Date: Fri, 2 Dec 2016 19:36:28 +0000 (UTC) From: Dimitry Andric <dim@FreeBSD.org> To: src-committers@freebsd.org, svn-src-projects@freebsd.org Subject: svn commit: r309437 - in projects/clang391-import/contrib/llvm: include/llvm/Support lib/CodeGen/SelectionDAG lib/Target/AMDGPU lib/Transforms/InstCombine lib/Transforms/Utils tools/clang/include/c... Message-ID: <201612021936.uB2JaSNP083228@repo.freebsd.org>
next in thread | raw e-mail | index | archive | help
Author: dim Date: Fri Dec 2 19:36:28 2016 New Revision: 309437 URL: https://svnweb.freebsd.org/changeset/base/309437 Log: Update llvm, clang, lld and lldb to release_39 branch r288513. Added: projects/clang391-import/contrib/llvm/tools/clang/lib/Headers/msa.h - copied unchanged from r309436, vendor/clang/dist/lib/Headers/msa.h Modified: projects/clang391-import/contrib/llvm/include/llvm/Support/Threading.h projects/clang391-import/contrib/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstructions.td projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp projects/clang391-import/contrib/llvm/lib/Transforms/Utils/SimplifyCFG.cpp projects/clang391-import/contrib/llvm/tools/clang/include/clang/Basic/DiagnosticSemaKinds.td projects/clang391-import/contrib/llvm/tools/clang/include/clang/Sema/Sema.h projects/clang391-import/contrib/llvm/tools/clang/lib/Basic/Targets.cpp projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CGStmtOpenMP.cpp projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CodeGenFunction.h projects/clang391-import/contrib/llvm/tools/clang/lib/Sema/SemaChecking.cpp projects/clang391-import/contrib/llvm/tools/clang/lib/Sema/SemaExprCXX.cpp projects/clang391-import/contrib/llvm/tools/lld/COFF/CMakeLists.txt Directory Properties: projects/clang391-import/contrib/llvm/ (props changed) projects/clang391-import/contrib/llvm/tools/clang/ (props changed) projects/clang391-import/contrib/llvm/tools/lld/ (props changed) projects/clang391-import/contrib/llvm/tools/lldb/ (props changed) Modified: projects/clang391-import/contrib/llvm/include/llvm/Support/Threading.h ============================================================================== --- projects/clang391-import/contrib/llvm/include/llvm/Support/Threading.h Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/include/llvm/Support/Threading.h Fri Dec 2 19:36:28 2016 (r309437) @@ -20,11 +20,11 @@ #include <ciso646> // So we can check the C++ standard lib macros. #include <functional> -// We use std::call_once on all Unix platforms except for NetBSD with -// libstdc++. That platform has a bug they are working to fix, and they'll -// remove the NetBSD checks once fixed. -#if defined(LLVM_ON_UNIX) && \ - !(defined(__NetBSD__) && !defined(_LIBCPP_VERSION)) && !defined(__ppc__) +// std::call_once from libc++ is used on all Unix platforms. Other +// implementations like libstdc++ are known to have problems on NetBSD, +// OpenBSD and PowerPC. +#if defined(LLVM_ON_UNIX) && (defined(_LIBCPP_VERSION) || \ + !(defined(__NetBSD__) || defined(__OpenBSD__) || defined(__ppc__))) #define LLVM_THREADING_USE_STD_CALL_ONCE 1 #else #define LLVM_THREADING_USE_STD_CALL_ONCE 0 Modified: projects/clang391-import/contrib/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -2185,24 +2185,29 @@ void DAGTypeLegalizer::ExpandIntRes_MUL( // options. This is a trivially-generalized version of the code from // Hacker's Delight (itself derived from Knuth's Algorithm M from section // 4.3.1). - SDValue Mask = - DAG.getConstant(APInt::getLowBitsSet(NVT.getSizeInBits(), - NVT.getSizeInBits() >> 1), dl, NVT); + unsigned Bits = NVT.getSizeInBits(); + unsigned HalfBits = Bits >> 1; + SDValue Mask = DAG.getConstant(APInt::getLowBitsSet(Bits, HalfBits), dl, + NVT); SDValue LLL = DAG.getNode(ISD::AND, dl, NVT, LL, Mask); SDValue RLL = DAG.getNode(ISD::AND, dl, NVT, RL, Mask); SDValue T = DAG.getNode(ISD::MUL, dl, NVT, LLL, RLL); SDValue TL = DAG.getNode(ISD::AND, dl, NVT, T, Mask); - SDValue Shift = - DAG.getConstant(NVT.getSizeInBits() >> 1, dl, - TLI.getShiftAmountTy(NVT, DAG.getDataLayout())); + EVT ShiftAmtTy = TLI.getShiftAmountTy(NVT, DAG.getDataLayout()); + if (APInt::getMaxValue(ShiftAmtTy.getSizeInBits()).ult(HalfBits)) { + // The type from TLI is too small to fit the shift amount we want. + // Override it with i32. The shift will have to be legalized. + ShiftAmtTy = MVT::i32; + } + SDValue Shift = DAG.getConstant(HalfBits, dl, ShiftAmtTy); SDValue TH = DAG.getNode(ISD::SRL, dl, NVT, T, Shift); SDValue LLH = DAG.getNode(ISD::SRL, dl, NVT, LL, Shift); SDValue RLH = DAG.getNode(ISD::SRL, dl, NVT, RL, Shift); SDValue U = DAG.getNode(ISD::ADD, dl, NVT, - DAG.getNode(ISD::MUL, dl, NVT, LLH, RLL), TL); + DAG.getNode(ISD::MUL, dl, NVT, LLH, RLL), TH); SDValue UL = DAG.getNode(ISD::AND, dl, NVT, U, Mask); SDValue UH = DAG.getNode(ISD::SRL, dl, NVT, U, Shift); @@ -2211,14 +2216,14 @@ void DAGTypeLegalizer::ExpandIntRes_MUL( SDValue VH = DAG.getNode(ISD::SRL, dl, NVT, V, Shift); SDValue W = DAG.getNode(ISD::ADD, dl, NVT, - DAG.getNode(ISD::MUL, dl, NVT, LL, RL), + DAG.getNode(ISD::MUL, dl, NVT, LLH, RLH), DAG.getNode(ISD::ADD, dl, NVT, UH, VH)); - Lo = DAG.getNode(ISD::ADD, dl, NVT, TH, + Lo = DAG.getNode(ISD::ADD, dl, NVT, TL, DAG.getNode(ISD::SHL, dl, NVT, V, Shift)); Hi = DAG.getNode(ISD::ADD, dl, NVT, W, DAG.getNode(ISD::ADD, dl, NVT, - DAG.getNode(ISD::MUL, dl, NVT, RH, LL), + DAG.getNode(ISD::MUL, dl, NVT, RH, LL), DAG.getNode(ISD::MUL, dl, NVT, RL, LH))); return; } Modified: projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -2203,7 +2203,8 @@ void SIInstrInfo::legalizeOperandsSMRD(M } void SIInstrInfo::legalizeOperands(MachineInstr &MI) const { - MachineRegisterInfo &MRI = MI.getParent()->getParent()->getRegInfo(); + MachineFunction &MF = *MI.getParent()->getParent(); + MachineRegisterInfo &MRI = MF.getRegInfo(); // Legalize VOP2 if (isVOP2(MI) || isVOPC(MI)) { @@ -2321,8 +2322,14 @@ void SIInstrInfo::legalizeOperands(Machi return; } - // Legalize MIMG - if (isMIMG(MI)) { + // Legalize MIMG and MUBUF/MTBUF for shaders. + // + // Shaders only generate MUBUF/MTBUF instructions via intrinsics or via + // scratch memory access. In both cases, the legalization never involves + // conversion to the addr64 form. + if (isMIMG(MI) || + (AMDGPU::isShader(MF.getFunction()->getCallingConv()) && + (isMUBUF(MI) || isMTBUF(MI)))) { MachineOperand *SRsrc = getNamedOperand(MI, AMDGPU::OpName::srsrc); if (SRsrc && !RI.isSGPRClass(MRI.getRegClass(SRsrc->getReg()))) { unsigned SGPR = readlaneVGPRToSGPR(SRsrc->getReg(), MI, MRI); @@ -2337,9 +2344,10 @@ void SIInstrInfo::legalizeOperands(Machi return; } - // Legalize MUBUF* instructions + // Legalize MUBUF* instructions by converting to addr64 form. // FIXME: If we start using the non-addr64 instructions for compute, we - // may need to legalize them here. + // may need to legalize them as above. This especially applies to the + // buffer_load_format_* variants and variants with idxen (or bothen). int SRsrcIdx = AMDGPU::getNamedOperandIdx(MI.getOpcode(), AMDGPU::OpName::srsrc); if (SRsrcIdx != -1) { Modified: projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstructions.td ============================================================================== --- projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstructions.td Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIInstructions.td Fri Dec 2 19:36:28 2016 (r309437) @@ -2029,6 +2029,7 @@ def SI_RETURN : PseudoInstSI < let hasSideEffects = 1; let SALU = 1; let hasNoSchedulingInfo = 1; + let DisableWQM = 1; } let Uses = [EXEC], Defs = [EXEC, VCC, M0], Modified: projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -219,13 +219,6 @@ char SIWholeQuadMode::scanInstructions(M markInstruction(MI, Flags, Worklist); GlobalFlags |= Flags; } - - if (WQMOutputs && MBB.succ_empty()) { - // This is a prolog shader. Make sure we go back to exact mode at the end. - Blocks[&MBB].OutNeeds = StateExact; - Worklist.push_back(&MBB); - GlobalFlags |= StateExact; - } } return GlobalFlags; Modified: projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineCompares.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -634,7 +634,7 @@ static bool canRewriteGEPAsOffset(Value } if (!isa<IntToPtrInst>(V) && !isa<PtrToIntInst>(V) && - !isa<GEPOperator>(V) && !isa<PHINode>(V)) + !isa<GetElementPtrInst>(V) && !isa<PHINode>(V)) // We've found some value that we can't explore which is different from // the base. Therefore we can't do this transformation. return false; Modified: projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/Transforms/InstCombine/InstCombineLoadStoreAlloca.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -579,6 +579,13 @@ static Instruction *unpackLoadToAggregat UndefValue::get(T), NewLoad, 0, Name)); } + // Bail out if the array is too large. Ideally we would like to optimize + // arrays of arbitrary size but this has a terrible impact on compile time. + // The threshold here is chosen arbitrarily, maybe needs a little bit of + // tuning. + if (NumElements > 1024) + return nullptr; + const DataLayout &DL = IC.getDataLayout(); auto EltSize = DL.getTypeAllocSize(ET); auto Align = LI.getAlignment(); @@ -1081,6 +1088,13 @@ static bool unpackStoreToAggregate(InstC return true; } + // Bail out if the array is too large. Ideally we would like to optimize + // arrays of arbitrary size but this has a terrible impact on compile time. + // The threshold here is chosen arbitrarily, maybe needs a little bit of + // tuning. + if (NumElements > 1024) + return false; + const DataLayout &DL = IC.getDataLayout(); auto EltSize = DL.getTypeAllocSize(AT->getElementType()); auto Align = SI.getAlignment(); Modified: projects/clang391-import/contrib/llvm/lib/Transforms/Utils/SimplifyCFG.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/lib/Transforms/Utils/SimplifyCFG.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/lib/Transforms/Utils/SimplifyCFG.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -2024,14 +2024,20 @@ static bool FoldTwoEntryPHINode(PHINode // Move all 'aggressive' instructions, which are defined in the // conditional parts of the if's up to the dominating block. - if (IfBlock1) + if (IfBlock1) { + for (auto &I : *IfBlock1) + I.dropUnknownNonDebugMetadata(); DomBlock->getInstList().splice(InsertPt->getIterator(), IfBlock1->getInstList(), IfBlock1->begin(), IfBlock1->getTerminator()->getIterator()); - if (IfBlock2) + } + if (IfBlock2) { + for (auto &I : *IfBlock2) + I.dropUnknownNonDebugMetadata(); DomBlock->getInstList().splice(InsertPt->getIterator(), IfBlock2->getInstList(), IfBlock2->begin(), IfBlock2->getTerminator()->getIterator()); + } while (PHINode *PN = dyn_cast<PHINode>(BB->begin())) { // Change the PHI node into a select instruction. Modified: projects/clang391-import/contrib/llvm/tools/clang/include/clang/Basic/DiagnosticSemaKinds.td ============================================================================== --- projects/clang391-import/contrib/llvm/tools/clang/include/clang/Basic/DiagnosticSemaKinds.td Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/tools/clang/include/clang/Basic/DiagnosticSemaKinds.td Fri Dec 2 19:36:28 2016 (r309437) @@ -7621,6 +7621,8 @@ def err_invalid_neon_type_code : Error< "incompatible constant for this __builtin_neon function">; def err_argument_invalid_range : Error< "argument should be a value from %0 to %1">; +def err_argument_not_multiple : Error< + "argument should be a multiple of %0">; def warn_neon_vector_initializer_non_portable : Warning< "vector initializers are not compatible with NEON intrinsics in big endian " "mode">, InGroup<DiagGroup<"nonportable-vector-initialization">>; Modified: projects/clang391-import/contrib/llvm/tools/clang/include/clang/Sema/Sema.h ============================================================================== --- projects/clang391-import/contrib/llvm/tools/clang/include/clang/Sema/Sema.h Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/tools/clang/include/clang/Sema/Sema.h Fri Dec 2 19:36:28 2016 (r309437) @@ -9417,6 +9417,8 @@ private: llvm::APSInt &Result); bool SemaBuiltinConstantArgRange(CallExpr *TheCall, int ArgNum, int Low, int High); + bool SemaBuiltinConstantArgMultiple(CallExpr *TheCall, int ArgNum, + unsigned Multiple); bool SemaBuiltinARMSpecialReg(unsigned BuiltinID, CallExpr *TheCall, int ArgNum, unsigned ExpectedFieldNum, bool AllowName); Modified: projects/clang391-import/contrib/llvm/tools/clang/lib/Basic/Targets.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/tools/clang/lib/Basic/Targets.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/tools/clang/lib/Basic/Targets.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -2081,21 +2081,23 @@ public: static GPUKind parseAMDGCNName(StringRef Name) { return llvm::StringSwitch<GPUKind>(Name) - .Case("tahiti", GK_SOUTHERN_ISLANDS) - .Case("pitcairn", GK_SOUTHERN_ISLANDS) - .Case("verde", GK_SOUTHERN_ISLANDS) - .Case("oland", GK_SOUTHERN_ISLANDS) - .Case("hainan", GK_SOUTHERN_ISLANDS) - .Case("bonaire", GK_SEA_ISLANDS) - .Case("kabini", GK_SEA_ISLANDS) - .Case("kaveri", GK_SEA_ISLANDS) - .Case("hawaii", GK_SEA_ISLANDS) - .Case("mullins", GK_SEA_ISLANDS) - .Case("tonga", GK_VOLCANIC_ISLANDS) - .Case("iceland", GK_VOLCANIC_ISLANDS) - .Case("carrizo", GK_VOLCANIC_ISLANDS) - .Case("fiji", GK_VOLCANIC_ISLANDS) - .Case("stoney", GK_VOLCANIC_ISLANDS) + .Case("tahiti", GK_SOUTHERN_ISLANDS) + .Case("pitcairn", GK_SOUTHERN_ISLANDS) + .Case("verde", GK_SOUTHERN_ISLANDS) + .Case("oland", GK_SOUTHERN_ISLANDS) + .Case("hainan", GK_SOUTHERN_ISLANDS) + .Case("bonaire", GK_SEA_ISLANDS) + .Case("kabini", GK_SEA_ISLANDS) + .Case("kaveri", GK_SEA_ISLANDS) + .Case("hawaii", GK_SEA_ISLANDS) + .Case("mullins", GK_SEA_ISLANDS) + .Case("tonga", GK_VOLCANIC_ISLANDS) + .Case("iceland", GK_VOLCANIC_ISLANDS) + .Case("carrizo", GK_VOLCANIC_ISLANDS) + .Case("fiji", GK_VOLCANIC_ISLANDS) + .Case("stoney", GK_VOLCANIC_ISLANDS) + .Case("polaris10", GK_VOLCANIC_ISLANDS) + .Case("polaris11", GK_VOLCANIC_ISLANDS) .Default(GK_NONE); } Modified: projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CGStmtOpenMP.cpp ============================================================================== --- projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CGStmtOpenMP.cpp Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CGStmtOpenMP.cpp Fri Dec 2 19:36:28 2016 (r309437) @@ -1767,17 +1767,11 @@ void CodeGenFunction::EmitOMPOuterLoop(b EmitBlock(LoopExit.getBlock()); // Tell the runtime we are done. - SourceLocation ELoc = S.getLocEnd(); - auto &&CodeGen = [DynamicOrOrdered, ELoc](CodeGenFunction &CGF) { + auto &&CodeGen = [DynamicOrOrdered, &S](CodeGenFunction &CGF) { if (!DynamicOrOrdered) - CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc); + CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd()); }; - CodeGen(*this); - - OpenMPDirectiveKind DKind = S.getDirectiveKind(); - if (DKind == OMPD_for || DKind == OMPD_parallel_for || - DKind == OMPD_distribute_parallel_for) - OMPCancelStack.back().CodeGen = CodeGen; + OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen); } void CodeGenFunction::EmitOMPForOuterLoop( @@ -1889,11 +1883,12 @@ void CodeGenFunction::EmitOMPDistributeO 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 &) { OMPLoopScope PreInitScope(CGF, S); + OMPCancelStackRAII CancelRegion(CGF, OMPD_distribute_parallel_for, + /*HasCancel=*/false); CGF.EmitStmt( cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt()); }); @@ -2082,15 +2077,10 @@ bool CodeGenFunction::EmitOMPWorksharing [](CodeGenFunction &) {}); EmitBlock(LoopExit.getBlock()); // Tell the runtime we are done. - SourceLocation ELoc = S.getLocEnd(); - auto &&CodeGen = [ELoc](CodeGenFunction &CGF) { - CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc); + auto &&CodeGen = [&S](CodeGenFunction &CGF) { + CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd()); }; - CodeGen(*this); - OpenMPDirectiveKind DKind = S.getDirectiveKind(); - if (DKind == OMPD_for || DKind == OMPD_parallel_for || - DKind == OMPD_distribute_parallel_for) - OMPCancelStack.back().CodeGen = CodeGen; + OMPCancelStack.emitExit(*this, S.getDirectiveKind(), CodeGen); } else { const bool IsMonotonic = Ordered || ScheduleKind.Schedule == OMPC_SCHEDULE_static || @@ -2140,11 +2130,11 @@ void CodeGenFunction::EmitOMPForDirectiv bool HasLastprivates = false; auto &&CodeGen = [&S, &HasLastprivates](CodeGenFunction &CGF, PrePostActionTy &) { + OMPCancelStackRAII CancelRegion(CGF, OMPD_for, S.hasCancel()); HasLastprivates = CGF.EmitOMPWorksharingLoop(S); }; { OMPLexicalScope Scope(*this, S, /*AsInlined=*/true); - OMPCancelStackRAII CancelRegion(*this); CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_for, CodeGen, S.hasCancel()); } @@ -2187,7 +2177,6 @@ void CodeGenFunction::EmitSections(const 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. @@ -2282,12 +2271,10 @@ void CodeGenFunction::EmitSections(const CGF.EmitOMPInnerLoop(S, /*RequiresCleanup=*/false, &Cond, &Inc, BodyGen, [](CodeGenFunction &) {}); // Tell the runtime we are done. - SourceLocation ELoc = S.getLocEnd(); - auto &&FinalCodeGen = [ELoc](CodeGenFunction &CGF) { - CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, ELoc); + auto &&CodeGen = [&S](CodeGenFunction &CGF) { + CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocEnd()); }; - FinalCodeGen(CGF); - CGF.OMPCancelStack.back().CodeGen = FinalCodeGen; + CGF.OMPCancelStack.emitExit(CGF, S.getDirectiveKind(), CodeGen); CGF.EmitOMPReductionClauseFinal(S); // Emit post-update of the reduction variables if IsLastIter != 0. emitPostUpdateForReductionClause( @@ -2309,6 +2296,7 @@ void CodeGenFunction::EmitSections(const HasCancel = OSD->hasCancel(); else if (auto *OPSD = dyn_cast<OMPParallelSectionsDirective>(&S)) HasCancel = OPSD->hasCancel(); + OMPCancelStackRAII CancelRegion(*this, S.getDirectiveKind(), HasCancel); CGM.getOpenMPRuntime().emitInlinedDirective(*this, OMPD_sections, CodeGen, HasCancel); // Emit barrier for lastprivates only if 'sections' directive has 'nowait' @@ -2412,7 +2400,7 @@ void CodeGenFunction::EmitOMPParallelFor // 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); + OMPCancelStackRAII CancelRegion(CGF, OMPD_parallel_for, S.hasCancel()); CGF.EmitOMPWorksharingLoop(S); }; emitCommonOMPParallelDirective(*this, S, OMPD_for, CodeGen); @@ -3412,14 +3400,14 @@ void CodeGenFunction::EmitOMPCancelDirec CodeGenFunction::JumpDest CodeGenFunction::getOMPCancelDestination(OpenMPDirectiveKind Kind) { - if (Kind == OMPD_parallel || Kind == OMPD_task) + if (Kind == OMPD_parallel || Kind == OMPD_task || + Kind == OMPD_target_parallel) return ReturnBlock; assert(Kind == OMPD_for || Kind == OMPD_section || Kind == OMPD_sections || 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; + Kind == OMPD_distribute_parallel_for || + Kind == OMPD_target_parallel_for); + return OMPCancelStack.getExitBlock(); } // Generate the instructions for '#pragma omp target data' directive. Modified: projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CodeGenFunction.h ============================================================================== --- projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CodeGenFunction.h Fri Dec 2 19:23:20 2016 (r309436) +++ projects/clang391-import/contrib/llvm/tools/clang/lib/CodeGen/CodeGenFunction.h Fri Dec 2 19:36:28 2016 (r309437) @@ -965,33 +965,92 @@ 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) {}) {} + /// Handles cancellation exit points in OpenMP-related constructs. + class OpenMPCancelExitStack { + /// Tracks cancellation exit point and join point for cancel-related exit + /// and normal exit. + struct CancelExit { + CancelExit() = default; + CancelExit(OpenMPDirectiveKind Kind, JumpDest ExitBlock, + JumpDest ContBlock) + : Kind(Kind), ExitBlock(ExitBlock), ContBlock(ContBlock) {} + OpenMPDirectiveKind Kind = OMPD_unknown; + /// true if the exit block has been emitted already by the special + /// emitExit() call, false if the default codegen is used. + bool HasBeenEmitted = false; + JumpDest ExitBlock; + JumpDest ContBlock; + }; + + SmallVector<CancelExit, 8> Stack; + + public: + OpenMPCancelExitStack() : Stack(1) {} + ~OpenMPCancelExitStack() = default; + /// Fetches the exit block for the current OpenMP construct. + JumpDest getExitBlock() const { return Stack.back().ExitBlock; } + /// Emits exit block with special codegen procedure specific for the related + /// OpenMP construct + emits code for normal construct cleanup. + void emitExit(CodeGenFunction &CGF, OpenMPDirectiveKind Kind, + const llvm::function_ref<void(CodeGenFunction &)> &CodeGen) { + if (Stack.back().Kind == Kind && getExitBlock().isValid()) { + assert(CGF.getOMPCancelDestination(Kind).isValid()); + assert(CGF.HaveInsertPoint()); + assert(!Stack.back().HasBeenEmitted); + auto IP = CGF.Builder.saveAndClearIP(); + CGF.EmitBlock(Stack.back().ExitBlock.getBlock()); + CodeGen(CGF); + CGF.EmitBranchThroughCleanup(Stack.back().ContBlock); + CGF.Builder.restoreIP(IP); + Stack.back().HasBeenEmitted = true; + } + CodeGen(CGF); + } + /// Enter the cancel supporting \a Kind construct. + /// \param Kind OpenMP directive that supports cancel constructs. + /// \param HasCancel true, if the construct has inner cancel directive, + /// false otherwise. + void enter(CodeGenFunction &CGF, OpenMPDirectiveKind Kind, bool HasCancel) { + Stack.push_back({Kind, + HasCancel ? CGF.getJumpDestInCurrentScope("cancel.exit") + : JumpDest(), + HasCancel ? CGF.getJumpDestInCurrentScope("cancel.cont") + : JumpDest()}); + } + /// Emits default exit point for the cancel construct (if the special one + /// has not be used) + join point for cancel/normal exits. + void exit(CodeGenFunction &CGF) { + if (getExitBlock().isValid()) { + assert(CGF.getOMPCancelDestination(Stack.back().Kind).isValid()); + bool HaveIP = CGF.HaveInsertPoint(); + if (!Stack.back().HasBeenEmitted) { + if (HaveIP) + CGF.EmitBranchThroughCleanup(Stack.back().ContBlock); + CGF.EmitBlock(Stack.back().ExitBlock.getBlock()); + CGF.EmitBranchThroughCleanup(Stack.back().ContBlock); + } + CGF.EmitBlock(Stack.back().ContBlock.getBlock()); + if (!HaveIP) { + CGF.Builder.CreateUnreachable(); + CGF.Builder.ClearInsertionPoint(); + } + } + Stack.pop_back(); + } }; - SmallVector<OMPCancel, 8> OMPCancelStack; + OpenMPCancelExitStack 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()); - } + OMPCancelStackRAII(CodeGenFunction &CGF, OpenMPDirectiveKind Kind, + bool HasCancel) + : CGF(CGF) { + CGF.OMPCancelStack.enter(CGF, Kind, HasCancel); } + ~OMPCancelStackRAII() { CGF.OMPCancelStack.exit(CGF); } }; CodeGenPGO PGO; Copied: projects/clang391-import/contrib/llvm/tools/clang/lib/Headers/msa.h (from r309436, vendor/clang/dist/lib/Headers/msa.h) ============================================================================== --- /dev/null 00:00:00 1970 (empty, because file is newly added) +++ projects/clang391-import/contrib/llvm/tools/clang/lib/Headers/msa.h Fri Dec 2 19:36:28 2016 (r309437, copy of r309436, vendor/clang/dist/lib/Headers/msa.h) @@ -0,0 +1,583 @@ +/*===---- msa.h - MIPS MSA intrinsics --------------------------------------=== + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to deal + * in the Software without restriction, including without limitation the rights + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell + * copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN + * THE SOFTWARE. + * + *===-----------------------------------------------------------------------=== + */ + +#ifndef _MSA_H +#define _MSA_H 1 + +#if defined(__mips_msa) +typedef signed char v16i8 __attribute__((vector_size(16), aligned(16))); +typedef signed char v16i8_b __attribute__((vector_size(16), aligned(1))); +typedef unsigned char v16u8 __attribute__((vector_size(16), aligned(16))); +typedef unsigned char v16u8_b __attribute__((vector_size(16), aligned(1))); +typedef short v8i16 __attribute__((vector_size(16), aligned(16))); +typedef short v8i16_h __attribute__((vector_size(16), aligned(2))); +typedef unsigned short v8u16 __attribute__((vector_size(16), aligned(16))); +typedef unsigned short v8u16_h __attribute__((vector_size(16), aligned(2))); +typedef int v4i32 __attribute__((vector_size(16), aligned(16))); +typedef int v4i32_w __attribute__((vector_size(16), aligned(4))); +typedef unsigned int v4u32 __attribute__((vector_size(16), aligned(16))); +typedef unsigned int v4u32_w __attribute__((vector_size(16), aligned(4))); +typedef long long v2i64 __attribute__((vector_size(16), aligned(16))); +typedef long long v2i64_d __attribute__((vector_size(16), aligned(8))); +typedef unsigned long long v2u64 __attribute__((vector_size(16), aligned(16))); +typedef unsigned long long v2u64_d __attribute__((vector_size(16), aligned(8))); +typedef float v4f32 __attribute__((vector_size(16), aligned(16))); +typedef float v4f32_w __attribute__((vector_size(16), aligned(4))); +typedef double v2f64 __attribute__ ((vector_size(16), aligned(16))); +typedef double v2f64_d __attribute__ ((vector_size(16), aligned(8))); + +#define __msa_sll_b __builtin_msa_sll_b +#define __msa_sll_h __builtin_msa_sll_h +#define __msa_sll_w __builtin_msa_sll_w +#define __msa_sll_d __builtin_msa_sll_d +#define __msa_slli_b __builtin_msa_slli_b +#define __msa_slli_h __builtin_msa_slli_h +#define __msa_slli_w __builtin_msa_slli_w +#define __msa_slli_d __builtin_msa_slli_d +#define __msa_sra_b __builtin_msa_sra_b +#define __msa_sra_h __builtin_msa_sra_h +#define __msa_sra_w __builtin_msa_sra_w +#define __msa_sra_d __builtin_msa_sra_d +#define __msa_srai_b __builtin_msa_srai_b +#define __msa_srai_h __builtin_msa_srai_h +#define __msa_srai_w __builtin_msa_srai_w +#define __msa_srai_d __builtin_msa_srai_d +#define __msa_srar_b __builtin_msa_srar_b +#define __msa_srar_h __builtin_msa_srar_h +#define __msa_srar_w __builtin_msa_srar_w +#define __msa_srar_d __builtin_msa_srar_d +#define __msa_srari_b __builtin_msa_srari_b +#define __msa_srari_h __builtin_msa_srari_h +#define __msa_srari_w __builtin_msa_srari_w +#define __msa_srari_d __builtin_msa_srari_d +#define __msa_srl_b __builtin_msa_srl_b +#define __msa_srl_h __builtin_msa_srl_h +#define __msa_srl_w __builtin_msa_srl_w +#define __msa_srl_d __builtin_msa_srl_d +#define __msa_srli_b __builtin_msa_srli_b +#define __msa_srli_h __builtin_msa_srli_h +#define __msa_srli_w __builtin_msa_srli_w +#define __msa_srli_d __builtin_msa_srli_d +#define __msa_srlr_b __builtin_msa_srlr_b +#define __msa_srlr_h __builtin_msa_srlr_h +#define __msa_srlr_w __builtin_msa_srlr_w +#define __msa_srlr_d __builtin_msa_srlr_d +#define __msa_srlri_b __builtin_msa_srlri_b +#define __msa_srlri_h __builtin_msa_srlri_h +#define __msa_srlri_w __builtin_msa_srlri_w +#define __msa_srlri_d __builtin_msa_srlri_d +#define __msa_bclr_b __builtin_msa_bclr_b +#define __msa_bclr_h __builtin_msa_bclr_h +#define __msa_bclr_w __builtin_msa_bclr_w +#define __msa_bclr_d __builtin_msa_bclr_d +#define __msa_bclri_b __builtin_msa_bclri_b +#define __msa_bclri_h __builtin_msa_bclri_h +#define __msa_bclri_w __builtin_msa_bclri_w +#define __msa_bclri_d __builtin_msa_bclri_d +#define __msa_bset_b __builtin_msa_bset_b +#define __msa_bset_h __builtin_msa_bset_h +#define __msa_bset_w __builtin_msa_bset_w +#define __msa_bset_d __builtin_msa_bset_d +#define __msa_bseti_b __builtin_msa_bseti_b +#define __msa_bseti_h __builtin_msa_bseti_h +#define __msa_bseti_w __builtin_msa_bseti_w +#define __msa_bseti_d __builtin_msa_bseti_d +#define __msa_bneg_b __builtin_msa_bneg_b +#define __msa_bneg_h __builtin_msa_bneg_h +#define __msa_bneg_w __builtin_msa_bneg_w +#define __msa_bneg_d __builtin_msa_bneg_d +#define __msa_bnegi_b __builtin_msa_bnegi_b +#define __msa_bnegi_h __builtin_msa_bnegi_h +#define __msa_bnegi_w __builtin_msa_bnegi_w +#define __msa_bnegi_d __builtin_msa_bnegi_d +#define __msa_binsl_b __builtin_msa_binsl_b +#define __msa_binsl_h __builtin_msa_binsl_h +#define __msa_binsl_w __builtin_msa_binsl_w +#define __msa_binsl_d __builtin_msa_binsl_d +#define __msa_binsli_b __builtin_msa_binsli_b +#define __msa_binsli_h __builtin_msa_binsli_h +#define __msa_binsli_w __builtin_msa_binsli_w +#define __msa_binsli_d __builtin_msa_binsli_d +#define __msa_binsr_b __builtin_msa_binsr_b +#define __msa_binsr_h __builtin_msa_binsr_h +#define __msa_binsr_w __builtin_msa_binsr_w +#define __msa_binsr_d __builtin_msa_binsr_d +#define __msa_binsri_b __builtin_msa_binsri_b +#define __msa_binsri_h __builtin_msa_binsri_h +#define __msa_binsri_w __builtin_msa_binsri_w +#define __msa_binsri_d __builtin_msa_binsri_d +#define __msa_addv_b __builtin_msa_addv_b +#define __msa_addv_h __builtin_msa_addv_h +#define __msa_addv_w __builtin_msa_addv_w +#define __msa_addv_d __builtin_msa_addv_d +#define __msa_addvi_b __builtin_msa_addvi_b +#define __msa_addvi_h __builtin_msa_addvi_h +#define __msa_addvi_w __builtin_msa_addvi_w +#define __msa_addvi_d __builtin_msa_addvi_d +#define __msa_subv_b __builtin_msa_subv_b +#define __msa_subv_h __builtin_msa_subv_h +#define __msa_subv_w __builtin_msa_subv_w +#define __msa_subv_d __builtin_msa_subv_d +#define __msa_subvi_b __builtin_msa_subvi_b +#define __msa_subvi_h __builtin_msa_subvi_h +#define __msa_subvi_w __builtin_msa_subvi_w +#define __msa_subvi_d __builtin_msa_subvi_d +#define __msa_max_s_b __builtin_msa_max_s_b +#define __msa_max_s_h __builtin_msa_max_s_h +#define __msa_max_s_w __builtin_msa_max_s_w +#define __msa_max_s_d __builtin_msa_max_s_d +#define __msa_maxi_s_b __builtin_msa_maxi_s_b +#define __msa_maxi_s_h __builtin_msa_maxi_s_h +#define __msa_maxi_s_w __builtin_msa_maxi_s_w +#define __msa_maxi_s_d __builtin_msa_maxi_s_d +#define __msa_max_u_b __builtin_msa_max_u_b +#define __msa_max_u_h __builtin_msa_max_u_h +#define __msa_max_u_w __builtin_msa_max_u_w +#define __msa_max_u_d __builtin_msa_max_u_d +#define __msa_maxi_u_b __builtin_msa_maxi_u_b +#define __msa_maxi_u_h __builtin_msa_maxi_u_h +#define __msa_maxi_u_w __builtin_msa_maxi_u_w +#define __msa_maxi_u_d __builtin_msa_maxi_u_d +#define __msa_min_s_b __builtin_msa_min_s_b +#define __msa_min_s_h __builtin_msa_min_s_h +#define __msa_min_s_w __builtin_msa_min_s_w +#define __msa_min_s_d __builtin_msa_min_s_d +#define __msa_mini_s_b __builtin_msa_mini_s_b +#define __msa_mini_s_h __builtin_msa_mini_s_h +#define __msa_mini_s_w __builtin_msa_mini_s_w +#define __msa_mini_s_d __builtin_msa_mini_s_d +#define __msa_min_u_b __builtin_msa_min_u_b +#define __msa_min_u_h __builtin_msa_min_u_h +#define __msa_min_u_w __builtin_msa_min_u_w +#define __msa_min_u_d __builtin_msa_min_u_d +#define __msa_mini_u_b __builtin_msa_mini_u_b +#define __msa_mini_u_h __builtin_msa_mini_u_h +#define __msa_mini_u_w __builtin_msa_mini_u_w +#define __msa_mini_u_d __builtin_msa_mini_u_d +#define __msa_max_a_b __builtin_msa_max_a_b +#define __msa_max_a_h __builtin_msa_max_a_h +#define __msa_max_a_w __builtin_msa_max_a_w +#define __msa_max_a_d __builtin_msa_max_a_d +#define __msa_min_a_b __builtin_msa_min_a_b +#define __msa_min_a_h __builtin_msa_min_a_h +#define __msa_min_a_w __builtin_msa_min_a_w +#define __msa_min_a_d __builtin_msa_min_a_d +#define __msa_ceq_b __builtin_msa_ceq_b +#define __msa_ceq_h __builtin_msa_ceq_h +#define __msa_ceq_w __builtin_msa_ceq_w +#define __msa_ceq_d __builtin_msa_ceq_d +#define __msa_ceqi_b __builtin_msa_ceqi_b +#define __msa_ceqi_h __builtin_msa_ceqi_h +#define __msa_ceqi_w __builtin_msa_ceqi_w +#define __msa_ceqi_d __builtin_msa_ceqi_d +#define __msa_clt_s_b __builtin_msa_clt_s_b +#define __msa_clt_s_h __builtin_msa_clt_s_h +#define __msa_clt_s_w __builtin_msa_clt_s_w +#define __msa_clt_s_d __builtin_msa_clt_s_d +#define __msa_clti_s_b __builtin_msa_clti_s_b +#define __msa_clti_s_h __builtin_msa_clti_s_h +#define __msa_clti_s_w __builtin_msa_clti_s_w +#define __msa_clti_s_d __builtin_msa_clti_s_d +#define __msa_clt_u_b __builtin_msa_clt_u_b +#define __msa_clt_u_h __builtin_msa_clt_u_h +#define __msa_clt_u_w __builtin_msa_clt_u_w +#define __msa_clt_u_d __builtin_msa_clt_u_d +#define __msa_clti_u_b __builtin_msa_clti_u_b +#define __msa_clti_u_h __builtin_msa_clti_u_h +#define __msa_clti_u_w __builtin_msa_clti_u_w +#define __msa_clti_u_d __builtin_msa_clti_u_d +#define __msa_cle_s_b __builtin_msa_cle_s_b +#define __msa_cle_s_h __builtin_msa_cle_s_h +#define __msa_cle_s_w __builtin_msa_cle_s_w +#define __msa_cle_s_d __builtin_msa_cle_s_d +#define __msa_clei_s_b __builtin_msa_clei_s_b +#define __msa_clei_s_h __builtin_msa_clei_s_h +#define __msa_clei_s_w __builtin_msa_clei_s_w +#define __msa_clei_s_d __builtin_msa_clei_s_d +#define __msa_cle_u_b __builtin_msa_cle_u_b +#define __msa_cle_u_h __builtin_msa_cle_u_h +#define __msa_cle_u_w __builtin_msa_cle_u_w +#define __msa_cle_u_d __builtin_msa_cle_u_d +#define __msa_clei_u_b __builtin_msa_clei_u_b +#define __msa_clei_u_h __builtin_msa_clei_u_h +#define __msa_clei_u_w __builtin_msa_clei_u_w +#define __msa_clei_u_d __builtin_msa_clei_u_d +#define __msa_ld_b __builtin_msa_ld_b +#define __msa_ld_h __builtin_msa_ld_h +#define __msa_ld_w __builtin_msa_ld_w +#define __msa_ld_d __builtin_msa_ld_d +#define __msa_st_b __builtin_msa_st_b +#define __msa_st_h __builtin_msa_st_h +#define __msa_st_w __builtin_msa_st_w +#define __msa_st_d __builtin_msa_st_d +#define __msa_sat_s_b __builtin_msa_sat_s_b +#define __msa_sat_s_h __builtin_msa_sat_s_h +#define __msa_sat_s_w __builtin_msa_sat_s_w +#define __msa_sat_s_d __builtin_msa_sat_s_d +#define __msa_sat_u_b __builtin_msa_sat_u_b +#define __msa_sat_u_h __builtin_msa_sat_u_h +#define __msa_sat_u_w __builtin_msa_sat_u_w +#define __msa_sat_u_d __builtin_msa_sat_u_d +#define __msa_add_a_b __builtin_msa_add_a_b +#define __msa_add_a_h __builtin_msa_add_a_h +#define __msa_add_a_w __builtin_msa_add_a_w +#define __msa_add_a_d __builtin_msa_add_a_d +#define __msa_adds_a_b __builtin_msa_adds_a_b +#define __msa_adds_a_h __builtin_msa_adds_a_h +#define __msa_adds_a_w __builtin_msa_adds_a_w +#define __msa_adds_a_d __builtin_msa_adds_a_d +#define __msa_adds_s_b __builtin_msa_adds_s_b +#define __msa_adds_s_h __builtin_msa_adds_s_h +#define __msa_adds_s_w __builtin_msa_adds_s_w +#define __msa_adds_s_d __builtin_msa_adds_s_d +#define __msa_adds_u_b __builtin_msa_adds_u_b +#define __msa_adds_u_h __builtin_msa_adds_u_h +#define __msa_adds_u_w __builtin_msa_adds_u_w +#define __msa_adds_u_d __builtin_msa_adds_u_d +#define __msa_ave_s_b __builtin_msa_ave_s_b +#define __msa_ave_s_h __builtin_msa_ave_s_h +#define __msa_ave_s_w __builtin_msa_ave_s_w +#define __msa_ave_s_d __builtin_msa_ave_s_d +#define __msa_ave_u_b __builtin_msa_ave_u_b +#define __msa_ave_u_h __builtin_msa_ave_u_h +#define __msa_ave_u_w __builtin_msa_ave_u_w +#define __msa_ave_u_d __builtin_msa_ave_u_d +#define __msa_aver_s_b __builtin_msa_aver_s_b +#define __msa_aver_s_h __builtin_msa_aver_s_h +#define __msa_aver_s_w __builtin_msa_aver_s_w +#define __msa_aver_s_d __builtin_msa_aver_s_d +#define __msa_aver_u_b __builtin_msa_aver_u_b +#define __msa_aver_u_h __builtin_msa_aver_u_h +#define __msa_aver_u_w __builtin_msa_aver_u_w +#define __msa_aver_u_d __builtin_msa_aver_u_d +#define __msa_subs_s_b __builtin_msa_subs_s_b +#define __msa_subs_s_h __builtin_msa_subs_s_h +#define __msa_subs_s_w __builtin_msa_subs_s_w +#define __msa_subs_s_d __builtin_msa_subs_s_d +#define __msa_subs_u_b __builtin_msa_subs_u_b +#define __msa_subs_u_h __builtin_msa_subs_u_h +#define __msa_subs_u_w __builtin_msa_subs_u_w +#define __msa_subs_u_d __builtin_msa_subs_u_d +#define __msa_subsuu_s_b __builtin_msa_subsuu_s_b +#define __msa_subsuu_s_h __builtin_msa_subsuu_s_h +#define __msa_subsuu_s_w __builtin_msa_subsuu_s_w +#define __msa_subsuu_s_d __builtin_msa_subsuu_s_d +#define __msa_subsus_u_b __builtin_msa_subsus_u_b +#define __msa_subsus_u_h __builtin_msa_subsus_u_h +#define __msa_subsus_u_w __builtin_msa_subsus_u_w +#define __msa_subsus_u_d __builtin_msa_subsus_u_d +#define __msa_asub_s_b __builtin_msa_asub_s_b +#define __msa_asub_s_h __builtin_msa_asub_s_h +#define __msa_asub_s_w __builtin_msa_asub_s_w +#define __msa_asub_s_d __builtin_msa_asub_s_d +#define __msa_asub_u_b __builtin_msa_asub_u_b +#define __msa_asub_u_h __builtin_msa_asub_u_h +#define __msa_asub_u_w __builtin_msa_asub_u_w +#define __msa_asub_u_d __builtin_msa_asub_u_d +#define __msa_mulv_b __builtin_msa_mulv_b +#define __msa_mulv_h __builtin_msa_mulv_h +#define __msa_mulv_w __builtin_msa_mulv_w +#define __msa_mulv_d __builtin_msa_mulv_d +#define __msa_maddv_b __builtin_msa_maddv_b +#define __msa_maddv_h __builtin_msa_maddv_h +#define __msa_maddv_w __builtin_msa_maddv_w +#define __msa_maddv_d __builtin_msa_maddv_d +#define __msa_msubv_b __builtin_msa_msubv_b +#define __msa_msubv_h __builtin_msa_msubv_h +#define __msa_msubv_w __builtin_msa_msubv_w +#define __msa_msubv_d __builtin_msa_msubv_d +#define __msa_div_s_b __builtin_msa_div_s_b +#define __msa_div_s_h __builtin_msa_div_s_h +#define __msa_div_s_w __builtin_msa_div_s_w +#define __msa_div_s_d __builtin_msa_div_s_d +#define __msa_div_u_b __builtin_msa_div_u_b +#define __msa_div_u_h __builtin_msa_div_u_h +#define __msa_div_u_w __builtin_msa_div_u_w +#define __msa_div_u_d __builtin_msa_div_u_d +#define __msa_hadd_s_h __builtin_msa_hadd_s_h +#define __msa_hadd_s_w __builtin_msa_hadd_s_w +#define __msa_hadd_s_d __builtin_msa_hadd_s_d +#define __msa_hadd_u_h __builtin_msa_hadd_u_h +#define __msa_hadd_u_w __builtin_msa_hadd_u_w +#define __msa_hadd_u_d __builtin_msa_hadd_u_d +#define __msa_hsub_s_h __builtin_msa_hsub_s_h +#define __msa_hsub_s_w __builtin_msa_hsub_s_w +#define __msa_hsub_s_d __builtin_msa_hsub_s_d +#define __msa_hsub_u_h __builtin_msa_hsub_u_h +#define __msa_hsub_u_w __builtin_msa_hsub_u_w +#define __msa_hsub_u_d __builtin_msa_hsub_u_d +#define __msa_mod_s_b __builtin_msa_mod_s_b +#define __msa_mod_s_h __builtin_msa_mod_s_h +#define __msa_mod_s_w __builtin_msa_mod_s_w +#define __msa_mod_s_d __builtin_msa_mod_s_d +#define __msa_mod_u_b __builtin_msa_mod_u_b +#define __msa_mod_u_h __builtin_msa_mod_u_h +#define __msa_mod_u_w __builtin_msa_mod_u_w +#define __msa_mod_u_d __builtin_msa_mod_u_d +#define __msa_dotp_s_h __builtin_msa_dotp_s_h +#define __msa_dotp_s_w __builtin_msa_dotp_s_w +#define __msa_dotp_s_d __builtin_msa_dotp_s_d +#define __msa_dotp_u_h __builtin_msa_dotp_u_h +#define __msa_dotp_u_w __builtin_msa_dotp_u_w +#define __msa_dotp_u_d __builtin_msa_dotp_u_d +#define __msa_dpadd_s_h __builtin_msa_dpadd_s_h +#define __msa_dpadd_s_w __builtin_msa_dpadd_s_w +#define __msa_dpadd_s_d __builtin_msa_dpadd_s_d +#define __msa_dpadd_u_h __builtin_msa_dpadd_u_h +#define __msa_dpadd_u_w __builtin_msa_dpadd_u_w +#define __msa_dpadd_u_d __builtin_msa_dpadd_u_d +#define __msa_dpsub_s_h __builtin_msa_dpsub_s_h +#define __msa_dpsub_s_w __builtin_msa_dpsub_s_w +#define __msa_dpsub_s_d __builtin_msa_dpsub_s_d +#define __msa_dpsub_u_h __builtin_msa_dpsub_u_h +#define __msa_dpsub_u_w __builtin_msa_dpsub_u_w +#define __msa_dpsub_u_d __builtin_msa_dpsub_u_d +#define __msa_sld_b __builtin_msa_sld_b +#define __msa_sld_h __builtin_msa_sld_h +#define __msa_sld_w __builtin_msa_sld_w +#define __msa_sld_d __builtin_msa_sld_d +#define __msa_sldi_b __builtin_msa_sldi_b +#define __msa_sldi_h __builtin_msa_sldi_h +#define __msa_sldi_w __builtin_msa_sldi_w +#define __msa_sldi_d __builtin_msa_sldi_d +#define __msa_splat_b __builtin_msa_splat_b +#define __msa_splat_h __builtin_msa_splat_h +#define __msa_splat_w __builtin_msa_splat_w +#define __msa_splat_d __builtin_msa_splat_d +#define __msa_splati_b __builtin_msa_splati_b +#define __msa_splati_h __builtin_msa_splati_h +#define __msa_splati_w __builtin_msa_splati_w +#define __msa_splati_d __builtin_msa_splati_d +#define __msa_pckev_b __builtin_msa_pckev_b +#define __msa_pckev_h __builtin_msa_pckev_h +#define __msa_pckev_w __builtin_msa_pckev_w +#define __msa_pckev_d __builtin_msa_pckev_d +#define __msa_pckod_b __builtin_msa_pckod_b +#define __msa_pckod_h __builtin_msa_pckod_h +#define __msa_pckod_w __builtin_msa_pckod_w +#define __msa_pckod_d __builtin_msa_pckod_d +#define __msa_ilvl_b __builtin_msa_ilvl_b +#define __msa_ilvl_h __builtin_msa_ilvl_h +#define __msa_ilvl_w __builtin_msa_ilvl_w +#define __msa_ilvl_d __builtin_msa_ilvl_d +#define __msa_ilvr_b __builtin_msa_ilvr_b +#define __msa_ilvr_h __builtin_msa_ilvr_h +#define __msa_ilvr_w __builtin_msa_ilvr_w +#define __msa_ilvr_d __builtin_msa_ilvr_d +#define __msa_ilvev_b __builtin_msa_ilvev_b +#define __msa_ilvev_h __builtin_msa_ilvev_h +#define __msa_ilvev_w __builtin_msa_ilvev_w +#define __msa_ilvev_d __builtin_msa_ilvev_d +#define __msa_ilvod_b __builtin_msa_ilvod_b +#define __msa_ilvod_h __builtin_msa_ilvod_h +#define __msa_ilvod_w __builtin_msa_ilvod_w +#define __msa_ilvod_d __builtin_msa_ilvod_d +#define __msa_vshf_b __builtin_msa_vshf_b +#define __msa_vshf_h __builtin_msa_vshf_h +#define __msa_vshf_w __builtin_msa_vshf_w +#define __msa_vshf_d __builtin_msa_vshf_d +#define __msa_and_v __builtin_msa_and_v +#define __msa_andi_b __builtin_msa_andi_b +#define __msa_or_v __builtin_msa_or_v +#define __msa_ori_b __builtin_msa_ori_b +#define __msa_nor_v __builtin_msa_nor_v +#define __msa_nori_b __builtin_msa_nori_b +#define __msa_xor_v __builtin_msa_xor_v +#define __msa_xori_b __builtin_msa_xori_b +#define __msa_bmnz_v __builtin_msa_bmnz_v +#define __msa_bmnzi_b __builtin_msa_bmnzi_b +#define __msa_bmz_v __builtin_msa_bmz_v +#define __msa_bmzi_b __builtin_msa_bmzi_b +#define __msa_bsel_v __builtin_msa_bsel_v +#define __msa_bseli_b __builtin_msa_bseli_b +#define __msa_shf_b __builtin_msa_shf_b +#define __msa_shf_h __builtin_msa_shf_h +#define __msa_shf_w __builtin_msa_shf_w +#define __msa_test_bnz_v __builtin_msa_bnz_v +#define __msa_test_bz_v __builtin_msa_bz_v +#define __msa_fill_b __builtin_msa_fill_b +#define __msa_fill_h __builtin_msa_fill_h +#define __msa_fill_w __builtin_msa_fill_w +#define __msa_fill_d __builtin_msa_fill_d +#define __msa_pcnt_b __builtin_msa_pcnt_b +#define __msa_pcnt_h __builtin_msa_pcnt_h +#define __msa_pcnt_w __builtin_msa_pcnt_w +#define __msa_pcnt_d __builtin_msa_pcnt_d +#define __msa_nloc_b __builtin_msa_nloc_b +#define __msa_nloc_h __builtin_msa_nloc_h +#define __msa_nloc_w __builtin_msa_nloc_w +#define __msa_nloc_d __builtin_msa_nloc_d +#define __msa_nlzc_b __builtin_msa_nlzc_b +#define __msa_nlzc_h __builtin_msa_nlzc_h +#define __msa_nlzc_w __builtin_msa_nlzc_w +#define __msa_nlzc_d __builtin_msa_nlzc_d +#define __msa_copy_s_b __builtin_msa_copy_s_b +#define __msa_copy_s_h __builtin_msa_copy_s_h +#define __msa_copy_s_w __builtin_msa_copy_s_w +#define __msa_copy_s_d __builtin_msa_copy_s_d +#define __msa_copy_u_b __builtin_msa_copy_u_b +#define __msa_copy_u_h __builtin_msa_copy_u_h +#define __msa_copy_u_w __builtin_msa_copy_u_w +#define __msa_copy_u_d __builtin_msa_copy_u_d +#define __msa_insert_b __builtin_msa_insert_b +#define __msa_insert_h __builtin_msa_insert_h +#define __msa_insert_w __builtin_msa_insert_w +#define __msa_insert_d __builtin_msa_insert_d +#define __msa_insve_b __builtin_msa_insve_b +#define __msa_insve_h __builtin_msa_insve_h +#define __msa_insve_w __builtin_msa_insve_w +#define __msa_insve_d __builtin_msa_insve_d +#define __msa_test_bnz_b __builtin_msa_bnz_b +#define __msa_test_bnz_h __builtin_msa_bnz_h +#define __msa_test_bnz_w __builtin_msa_bnz_w +#define __msa_test_bnz_d __builtin_msa_bnz_d +#define __msa_test_bz_b __builtin_msa_bz_b +#define __msa_test_bz_h __builtin_msa_bz_h +#define __msa_test_bz_w __builtin_msa_bz_w +#define __msa_test_bz_d __builtin_msa_bz_d +#define __msa_ldi_b __builtin_msa_ldi_b +#define __msa_ldi_h __builtin_msa_ldi_h +#define __msa_ldi_w __builtin_msa_ldi_w +#define __msa_ldi_d __builtin_msa_ldi_d +#define __msa_fcaf_w __builtin_msa_fcaf_w +#define __msa_fcaf_d __builtin_msa_fcaf_d +#define __msa_fcor_w __builtin_msa_fcor_w +#define __msa_fcor_d __builtin_msa_fcor_d +#define __msa_fcun_w __builtin_msa_fcun_w +#define __msa_fcun_d __builtin_msa_fcun_d +#define __msa_fcune_w __builtin_msa_fcune_w +#define __msa_fcune_d __builtin_msa_fcune_d +#define __msa_fcueq_w __builtin_msa_fcueq_w +#define __msa_fcueq_d __builtin_msa_fcueq_d +#define __msa_fceq_w __builtin_msa_fceq_w +#define __msa_fceq_d __builtin_msa_fceq_d +#define __msa_fcne_w __builtin_msa_fcne_w +#define __msa_fcne_d __builtin_msa_fcne_d +#define __msa_fclt_w __builtin_msa_fclt_w +#define __msa_fclt_d __builtin_msa_fclt_d +#define __msa_fcult_w __builtin_msa_fcult_w +#define __msa_fcult_d __builtin_msa_fcult_d +#define __msa_fcle_w __builtin_msa_fcle_w *** DIFF OUTPUT TRUNCATED AT 1000 LINES ***
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201612021936.uB2JaSNP083228>