValue getMemsetStores(SelectionDAG &DAG, const SDLoc &dl,
assert(Value.getValueType() == VT && "Value with wrong type.");
SDValue Store = DAG.getStore(
Chain, dl, Value,
- DAG.getMemBasePlusOffset(Dst, TypeSize::getFixed(DstOff), dl),
+ DAG.getObjectPtrOffset(dl, Dst, TypeSize::getFixed(DstOff)),
DstPtrInfo.getWithOffset(DstOff), Alignment,
isVol ? MachineMemOperand::MOVolatile : MachineMemOperand::MONone,
NewAAInfo);
@@ -11848,25 +11853,38 @@ SDValue SelectionDAG::getTargetInsertSubreg(int SRIdx, const SDLoc &DL, EVT VT,
/// getNodeIfExists - Get the specified node if it's already available, or
/// else return NULL.
SDNode *SelectionDAG::getNodeIfExists(unsigned Opcode, SDVTList VTList,
- ArrayRef<SDValue> Ops) {
+ ArrayRef<SDValue> Ops,
+ bool AllowCommute) {
SDNodeFlags Flags;
if (Inserter)
Flags = Inserter->getFlags();
- return getNodeIfExists(Opcode, VTList, Ops, Flags);
+ return getNodeIfExists(Opcode, VTList, Ops, Flags, AllowCommute);
}
SDNode *SelectionDAG::getNodeIfExists(unsigned Opcode, SDVTList VTList,
ArrayRef<SDValue> Ops,
- const SDNodeFlags Flags) {
- if (VTList.VTs[VTList.NumVTs - 1] != MVT::Glue) {
+ const SDNodeFlags Flags,
+ bool AllowCommute) {
+ if (VTList.VTs[VTList.NumVTs - 1] == MVT::Glue)
+ return nullptr;
+
+ auto Lookup = [&](ArrayRef<SDValue> LookupOps) -> SDNode * {
FoldingSetNodeID ID;
- AddNodeIDNode(ID, Opcode, VTList, Ops);
+ AddNodeIDNode(ID, Opcode, VTList, LookupOps);
void *IP = nullptr;
- if (SDNode *E = FindNodeOrInsertPos(ID, SDLoc(), IP)) {
+ if (SDNode *E = FindNodeOrInsertPos(ID, IP)) {
E->intersectFlagsWith(Flags);
return E;
}
- }
+ return nullptr;
+ };
+
+ if (SDNode *Existing = Lookup(Ops))
+ return Existing;
+
+ if (AllowCommute && TLI->isCommutativeBinOp(Opcode))
+ return Lookup({Ops[1], Ops[0]});
+
return nullptr;
}
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index c21890a..cb0038c 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -3971,8 +3971,14 @@ void SelectionDAGBuilder::visitSIToFP(const User &I) {
}
void SelectionDAGBuilder::visitPtrToAddr(const User &I) {
- // FIXME: this is not correct for pointers with addr width != pointer width
- visitPtrToInt(I);
+ SDValue N = getValue(I.getOperand(0));
+ // By definition the type of the ptrtoaddr must be equal to the address type.
+ const auto &TLI = DAG.getTargetLoweringInfo();
+ EVT AddrVT = TLI.getValueType(DAG.getDataLayout(), I.getType());
+ // The address width must be smaller or equal to the pointer representation
+ // width, so we lower ptrtoaddr as a truncate (possibly folded to a no-op).
+ N = DAG.getNode(ISD::TRUNCATE, getCurSDLoc(), AddrVT, N);
+ setValue(&I, N);
}
void SelectionDAGBuilder::visitPtrToInt(const User &I) {
@@ -6996,6 +7002,13 @@ void SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I,
getValue(I.getArgOperand(0)),
getValue(I.getArgOperand(1)),
getValue(I.getArgOperand(2)), Flags));
+ } else if (TLI.isOperationLegalOrCustom(ISD::FMULADD, VT)) {
+ // TODO: Support splitting the vector.
+ setValue(&I, DAG.getNode(ISD::FMULADD, sdl,
+ getValue(I.getArgOperand(0)).getValueType(),
+ getValue(I.getArgOperand(0)),
+ getValue(I.getArgOperand(1)),
+ getValue(I.getArgOperand(2)), Flags));
} else {
// TODO: Intrinsic calls should have fast-math-flags.
SDValue Mul = DAG.getNode(
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
index fcfbfe6..39cbfad 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGDumper.cpp
@@ -310,6 +310,7 @@ std::string SDNode::getOperationName(const SelectionDAG *G) const {
case ISD::FMA: return "fma";
case ISD::STRICT_FMA: return "strict_fma";
case ISD::FMAD: return "fmad";
+ case ISD::FMULADD: return "fmuladd";
case ISD::FREM: return "frem";
case ISD::STRICT_FREM: return "strict_frem";
case ISD::FCOPYSIGN: return "fcopysign";
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
index 175753f..6c11c5b 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGISel.cpp
@@ -234,6 +234,19 @@ static bool dontUseFastISelFor(const Function &Fn) {
});
}
+static bool maintainPGOProfile(const TargetMachine &TM,
+ CodeGenOptLevel OptLevel) {
+ if (OptLevel != CodeGenOptLevel::None)
+ return true;
+ if (TM.getPGOOption()) {
+ const PGOOptions &Options = *TM.getPGOOption();
+ return Options.Action == PGOOptions::PGOAction::IRUse ||
+ Options.Action == PGOOptions::PGOAction::SampleUse ||
+ Options.CSAction == PGOOptions::CSPGOAction::CSIRUse;
+ }
+ return false;
+}
+
namespace llvm {
//===--------------------------------------------------------------------===//
@@ -395,6 +408,7 @@ SelectionDAGISel::~SelectionDAGISel() { delete CurDAG; }
void SelectionDAGISelLegacy::getAnalysisUsage(AnalysisUsage &AU) const {
CodeGenOptLevel OptLevel = Selector->OptLevel;
+ bool RegisterPGOPasses = maintainPGOProfile(Selector->TM, Selector->OptLevel);
if (OptLevel != CodeGenOptLevel::None)
AU.addRequired<AAResultsWrapperPass>();
AU.addRequired<GCModuleInfo>();
@@ -403,15 +417,15 @@ void SelectionDAGISelLegacy::getAnalysisUsage(AnalysisUsage &AU) const {
AU.addRequired<TargetLibraryInfoWrapperPass>();
AU.addRequired<TargetTransformInfoWrapperPass>();
AU.addRequired<AssumptionCacheTracker>();
- if (UseMBPI && OptLevel != CodeGenOptLevel::None)
- AU.addRequired<BranchProbabilityInfoWrapperPass>();
+ if (UseMBPI && RegisterPGOPasses)
+ AU.addRequired<BranchProbabilityInfoWrapperPass>();
AU.addRequired<ProfileSummaryInfoWrapperPass>();
// AssignmentTrackingAnalysis only runs if assignment tracking is enabled for
// the module.
AU.addRequired<AssignmentTrackingAnalysis>();
AU.addPreserved<AssignmentTrackingAnalysis>();
- if (OptLevel != CodeGenOptLevel::None)
- LazyBlockFrequencyInfoPass::getLazyBFIAnalysisUsage(AU);
+ if (RegisterPGOPasses)
+ LazyBlockFrequencyInfoPass::getLazyBFIAnalysisUsage(AU);
MachineFunctionPass::getAnalysisUsage(AU);
}
@@ -464,6 +478,7 @@ void SelectionDAGISel::initializeAnalysisResults(
(void)MatchFilterFuncName;
#endif
+ bool RegisterPGOPasses = maintainPGOProfile(TM, OptLevel);
TII = MF->getSubtarget().getInstrInfo();
TLI = MF->getSubtarget().getTargetLowering();
RegInfo = &MF->getRegInfo();
@@ -474,7 +489,7 @@ void SelectionDAGISel::initializeAnalysisResults(
auto *PSI = MAMP.getCachedResult<ProfileSummaryAnalysis>(*Fn.getParent());
BlockFrequencyInfo *BFI = nullptr;
FAM.getResult<BlockFrequencyAnalysis>(Fn);
- if (PSI && PSI->hasProfileSummary() && OptLevel != CodeGenOptLevel::None)
+ if (PSI && PSI->hasProfileSummary() && RegisterPGOPasses)
BFI = &FAM.getResult<BlockFrequencyAnalysis>(Fn);
FunctionVarLocs const *FnVarLocs = nullptr;
@@ -492,7 +507,7 @@ void SelectionDAGISel::initializeAnalysisResults(
// into account). That's unfortunate but OK because it just means we won't
// ask for passes that have been required anyway.
- if (UseMBPI && OptLevel != CodeGenOptLevel::None)
+ if (UseMBPI && RegisterPGOPasses)
FuncInfo->BPI = &FAM.getResult<BranchProbabilityAnalysis>(Fn);
else
FuncInfo->BPI = nullptr;
@@ -518,6 +533,7 @@ void SelectionDAGISel::initializeAnalysisResults(MachineFunctionPass &MFP) {
(void)MatchFilterFuncName;
#endif
+ bool RegisterPGOPasses = maintainPGOProfile(TM, OptLevel);
TII = MF->getSubtarget().getInstrInfo();
TLI = MF->getSubtarget().getTargetLowering();
RegInfo = &MF->getRegInfo();
@@ -528,7 +544,7 @@ void SelectionDAGISel::initializeAnalysisResults(MachineFunctionPass &MFP) {
AC = &MFP.getAnalysis<AssumptionCacheTracker>().getAssumptionCache(Fn);
auto *PSI = &MFP.getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI();
BlockFrequencyInfo *BFI = nullptr;
- if (PSI && PSI->hasProfileSummary() && OptLevel != CodeGenOptLevel::None)
+ if (PSI && PSI->hasProfileSummary() && RegisterPGOPasses)
BFI = &MFP.getAnalysis<LazyBlockFrequencyInfoPass>().getBFI();
FunctionVarLocs const *FnVarLocs = nullptr;
@@ -549,7 +565,7 @@ void SelectionDAGISel::initializeAnalysisResults(MachineFunctionPass &MFP) {
// into account). That's unfortunate but OK because it just means we won't
// ask for passes that have been required anyway.
- if (UseMBPI && OptLevel != CodeGenOptLevel::None)
+ if (UseMBPI && RegisterPGOPasses)
FuncInfo->BPI =
&MFP.getAnalysis<BranchProbabilityInfoWrapperPass>().getBPI();
else
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index cc503d3..920dff9 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -7676,6 +7676,7 @@ SDValue TargetLowering::getNegatedExpression(SDValue Op, SelectionDAG &DAG,
break;
}
case ISD::FMA:
+ case ISD::FMULADD:
case ISD::FMAD: {
if (!Flags.hasNoSignedZeros())
break;
diff --git a/llvm/lib/CodeGen/StackFrameLayoutAnalysisPass.cpp b/llvm/lib/CodeGen/StackFrameLayoutAnalysisPass.cpp
index 64e5cd5..95a9c3f 100644
--- a/llvm/lib/CodeGen/StackFrameLayoutAnalysisPass.cpp
+++ b/llvm/lib/CodeGen/StackFrameLayoutAnalysisPass.cpp
@@ -306,10 +306,7 @@ char &llvm::StackFrameLayoutAnalysisPassID = StackFrameLayoutAnalysisLegacy::ID;
INITIALIZE_PASS(StackFrameLayoutAnalysisLegacy, "stack-frame-layout",
"Stack Frame Layout", false, false)
-namespace llvm {
/// Returns a newly-created StackFrameLayout pass.
-MachineFunctionPass *createStackFrameLayoutAnalysisPass() {
+MachineFunctionPass *llvm::createStackFrameLayoutAnalysisPass() {
return new StackFrameLayoutAnalysisLegacy();
}
-
-} // namespace llvm
diff --git a/llvm/lib/CodeGen/StaticDataAnnotator.cpp b/llvm/lib/CodeGen/StaticDataAnnotator.cpp
index 53a9ab4..eac20120 100644
--- a/llvm/lib/CodeGen/StaticDataAnnotator.cpp
+++ b/llvm/lib/CodeGen/StaticDataAnnotator.cpp
@@ -75,22 +75,11 @@ bool StaticDataAnnotator::runOnModule(Module &M) {
bool Changed = false;
for (auto &GV : M.globals()) {
- if (GV.isDeclarationForLinker())
+ if (!llvm::memprof::IsAnnotationOK(GV))
continue;
- // The implementation below assumes prior passes don't set section prefixes,
- // and specifically do 'assign' rather than 'update'. So report error if a
- // section prefix is already set.
- if (auto maybeSectionPrefix = GV.getSectionPrefix();
- maybeSectionPrefix && !maybeSectionPrefix->empty())
- llvm::report_fatal_error("Global variable " + GV.getName() +
- " already has a section prefix " +
- *maybeSectionPrefix);
-
StringRef SectionPrefix = SDPI->getConstantSectionPrefix(&GV, PSI);
- if (SectionPrefix.empty())
- continue;
-
+ // setSectionPrefix returns true if the section prefix is updated.
Changed |= GV.setSectionPrefix(SectionPrefix);
}
diff --git a/llvm/lib/CodeGen/StaticDataSplitter.cpp b/llvm/lib/CodeGen/StaticDataSplitter.cpp
index e22dc25..1593a40 100644
--- a/llvm/lib/CodeGen/StaticDataSplitter.cpp
+++ b/llvm/lib/CodeGen/StaticDataSplitter.cpp
@@ -130,10 +130,8 @@ StaticDataSplitter::getConstant(const MachineOperand &Op,
if (Op.isGlobal()) {
// Find global variables with local linkage.
const GlobalVariable *GV = getLocalLinkageGlobalVariable(Op.getGlobal());
- // Skip 'llvm.'-prefixed global variables conservatively because they are
- // often handled specially, and skip those not in static data
- // sections.
- if (!GV || GV->getName().starts_with("llvm.") ||
+ // Skip those not eligible for annotation or not in static data sections.
+ if (!GV || !llvm::memprof::IsAnnotationOK(*GV) ||
!inStaticDataSection(*GV, TM))
return nullptr;
return GV;
diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index c23281a..060b1dd 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -815,7 +815,8 @@ void TargetLoweringBase::initActions() {
ISD::FTAN, ISD::FACOS,
ISD::FASIN, ISD::FATAN,
ISD::FCOSH, ISD::FSINH,
- ISD::FTANH, ISD::FATAN2},
+ ISD::FTANH, ISD::FATAN2,
+ ISD::FMULADD},
VT, Expand);
// Overflow operations default to expand
diff --git a/llvm/lib/CodeGen/TargetRegisterInfo.cpp b/llvm/lib/CodeGen/TargetRegisterInfo.cpp
index c9e4618..971f822 100644
--- a/llvm/lib/CodeGen/TargetRegisterInfo.cpp
+++ b/llvm/lib/CodeGen/TargetRegisterInfo.cpp
@@ -102,10 +102,8 @@ bool TargetRegisterInfo::checkAllSuperRegsMarked(const BitVector &RegisterSet,
return true;
}
-namespace llvm {
-
-Printable printReg(Register Reg, const TargetRegisterInfo *TRI,
- unsigned SubIdx, const MachineRegisterInfo *MRI) {
+Printable llvm::printReg(Register Reg, const TargetRegisterInfo *TRI,
+ unsigned SubIdx, const MachineRegisterInfo *MRI) {
return Printable([Reg, TRI, SubIdx, MRI](raw_ostream &OS) {
if (!Reg)
OS << "$noreg";
@@ -135,7 +133,7 @@ Printable printReg(Register Reg, const TargetRegisterInfo *TRI,
});
}
-Printable printRegUnit(unsigned Unit, const TargetRegisterInfo *TRI) {
+Printable llvm::printRegUnit(unsigned Unit, const TargetRegisterInfo *TRI) {
return Printable([Unit, TRI](raw_ostream &OS) {
// Generic printout when TRI is missing.
if (!TRI) {
@@ -158,7 +156,7 @@ Printable printRegUnit(unsigned Unit, const TargetRegisterInfo *TRI) {
});
}
-Printable printVRegOrUnit(unsigned Unit, const TargetRegisterInfo *TRI) {
+Printable llvm::printVRegOrUnit(unsigned Unit, const TargetRegisterInfo *TRI) {
return Printable([Unit, TRI](raw_ostream &OS) {
if (Register::isVirtualRegister(Unit)) {
OS << '%' << Register(Unit).virtRegIndex();
@@ -168,8 +166,9 @@ Printable printVRegOrUnit(unsigned Unit, const TargetRegisterInfo *TRI) {
});
}
-Printable printRegClassOrBank(Register Reg, const MachineRegisterInfo &RegInfo,
- const TargetRegisterInfo *TRI) {
+Printable llvm::printRegClassOrBank(Register Reg,
+ const MachineRegisterInfo &RegInfo,
+ const TargetRegisterInfo *TRI) {
return Printable([Reg, &RegInfo, TRI](raw_ostream &OS) {
if (RegInfo.getRegClassOrNull(Reg))
OS << StringRef(TRI->getRegClassName(RegInfo.getRegClass(Reg))).lower();
@@ -183,8 +182,6 @@ Printable printRegClassOrBank(Register Reg, const MachineRegisterInfo &RegInfo,
});
}
-} // end namespace llvm
-
/// getAllocatableClass - Return the maximal subclass of the given register
/// class that is alloctable, or NULL.
const TargetRegisterClass *