diff options
author | Dimitry Andric <dim@FreeBSD.org> | 2016-11-25 19:07:40 +0000 |
---|---|---|
committer | Dimitry Andric <dim@FreeBSD.org> | 2016-11-25 19:07:40 +0000 |
commit | 17c7957f023f02fc2c88f51f8908c19b52609275 (patch) | |
tree | c654618ff2d38e26916b49614d89fe01f4a4818d /lib/CodeGen | |
parent | c477790a57f44875b9de2043f2eb47dff2d20133 (diff) |
Diffstat (limited to 'lib/CodeGen')
-rw-r--r-- | lib/CodeGen/CGExpr.cpp | 17 | ||||
-rw-r--r-- | lib/CodeGen/CGStmt.cpp | 33 | ||||
-rw-r--r-- | lib/CodeGen/CGStmtOpenMP.cpp | 61 | ||||
-rw-r--r-- | lib/CodeGen/CodeGenFunction.cpp | 45 | ||||
-rw-r--r-- | lib/CodeGen/CodeGenFunction.h | 33 |
5 files changed, 166 insertions, 23 deletions
diff --git a/lib/CodeGen/CGExpr.cpp b/lib/CodeGen/CGExpr.cpp index 3e1ae3604f94..5f3b290d8eb1 100644 --- a/lib/CodeGen/CGExpr.cpp +++ b/lib/CodeGen/CGExpr.cpp @@ -2105,12 +2105,11 @@ LValue CodeGenFunction::EmitDeclRefLValue(const DeclRefExpr *E) { if (auto *FD = LambdaCaptureFields.lookup(VD)) return EmitCapturedFieldLValue(*this, FD, CXXABIThisValue); else if (CapturedStmtInfo) { - auto it = LocalDeclMap.find(VD); - if (it != LocalDeclMap.end()) { - if (auto RefTy = VD->getType()->getAs<ReferenceType>()) { - return EmitLoadOfReferenceLValue(it->second, RefTy); - } - return MakeAddrLValue(it->second, T); + auto I = LocalDeclMap.find(VD); + if (I != LocalDeclMap.end()) { + if (auto RefTy = VD->getType()->getAs<ReferenceType>()) + return EmitLoadOfReferenceLValue(I->second, RefTy); + return MakeAddrLValue(I->second, T); } LValue CapLVal = EmitCapturedFieldLValue(*this, CapturedStmtInfo->lookup(VD), @@ -2249,13 +2248,15 @@ LValue CodeGenFunction::EmitUnaryOpLValue(const UnaryOperator *E) { return LV; } - assert(E->getSubExpr()->getType()->isAnyComplexType()); + QualType T = ExprTy->castAs<ComplexType>()->getElementType(); Address Component = (E->getOpcode() == UO_Real ? emitAddrOfRealComponent(LV.getAddress(), LV.getType()) : emitAddrOfImagComponent(LV.getAddress(), LV.getType())); - return MakeAddrLValue(Component, ExprTy, LV.getAlignmentSource()); + LValue ElemLV = MakeAddrLValue(Component, T, LV.getAlignmentSource()); + ElemLV.getQuals().addQualifiers(LV.getQuals()); + return ElemLV; } case UO_PreInc: case UO_PreDec: { diff --git a/lib/CodeGen/CGStmt.cpp b/lib/CodeGen/CGStmt.cpp index 77879021f9af..d815863e929d 100644 --- a/lib/CodeGen/CGStmt.cpp +++ b/lib/CodeGen/CGStmt.cpp @@ -1323,6 +1323,10 @@ static CSFC_Result CollectStatementsForCase(const Stmt *S, // Handle this as two cases: we might be looking for the SwitchCase (if so // the skipped statements must be skippable) or we might already have it. CompoundStmt::const_body_iterator I = CS->body_begin(), E = CS->body_end(); + bool StartedInLiveCode = FoundCase; + unsigned StartSize = ResultStmts.size(); + + // If we've not found the case yet, scan through looking for it. if (Case) { // Keep track of whether we see a skipped declaration. The code could be // using the declaration even if it is skipped, so we can't optimize out @@ -1332,7 +1336,7 @@ static CSFC_Result CollectStatementsForCase(const Stmt *S, // If we're looking for the case, just see if we can skip each of the // substatements. for (; Case && I != E; ++I) { - HadSkippedDecl |= isa<DeclStmt>(*I); + HadSkippedDecl |= CodeGenFunction::mightAddDeclToScope(*I); switch (CollectStatementsForCase(*I, Case, FoundCase, ResultStmts)) { case CSFC_Failure: return CSFC_Failure; @@ -1368,11 +1372,19 @@ static CSFC_Result CollectStatementsForCase(const Stmt *S, break; } } + + if (!FoundCase) + return CSFC_Success; + + assert(!HadSkippedDecl && "fallthrough after skipping decl"); } // If we have statements in our range, then we know that the statements are // live and need to be added to the set of statements we're tracking. + bool AnyDecls = false; for (; I != E; ++I) { + AnyDecls |= CodeGenFunction::mightAddDeclToScope(*I); + switch (CollectStatementsForCase(*I, nullptr, FoundCase, ResultStmts)) { case CSFC_Failure: return CSFC_Failure; case CSFC_FallThrough: @@ -1390,7 +1402,24 @@ static CSFC_Result CollectStatementsForCase(const Stmt *S, } } - return Case ? CSFC_Success : CSFC_FallThrough; + // If we're about to fall out of a scope without hitting a 'break;', we + // can't perform the optimization if there were any decls in that scope + // (we'd lose their end-of-lifetime). + if (AnyDecls) { + // If the entire compound statement was live, there's one more thing we + // can try before giving up: emit the whole thing as a single statement. + // We can do that unless the statement contains a 'break;'. + // FIXME: Such a break must be at the end of a construct within this one. + // We could emit this by just ignoring the BreakStmts entirely. + if (StartedInLiveCode && !CodeGenFunction::containsBreak(S)) { + ResultStmts.resize(StartSize); + ResultStmts.push_back(S); + } else { + return CSFC_Failure; + } + } + + return CSFC_FallThrough; } // Okay, this is some other statement that we don't handle explicitly, like a diff --git a/lib/CodeGen/CGStmtOpenMP.cpp b/lib/CodeGen/CGStmtOpenMP.cpp index 26f780fdd7de..8937685fdc7b 100644 --- a/lib/CodeGen/CGStmtOpenMP.cpp +++ b/lib/CodeGen/CGStmtOpenMP.cpp @@ -232,8 +232,15 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) { assert(I->capturesVariableArrayType()); II = &getContext().Idents.get("vla"); } - if (ArgType->isVariablyModifiedType()) - ArgType = getContext().getVariableArrayDecayedType(ArgType); + if (ArgType->isVariablyModifiedType()) { + bool IsReference = ArgType->isLValueReferenceType(); + ArgType = + getContext().getCanonicalParamType(ArgType.getNonReferenceType()); + if (IsReference && !ArgType->isPointerType()) { + ArgType = getContext().getLValueReferenceType( + ArgType, /*SpelledAsLValue=*/false); + } + } Args.push_back(ImplicitParamDecl::Create(getContext(), nullptr, FD->getLocation(), II, ArgType)); ++I; @@ -287,8 +294,14 @@ CodeGenFunction::GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S) { QualType VarTy = Var->getType(); Address ArgAddr = ArgLVal.getAddress(); if (!VarTy->isReferenceType()) { - ArgAddr = EmitLoadOfReference( - ArgAddr, ArgLVal.getType()->castAs<ReferenceType>()); + if (ArgLVal.getType()->isLValueReferenceType()) { + ArgAddr = EmitLoadOfReference( + ArgAddr, ArgLVal.getType()->castAs<ReferenceType>()); + } else if (!VarTy->isVariablyModifiedType() || !VarTy->isPointerType()) { + assert(ArgLVal.getType()->isPointerType()); + ArgAddr = EmitLoadOfPointer( + ArgAddr, ArgLVal.getType()->castAs<PointerType>()); + } } setAddrOfLocalVar( Var, Address(ArgAddr.getPointer(), getContext().getDeclAlign(Var))); @@ -1754,9 +1767,17 @@ void CodeGenFunction::EmitOMPOuterLoop(bool DynamicOrOrdered, bool IsMonotonic, EmitBlock(LoopExit.getBlock()); // Tell the runtime we are done. - if (!DynamicOrOrdered) - RT.emitForStaticFinish(*this, S.getLocEnd()); + SourceLocation ELoc = S.getLocEnd(); + auto &&CodeGen = [DynamicOrOrdered, ELoc](CodeGenFunction &CGF) { + if (!DynamicOrOrdered) + CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc); + }; + CodeGen(*this); + OpenMPDirectiveKind DKind = S.getDirectiveKind(); + if (DKind == OMPD_for || DKind == OMPD_parallel_for || + DKind == OMPD_distribute_parallel_for) + OMPCancelStack.back().CodeGen = CodeGen; } void CodeGenFunction::EmitOMPForOuterLoop( @@ -1868,6 +1889,7 @@ void CodeGenFunction::EmitOMPDistributeOuterLoop( void CodeGenFunction::EmitOMPDistributeParallelForDirective( const OMPDistributeParallelForDirective &S) { OMPLexicalScope Scope(*this, S, /*AsInlined=*/true); + OMPCancelStackRAII CancelRegion(*this); CGM.getOpenMPRuntime().emitInlinedDirective( *this, OMPD_distribute_parallel_for, [&S](CodeGenFunction &CGF, PrePostActionTy &) { @@ -2060,7 +2082,15 @@ bool CodeGenFunction::EmitOMPWorksharingLoop(const OMPLoopDirective &S) { [](CodeGenFunction &) {}); EmitBlock(LoopExit.getBlock()); // Tell the runtime we are done. - RT.emitForStaticFinish(*this, S.getLocStart()); + SourceLocation ELoc = S.getLocEnd(); + auto &&CodeGen = [ELoc](CodeGenFunction &CGF) { + CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc); + }; + CodeGen(*this); + OpenMPDirectiveKind DKind = S.getDirectiveKind(); + if (DKind == OMPD_for || DKind == OMPD_parallel_for || + DKind == OMPD_distribute_parallel_for) + OMPCancelStack.back().CodeGen = CodeGen; } else { const bool IsMonotonic = Ordered || ScheduleKind.Schedule == OMPC_SCHEDULE_static || @@ -2114,6 +2144,7 @@ void CodeGenFunction::EmitOMPForDirective(const OMPForDirective &S) { }; { OMPLexicalScope Scope(*this, S, /*AsInlined=*/true); + OMPCancelStackRAII CancelRegion(*this); CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_for, CodeGen, S.hasCancel()); } @@ -2156,6 +2187,7 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) { bool HasLastprivates = false; auto &&CodeGen = [&S, Stmt, CS, &HasLastprivates](CodeGenFunction &CGF, PrePostActionTy &) { + OMPCancelStackRAII CancelRegion(CGF); auto &C = CGF.CGM.getContext(); auto KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1); // Emit helper vars inits. @@ -2250,7 +2282,12 @@ void CodeGenFunction::EmitSections(const OMPExecutableDirective &S) { CGF.EmitOMPInnerLoop(S, /*RequiresCleanup=*/false, &Cond, &Inc, BodyGen, [](CodeGenFunction &) {}); // Tell the runtime we are done. - CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocStart()); + SourceLocation ELoc = S.getLocEnd(); + auto &&FinalCodeGen = [ELoc](CodeGenFunction &CGF) { + CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc); + }; + FinalCodeGen(CGF); + CGF.OMPCancelStack.back().CodeGen = FinalCodeGen; CGF.EmitOMPReductionClauseFinal(S); // Emit post-update of the reduction variables if IsLastIter != 0. emitPostUpdateForReductionClause( @@ -2375,6 +2412,7 @@ void CodeGenFunction::EmitOMPParallelForDirective( // Emit directive as a combined directive that consists of two implicit // directives: 'parallel' with 'for' directive. auto &&CodeGen = [&S](CodeGenFunction &CGF, PrePostActionTy &) { + OMPCancelStackRAII CancelRegion(CGF); CGF.EmitOMPWorksharingLoop(S); }; emitCommonOMPParallelDirective(*this, S, OMPD_for, CodeGen); @@ -3377,8 +3415,11 @@ CodeGenFunction::getOMPCancelDestination(OpenMPDirectiveKind Kind) { if (Kind == OMPD_parallel || Kind == OMPD_task) return ReturnBlock; assert(Kind == OMPD_for || Kind == OMPD_section || Kind == OMPD_sections || - Kind == OMPD_parallel_sections || Kind == OMPD_parallel_for); - return BreakContinueStack.back().BreakBlock; + Kind == OMPD_parallel_sections || Kind == OMPD_parallel_for || + Kind == OMPD_distribute_parallel_for); + if (!OMPCancelStack.back().ExitBlock.isValid()) + OMPCancelStack.back().ExitBlock = getJumpDestInCurrentScope("cancel.exit"); + return OMPCancelStack.back().ExitBlock; } // Generate the instructions for '#pragma omp target data' directive. diff --git a/lib/CodeGen/CodeGenFunction.cpp b/lib/CodeGen/CodeGenFunction.cpp index 183ee12ea232..11e4ad9ecefa 100644 --- a/lib/CodeGen/CodeGenFunction.cpp +++ b/lib/CodeGen/CodeGenFunction.cpp @@ -25,6 +25,7 @@ #include "clang/AST/Decl.h" #include "clang/AST/DeclCXX.h" #include "clang/AST/StmtCXX.h" +#include "clang/AST/StmtObjC.h" #include "clang/Basic/Builtins.h" #include "clang/Basic/TargetInfo.h" #include "clang/CodeGen/CGFunctionInfo.h" @@ -436,6 +437,23 @@ void CodeGenFunction::EmitMCountInstrumentation() { EmitNounwindRuntimeCall(MCountFn); } +// Returns the address space id that should be produced to the +// kernel_arg_addr_space metadata. This is always fixed to the ids +// as specified in the SPIR 2.0 specification in order to differentiate +// for example in clGetKernelArgInfo() implementation between the address +// spaces with targets without unique mapping to the OpenCL address spaces +// (basically all single AS CPUs). +static unsigned ArgInfoAddressSpace(unsigned LangAS) { + switch (LangAS) { + case LangAS::opencl_global: return 1; + case LangAS::opencl_constant: return 2; + case LangAS::opencl_local: return 3; + case LangAS::opencl_generic: return 4; // Not in SPIR 2.0 specs. + default: + return 0; // Assume private. + } +} + // OpenCL v1.2 s5.6.4.6 allows the compiler to store kernel argument // information in the program executable. The argument information stored // includes the argument name, its type, the address and access qualifiers used. @@ -476,7 +494,7 @@ static void GenOpenCLArgMetadata(const FunctionDecl *FD, llvm::Function *Fn, // Get address qualifier. addressQuals.push_back(llvm::ConstantAsMetadata::get(Builder.getInt32( - ASTCtx.getTargetAddressSpace(pointeeTy.getAddressSpace())))); + ArgInfoAddressSpace(pointeeTy.getAddressSpace())))); // Get argument type name. std::string typeName = @@ -513,8 +531,7 @@ static void GenOpenCLArgMetadata(const FunctionDecl *FD, llvm::Function *Fn, uint32_t AddrSpc = 0; bool isPipe = ty->isPipeType(); if (ty->isImageType() || isPipe) - AddrSpc = - CGM.getContext().getTargetAddressSpace(LangAS::opencl_global); + AddrSpc = ArgInfoAddressSpace(LangAS::opencl_global); addressQuals.push_back( llvm::ConstantAsMetadata::get(Builder.getInt32(AddrSpc))); @@ -1143,6 +1160,28 @@ bool CodeGenFunction::containsBreak(const Stmt *S) { return false; } +bool CodeGenFunction::mightAddDeclToScope(const Stmt *S) { + if (!S) return false; + + // Some statement kinds add a scope and thus never add a decl to the current + // scope. Note, this list is longer than the list of statements that might + // have an unscoped decl nested within them, but this way is conservatively + // correct even if more statement kinds are added. + if (isa<IfStmt>(S) || isa<SwitchStmt>(S) || isa<WhileStmt>(S) || + isa<DoStmt>(S) || isa<ForStmt>(S) || isa<CompoundStmt>(S) || + isa<CXXForRangeStmt>(S) || isa<CXXTryStmt>(S) || + isa<ObjCForCollectionStmt>(S) || isa<ObjCAtTryStmt>(S)) + return false; + + if (isa<DeclStmt>(S)) + return true; + + for (const Stmt *SubStmt : S->children()) + if (mightAddDeclToScope(SubStmt)) + return true; + + return false; +} /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it diff --git a/lib/CodeGen/CodeGenFunction.h b/lib/CodeGen/CodeGenFunction.h index 45068fa55444..f61ba69e3a0c 100644 --- a/lib/CodeGen/CodeGenFunction.h +++ b/lib/CodeGen/CodeGenFunction.h @@ -965,6 +965,35 @@ private: }; SmallVector<BreakContinue, 8> BreakContinueStack; + /// Data for exit block for proper support of OpenMP cancellation constructs. + struct OMPCancel { + JumpDest ExitBlock; + llvm::function_ref<void(CodeGenFunction &CGF)> CodeGen; + OMPCancel() : CodeGen([](CodeGenFunction &CGF) {}) {} + }; + SmallVector<OMPCancel, 8> OMPCancelStack; + + /// Controls insertion of cancellation exit blocks in worksharing constructs. + class OMPCancelStackRAII { + CodeGenFunction &CGF; + + public: + OMPCancelStackRAII(CodeGenFunction &CGF) : CGF(CGF) { + CGF.OMPCancelStack.push_back({}); + } + ~OMPCancelStackRAII() { + if (CGF.HaveInsertPoint() && + CGF.OMPCancelStack.back().ExitBlock.isValid()) { + auto CJD = CGF.getJumpDestInCurrentScope("cancel.cont"); + CGF.EmitBranchThroughCleanup(CJD); + CGF.EmitBlock(CGF.OMPCancelStack.back().ExitBlock.getBlock()); + CGF.OMPCancelStack.back().CodeGen(CGF); + CGF.EmitBranchThroughCleanup(CJD); + CGF.EmitBlock(CJD.getBlock()); + } + } + }; + CodeGenPGO PGO; /// Calculate branch weights appropriate for PGO data @@ -3163,6 +3192,10 @@ public: /// If the statement (recursively) contains a switch or loop with a break /// inside of it, this is fine. static bool containsBreak(const Stmt *S); + + /// Determine if the given statement might introduce a declaration into the + /// current scope, by being a (possibly-labelled) DeclStmt. + static bool mightAddDeclToScope(const Stmt *S); /// ConstantFoldsToSimpleInteger - If the specified expression does not fold /// to a constant, or if it does but contains a label, return false. If it |