diff options
Diffstat (limited to 'llvm')
68 files changed, 5689 insertions, 390 deletions
diff --git a/llvm/docs/ReleaseNotes.md b/llvm/docs/ReleaseNotes.md index 46a9f32..640516a 100644 --- a/llvm/docs/ReleaseNotes.md +++ b/llvm/docs/ReleaseNotes.md @@ -134,6 +134,8 @@ Changes to the WebAssembly Backend Changes to the Windows Target ----------------------------- +* `-fpseudo-probe-for-profiling` is now supported for COFF. + Changes to the X86 Backend -------------------------- @@ -161,6 +163,8 @@ Changes to the Debug Info Changes to the LLVM tools --------------------------------- +* `llvm-profgen` now supports decoding pseudo probe for COFF binaries. + * `llvm-readelf` now dumps all hex format values in lower-case mode. * Some code paths for supporting Python 2.7 in `llvm-lit` have been removed. * Support for `%T` in lit has been removed. diff --git a/llvm/include/llvm/ExecutionEngine/Orc/Shared/ExecutorAddress.h b/llvm/include/llvm/ExecutionEngine/Orc/Shared/ExecutorAddress.h index a4a7fa4..a5f6c4f 100644 --- a/llvm/include/llvm/ExecutionEngine/Orc/Shared/ExecutorAddress.h +++ b/llvm/include/llvm/ExecutionEngine/Orc/Shared/ExecutorAddress.h @@ -272,6 +272,9 @@ struct ExecutorAddrRange { } bool contains(ExecutorAddr Addr) const { return Start <= Addr && Addr < End; } + bool contains(const ExecutorAddrRange &Other) { + return (Other.Start >= Start && Other.End <= End); + } bool overlaps(const ExecutorAddrRange &Other) { return !(Other.End <= Start || End <= Other.Start); } diff --git a/llvm/include/llvm/IR/IntrinsicsRISCVXsf.td b/llvm/include/llvm/IR/IntrinsicsRISCVXsf.td index bf20080..4a0272c 100644 --- a/llvm/include/llvm/IR/IntrinsicsRISCVXsf.td +++ b/llvm/include/llvm/IR/IntrinsicsRISCVXsf.td @@ -180,4 +180,98 @@ let TargetPrefix = "riscv" in { // XSfvfnrclipxfqf defm int_riscv_sf_vfnrclip_x_f_qf : RISCVSFCustomVFNRCLIP; defm int_riscv_sf_vfnrclip_xu_f_qf : RISCVSFCustomVFNRCLIP; + + // XSfmm + // Output: (output_len) + // Input: (input_len, vsew, twiden) + class RISCVSFVSet + : DefaultAttrsIntrinsic<[llvm_anyint_ty], + [LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>], + [ImmArg<ArgIndex<1>>, ImmArg<ArgIndex<2>>, IntrNoMem]>; + + // Input: (tss, base, tn) + class RISCVSFTileLoad + : DefaultAttrsIntrinsic<[], + [llvm_anyint_ty, llvm_ptr_ty, LLVMMatchType<0>], + [NoCapture<ArgIndex<1>>, IntrHasSideEffects]>, + RISCVVIntrinsic; + + // Input: (tss, base, tn) + class RISCVSFTileStore + : DefaultAttrsIntrinsic<[], + [llvm_anyint_ty, llvm_ptr_ty, LLVMMatchType<0>], + [NoCapture<ArgIndex<1>>, IntrWriteMem, + IntrHasSideEffects]>, + RISCVVIntrinsic; + + // Output: () + // Input: (mtd, mat1, mat2, tm, tn, tk, twiden) + class RISCVSFCustomMatMul<bit is_float = false> + : DefaultAttrsIntrinsic<[], [llvm_anyint_ty, llvm_anyvector_ty, + !if(is_float, LLVMMatchType<1>, + llvm_anyvector_ty), + LLVMMatchType<0>, LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>], + [IntrNoMem, IntrHasSideEffects, + ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<6>>]>, + RISCVVIntrinsic; + + def int_riscv_sf_vsettnt : RISCVSFVSet; + def int_riscv_sf_vsettm : RISCVSFVSet; + def int_riscv_sf_vsettk : RISCVSFVSet; + + def int_riscv_sf_vlte8 : RISCVSFTileLoad; + def int_riscv_sf_vlte16 : RISCVSFTileLoad; + def int_riscv_sf_vlte32 : RISCVSFTileLoad; + def int_riscv_sf_vlte64 : RISCVSFTileLoad; + def int_riscv_sf_vste8 : RISCVSFTileStore; + def int_riscv_sf_vste16 : RISCVSFTileStore; + def int_riscv_sf_vste32 : RISCVSFTileStore; + def int_riscv_sf_vste64 : RISCVSFTileStore; + + // Output: (vd) + // Input: (tss, tn) + def int_riscv_sf_vtmv_v_t + : DefaultAttrsIntrinsic<[llvm_anyvector_ty], + [llvm_anyint_ty, LLVMMatchType<1>], + [IntrNoMem, IntrHasSideEffects]>, + RISCVVIntrinsic { + let VLOperand = 2; + } + // Output: () + // Input: (tss, vs2, tn) + def int_riscv_sf_vtmv_t_v + : DefaultAttrsIntrinsic<[], [LLVMMatchType<1>, llvm_anyvector_ty, + llvm_anyint_ty], [IntrNoMem, IntrHasSideEffects]>, + RISCVVIntrinsic { + let VLOperand = 2; + } + + foreach a = ["u", "s"] in { + foreach b = ["u", "s"] in { + def int_riscv_sf_mm_ # a # _ # b : RISCVSFCustomMatMul; + } + } + + def int_riscv_sf_mm_f_f : RISCVSFCustomMatMul<true>; + foreach e1 = [5, 4] in + foreach e2 = [5, 4] in + def int_riscv_sf_mm_e # e1 # m # !sub(7, e1) # _e # e2 # m # !sub(7, e2) + : RISCVSFCustomMatMul<true>; + + // Output: () + // Input: (mtd) + def int_riscv_sf_vtzero_t + : DefaultAttrsIntrinsic<[], + [llvm_anyint_ty, LLVMMatchType<0>,LLVMMatchType<0>, + LLVMMatchType<0>, LLVMMatchType<0>], + [ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<3>>, + ImmArg<ArgIndex<4>>, IntrNoMem, IntrHasSideEffects]>, + RISCVVIntrinsic; + + // Output: () + // Input: () + def int_riscv_sf_vtdiscard + : DefaultAttrsIntrinsic<[], [], [IntrNoMem, IntrHasSideEffects]>, + RISCVVIntrinsic; } // TargetPrefix = "riscv" diff --git a/llvm/lib/Analysis/DependenceAnalysis.cpp b/llvm/lib/Analysis/DependenceAnalysis.cpp index 8d20b0e..805b682 100644 --- a/llvm/lib/Analysis/DependenceAnalysis.cpp +++ b/llvm/lib/Analysis/DependenceAnalysis.cpp @@ -1180,32 +1180,41 @@ bool DependenceInfo::isKnownLessThan(const SCEV *S, const SCEV *Size) const { S = SE->getTruncateOrZeroExtend(S, MaxType); Size = SE->getTruncateOrZeroExtend(Size, MaxType); - // Special check for addrecs using BE taken count - if (const SCEVAddRecExpr *AddRec = dyn_cast<SCEVAddRecExpr>(S)) - if (AddRec->isAffine() && AddRec->hasNoSignedWrap()) { - const SCEV *BECount = SE->getBackedgeTakenCount(AddRec->getLoop()); - const SCEV *Start = AddRec->getStart(); - const SCEV *Step = AddRec->getStepRecurrence(*SE); - const SCEV *End = AddRec->evaluateAtIteration(BECount, *SE); - const SCEV *Diff0 = SE->getMinusSCEV(Start, Size); - const SCEV *Diff1 = SE->getMinusSCEV(End, Size); - - // If the value of Step is non-negative and the AddRec is non-wrap, it - // reaches its maximum at the last iteration. So it's enouth to check - // whether End - Size is negative. - if (SE->isKnownNonNegative(Step) && SE->isKnownNegative(Diff1)) - return true; + auto CheckAddRecBECount = [&]() { + const SCEVAddRecExpr *AddRec = dyn_cast<SCEVAddRecExpr>(S); + if (!AddRec || !AddRec->isAffine() || !AddRec->hasNoSignedWrap()) + return false; + const SCEV *BECount = collectUpperBound(AddRec->getLoop(), MaxType); + // If the BTC cannot be computed, check the base case for S. + if (!BECount || isa<SCEVCouldNotCompute>(BECount)) + return false; + const SCEV *Start = AddRec->getStart(); + const SCEV *Step = AddRec->getStepRecurrence(*SE); + const SCEV *End = AddRec->evaluateAtIteration(BECount, *SE); + const SCEV *Diff0 = SE->getMinusSCEV(Start, Size); + const SCEV *Diff1 = SE->getMinusSCEV(End, Size); + + // If the value of Step is non-negative and the AddRec is non-wrap, it + // reaches its maximum at the last iteration. So it's enouth to check + // whether End - Size is negative. + if (SE->isKnownNonNegative(Step) && SE->isKnownNegative(Diff1)) + return true; - // If the value of Step is non-positive and the AddRec is non-wrap, the - // initial value is its maximum. - if (SE->isKnownNonPositive(Step) && SE->isKnownNegative(Diff0)) - return true; + // If the value of Step is non-positive and the AddRec is non-wrap, the + // initial value is its maximum. + if (SE->isKnownNonPositive(Step) && SE->isKnownNegative(Diff0)) + return true; - // Even if we don't know the sign of Step, either Start or End must be - // the maximum value of the AddRec since it is non-wrap. - if (SE->isKnownNegative(Diff0) && SE->isKnownNegative(Diff1)) - return true; - } + // Even if we don't know the sign of Step, either Start or End must be + // the maximum value of the AddRec since it is non-wrap. + if (SE->isKnownNegative(Diff0) && SE->isKnownNegative(Diff1)) + return true; + + return false; + }; + + if (CheckAddRecBECount()) + return true; // Check using normal isKnownNegative const SCEV *LimitedBound = SE->getMinusSCEV(S, Size); diff --git a/llvm/lib/Analysis/IVDescriptors.cpp b/llvm/lib/Analysis/IVDescriptors.cpp index b8c540c..9f8ac6e 100644 --- a/llvm/lib/Analysis/IVDescriptors.cpp +++ b/llvm/lib/Analysis/IVDescriptors.cpp @@ -849,17 +849,12 @@ RecurrenceDescriptor::isMinMaxPattern(Instruction *I, RecurKind Kind, /// %sum.2 = select %cmp, %add, %sum.1 RecurrenceDescriptor::InstDesc RecurrenceDescriptor::isConditionalRdxPattern(Instruction *I) { - SelectInst *SI = dyn_cast<SelectInst>(I); - if (!SI) - return InstDesc(false, I); - - CmpInst *CI = dyn_cast<CmpInst>(SI->getCondition()); + Value *TrueVal, *FalseVal; // Only handle single use cases for now. - if (!CI || !CI->hasOneUse()) + if (!match(I, + m_Select(m_OneUse(m_Cmp()), m_Value(TrueVal), m_Value(FalseVal)))) return InstDesc(false, I); - Value *TrueVal = SI->getTrueValue(); - Value *FalseVal = SI->getFalseValue(); // Handle only when either of operands of select instruction is a PHI // node for now. if ((isa<PHINode>(TrueVal) && isa<PHINode>(FalseVal)) || @@ -886,7 +881,7 @@ RecurrenceDescriptor::isConditionalRdxPattern(Instruction *I) { if (!IPhi || IPhi != FalseVal) return InstDesc(false, I); - return InstDesc(true, SI); + return InstDesc(true, I); } RecurrenceDescriptor::InstDesc RecurrenceDescriptor::isRecurrenceInstr( diff --git a/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp b/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp index 3812823..9b9a6a2 100644 --- a/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp +++ b/llvm/lib/CodeGen/GlobalISel/GISelValueTracking.cpp @@ -2013,6 +2013,43 @@ unsigned GISelValueTracking::computeNumSignBits(Register R, 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/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp index e153842..787a81a 100644 --- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp @@ -5044,7 +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 @@ -5076,7 +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(); diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp index c9aeef7..876066e9 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); diff --git a/llvm/lib/IR/ConstantFold.cpp b/llvm/lib/IR/ConstantFold.cpp index 3842b1a..6a9ef2e 100644 --- a/llvm/lib/IR/ConstantFold.cpp +++ b/llvm/lib/IR/ConstantFold.cpp @@ -741,7 +741,8 @@ Constant *llvm::ConstantFoldBinaryInstruction(unsigned Opcode, Constant *C1, assert(!CI2->isZero() && "And zero handled above"); if (ConstantExpr *CE1 = dyn_cast<ConstantExpr>(C1)) { // If and'ing the address of a global with a constant, fold it. - if (CE1->getOpcode() == Instruction::PtrToInt && + if ((CE1->getOpcode() == Instruction::PtrToInt || + CE1->getOpcode() == Instruction::PtrToAddr) && isa<GlobalValue>(CE1->getOperand(0))) { GlobalValue *GV = cast<GlobalValue>(CE1->getOperand(0)); diff --git a/llvm/lib/Target/AArch64/Utils/AArch64SMEAttributes.cpp b/llvm/lib/Target/AArch64/Utils/AArch64SMEAttributes.cpp index dd6fa16..d71f728 100644 --- a/llvm/lib/Target/AArch64/Utils/AArch64SMEAttributes.cpp +++ b/llvm/lib/Target/AArch64/Utils/AArch64SMEAttributes.cpp @@ -130,6 +130,12 @@ SMECallAttrs::SMECallAttrs(const CallBase &CB, const AArch64TargetLowering *TLI) if (auto *CalledFunction = CB.getCalledFunction()) CalledFn = SMEAttrs(*CalledFunction, TLI); + // An `invoke` of an agnostic ZA function may not return normally (it may + // resume in an exception block). In this case, it acts like a private ZA + // callee and may require a ZA save to be set up before it is called. + if (isa<InvokeInst>(CB)) + CalledFn.set(SMEAttrs::ZA_State_Agnostic, /*Enable=*/false); + // FIXME: We probably should not allow SME attributes on direct calls but // clang duplicates streaming mode attributes at each callsite. assert((IsIndirect || diff --git a/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp b/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp index 66ca436..de433e4 100644 --- a/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp +++ b/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp @@ -1100,6 +1100,12 @@ static bool lowerRISCVVMachineInstrToMCInst(const MachineInstr *MI, --NumOps; if (RISCVII::hasRoundModeOp(TSFlags)) --NumOps; + if (RISCVII::hasTWidenOp(TSFlags)) + --NumOps; + if (RISCVII::hasTMOp(TSFlags)) + --NumOps; + if (RISCVII::hasTKOp(TSFlags)) + --NumOps; bool hasVLOutput = RISCVInstrInfo::isFaultOnlyFirstLoad(*MI); for (unsigned OpNo = 0; OpNo != NumOps; ++OpNo) { diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp index 974252a..9a6afa1 100644 --- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp +++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.cpp @@ -516,6 +516,44 @@ void RISCVDAGToDAGISel::selectVSETVLI(SDNode *Node) { CurDAG->getMachineNode(Opcode, DL, XLenVT, VLOperand, VTypeIOp)); } +void RISCVDAGToDAGISel::selectXSfmmVSET(SDNode *Node) { + if (!Subtarget->hasVendorXSfmmbase()) + return; + + assert(Node->getOpcode() == ISD::INTRINSIC_WO_CHAIN && "Unexpected opcode"); + + SDLoc DL(Node); + MVT XLenVT = Subtarget->getXLenVT(); + + unsigned IntNo = Node->getConstantOperandVal(0); + + assert((IntNo == Intrinsic::riscv_sf_vsettnt || + IntNo == Intrinsic::riscv_sf_vsettm || + IntNo == Intrinsic::riscv_sf_vsettk) && + "Unexpected XSfmm vset intrinsic"); + + unsigned SEW = RISCVVType::decodeVSEW(Node->getConstantOperandVal(2)); + unsigned Widen = RISCVVType::decodeTWiden(Node->getConstantOperandVal(3)); + unsigned PseudoOpCode = + IntNo == Intrinsic::riscv_sf_vsettnt ? RISCV::PseudoSF_VSETTNT + : IntNo == Intrinsic::riscv_sf_vsettm ? RISCV::PseudoSF_VSETTM + : RISCV::PseudoSF_VSETTK; + + if (IntNo == Intrinsic::riscv_sf_vsettnt) { + unsigned VTypeI = RISCVVType::encodeXSfmmVType(SEW, Widen, 0); + SDValue VTypeIOp = CurDAG->getTargetConstant(VTypeI, DL, XLenVT); + + ReplaceNode(Node, CurDAG->getMachineNode(PseudoOpCode, DL, XLenVT, + Node->getOperand(1), VTypeIOp)); + } else { + SDValue Log2SEW = CurDAG->getTargetConstant(Log2_32(SEW), DL, XLenVT); + SDValue TWiden = CurDAG->getTargetConstant(Widen, DL, XLenVT); + ReplaceNode(Node, + CurDAG->getMachineNode(PseudoOpCode, DL, XLenVT, + Node->getOperand(1), Log2SEW, TWiden)); + } +} + bool RISCVDAGToDAGISel::tryShrinkShlLogicImm(SDNode *Node) { MVT VT = Node->getSimpleValueType(0); unsigned Opcode = Node->getOpcode(); @@ -847,6 +885,11 @@ bool RISCVDAGToDAGISel::tryIndexedLoad(SDNode *Node) { return true; } +static Register getTileReg(uint64_t TileNum) { + assert(TileNum <= 15 && "Invalid tile number"); + return RISCV::T0 + TileNum; +} + void RISCVDAGToDAGISel::selectSF_VC_X_SE(SDNode *Node) { if (!Subtarget->hasVInstructions()) return; @@ -2035,6 +2078,10 @@ void RISCVDAGToDAGISel::Select(SDNode *Node) { case Intrinsic::riscv_vsetvli: case Intrinsic::riscv_vsetvlimax: return selectVSETVLI(Node); + case Intrinsic::riscv_sf_vsettnt: + case Intrinsic::riscv_sf_vsettm: + case Intrinsic::riscv_sf_vsettk: + return selectXSfmmVSET(Node); } break; } @@ -2458,6 +2505,142 @@ void RISCVDAGToDAGISel::Select(SDNode *Node) { case Intrinsic::riscv_sf_vc_i_se: selectSF_VC_X_SE(Node); return; + case Intrinsic::riscv_sf_vlte8: + case Intrinsic::riscv_sf_vlte16: + case Intrinsic::riscv_sf_vlte32: + case Intrinsic::riscv_sf_vlte64: { + unsigned Log2SEW; + unsigned PseudoInst; + switch (IntNo) { + case Intrinsic::riscv_sf_vlte8: + PseudoInst = RISCV::PseudoSF_VLTE8; + Log2SEW = 3; + break; + case Intrinsic::riscv_sf_vlte16: + PseudoInst = RISCV::PseudoSF_VLTE16; + Log2SEW = 4; + break; + case Intrinsic::riscv_sf_vlte32: + PseudoInst = RISCV::PseudoSF_VLTE32; + Log2SEW = 5; + break; + case Intrinsic::riscv_sf_vlte64: + PseudoInst = RISCV::PseudoSF_VLTE64; + Log2SEW = 6; + break; + } + + SDValue SEWOp = CurDAG->getTargetConstant(Log2SEW, DL, XLenVT); + SDValue TWidenOp = CurDAG->getTargetConstant(1, DL, XLenVT); + SDValue Operands[] = {Node->getOperand(2), + Node->getOperand(3), + Node->getOperand(4), + SEWOp, + TWidenOp, + Node->getOperand(0)}; + + MachineSDNode *TileLoad = + CurDAG->getMachineNode(PseudoInst, DL, Node->getVTList(), Operands); + if (auto *MemOp = dyn_cast<MemSDNode>(Node)) + CurDAG->setNodeMemRefs(TileLoad, {MemOp->getMemOperand()}); + + ReplaceNode(Node, TileLoad); + return; + } + case Intrinsic::riscv_sf_mm_s_s: + case Intrinsic::riscv_sf_mm_s_u: + case Intrinsic::riscv_sf_mm_u_s: + case Intrinsic::riscv_sf_mm_u_u: + case Intrinsic::riscv_sf_mm_e5m2_e5m2: + case Intrinsic::riscv_sf_mm_e5m2_e4m3: + case Intrinsic::riscv_sf_mm_e4m3_e5m2: + case Intrinsic::riscv_sf_mm_e4m3_e4m3: + case Intrinsic::riscv_sf_mm_f_f: { + bool HasFRM = false; + unsigned PseudoInst; + switch (IntNo) { + case Intrinsic::riscv_sf_mm_s_s: + PseudoInst = RISCV::PseudoSF_MM_S_S; + break; + case Intrinsic::riscv_sf_mm_s_u: + PseudoInst = RISCV::PseudoSF_MM_S_U; + break; + case Intrinsic::riscv_sf_mm_u_s: + PseudoInst = RISCV::PseudoSF_MM_U_S; + break; + case Intrinsic::riscv_sf_mm_u_u: + PseudoInst = RISCV::PseudoSF_MM_U_U; + break; + case Intrinsic::riscv_sf_mm_e5m2_e5m2: + PseudoInst = RISCV::PseudoSF_MM_E5M2_E5M2; + HasFRM = true; + break; + case Intrinsic::riscv_sf_mm_e5m2_e4m3: + PseudoInst = RISCV::PseudoSF_MM_E5M2_E4M3; + HasFRM = true; + break; + case Intrinsic::riscv_sf_mm_e4m3_e5m2: + PseudoInst = RISCV::PseudoSF_MM_E4M3_E5M2; + HasFRM = true; + break; + case Intrinsic::riscv_sf_mm_e4m3_e4m3: + PseudoInst = RISCV::PseudoSF_MM_E4M3_E4M3; + HasFRM = true; + break; + case Intrinsic::riscv_sf_mm_f_f: + if (Node->getOperand(3).getValueType().getScalarType() == MVT::bf16) + PseudoInst = RISCV::PseudoSF_MM_F_F_ALT; + else + PseudoInst = RISCV::PseudoSF_MM_F_F; + HasFRM = true; + break; + } + uint64_t TileNum = Node->getConstantOperandVal(2); + SDValue Op1 = Node->getOperand(3); + SDValue Op2 = Node->getOperand(4); + MVT VT = Op1->getSimpleValueType(0); + unsigned Log2SEW = Log2_32(VT.getScalarSizeInBits()); + SDValue TmOp = Node->getOperand(5); + SDValue TnOp = Node->getOperand(6); + SDValue TkOp = Node->getOperand(7); + SDValue TWidenOp = Node->getOperand(8); + SDValue Chain = Node->getOperand(0); + + // sf.mm.f.f with sew=32, twiden=2 is invalid + if (IntNo == Intrinsic::riscv_sf_mm_f_f && Log2SEW == 5 && + TWidenOp->getAsZExtVal() == 2) + reportFatalUsageError("sf.mm.f.f doesn't support (sew=32, twiden=2)"); + + SmallVector<SDValue, 10> Operands( + {CurDAG->getRegister(getTileReg(TileNum), XLenVT), Op1, Op2}); + if (HasFRM) + Operands.push_back( + CurDAG->getTargetConstant(RISCVFPRndMode::DYN, DL, XLenVT)); + Operands.append({TmOp, TnOp, TkOp, + CurDAG->getTargetConstant(Log2SEW, DL, XLenVT), TWidenOp, + Chain}); + + auto *NewNode = + CurDAG->getMachineNode(PseudoInst, DL, Node->getVTList(), Operands); + + ReplaceNode(Node, NewNode); + return; + } + case Intrinsic::riscv_sf_vtzero_t: { + uint64_t TileNum = Node->getConstantOperandVal(2); + SDValue Tm = Node->getOperand(3); + SDValue Tn = Node->getOperand(4); + SDValue Log2SEW = Node->getOperand(5); + SDValue TWiden = Node->getOperand(6); + SDValue Chain = Node->getOperand(0); + auto *NewNode = CurDAG->getMachineNode( + RISCV::PseudoSF_VTZERO_T, DL, Node->getVTList(), + {CurDAG->getRegister(getTileReg(TileNum), XLenVT), Tm, Tn, Log2SEW, + TWiden, Chain}); + + ReplaceNode(Node, NewNode); + return; + } } break; } diff --git a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h index f03b44c..19ee103 100644 --- a/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h +++ b/llvm/lib/Target/RISCV/RISCVISelDAGToDAG.h @@ -165,6 +165,7 @@ public: void selectVSXSEG(SDNode *Node, unsigned NF, bool IsMasked, bool IsOrdered); void selectVSETVLI(SDNode *Node); + void selectXSfmmVSET(SDNode *Node); void selectSF_VC_X_SE(SDNode *Node); diff --git a/llvm/lib/Target/RISCV/RISCVInstrFormats.td b/llvm/lib/Target/RISCV/RISCVInstrFormats.td index 5b06303..fee1d15 100644 --- a/llvm/lib/Target/RISCV/RISCVInstrFormats.td +++ b/llvm/lib/Target/RISCV/RISCVInstrFormats.td @@ -177,7 +177,7 @@ def EltDepsVL : EltDeps<vl=1, mask=0>; def EltDepsMask : EltDeps<vl=0, mask=1>; def EltDepsVLMask : EltDeps<vl=1, mask=1>; -class EEW <bits<2> val> { +class EEW<bits<2> val> { bits<2> Value = val; } def EEW1 : EEW<0>; @@ -185,6 +185,13 @@ def EEWSEWx1 : EEW<1>; def EEWSEWx2 : EEW<2>; def EEWSEWx4 : EEW<3>; +class AltFmtType<bits<2> val> { + bits<2> Value = val; +} +def DONT_CARE_ALTFMT : AltFmtType<0>; +def IS_NOT_ALTFMT : AltFmtType<1>; +def IS_ALTFMT : AltFmtType<2>; + class RVInstCommon<dag outs, dag ins, string opcodestr, string argstr, list<dag> pattern, InstFormat format> : Instruction { let Namespace = "RISCV"; @@ -271,8 +278,8 @@ class RVInstCommon<dag outs, dag ins, string opcodestr, string argstr, // 0 -> Don't care about altfmt bit in VTYPE. // 1 -> Is not altfmt. // 2 -> Is altfmt(BF16). - bits<2> AltFmtType = 0; - let TSFlags{28-27} = AltFmtType; + AltFmtType AltFmtType = DONT_CARE_ALTFMT; + let TSFlags{28-27} = AltFmtType.Value; // XSfmmbase bit HasTWidenOp = 0; diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td index c1b23af..65865ce 100644 --- a/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td +++ b/llvm/lib/Target/RISCV/RISCVInstrInfoVPseudos.td @@ -128,9 +128,6 @@ defvar TAIL_AGNOSTIC = 1; defvar TU_MU = 0; defvar TA_MU = 1; defvar TA_MA = 3; -defvar DONT_CARE_ALTFMT = 0; -defvar IS_NOT_ALTFMT = 1; -defvar IS_ALTFMT = 2; //===----------------------------------------------------------------------===// // Utilities. diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoXSfmm.td b/llvm/lib/Target/RISCV/RISCVInstrInfoXSfmm.td index 5ad22e6b..d77a44a 100644 --- a/llvm/lib/Target/RISCV/RISCVInstrInfoXSfmm.td +++ b/llvm/lib/Target/RISCV/RISCVInstrInfoXSfmm.td @@ -418,3 +418,54 @@ let hasSideEffects = 1, mayLoad = 0, mayStore = 0 in { ixlenimm:$twiden)>; def PseudoSF_VTDISCARD : RISCVVPseudo<(outs), (ins), []>; } + +class VPatXSfmmTileStore<string intrinsic_name, + string inst_name, + int log2sew> : + Pat<(!cast<Intrinsic>(intrinsic_name) + (XLenVT GPR:$rs2), + (XLenVT GPR:$rs1), + (XLenVT AVL:$tn)), + (!cast<Instruction>(inst_name) + (XLenVT GPR:$rs2), + (XLenVT GPR:$rs1), + GPR:$tn, log2sew, 1)>; + +class VPatXSfmmTileMove_T_V<string intrinsic_name, + string inst_name, + ValueType reg_type, + int log2sew> : + Pat<(!cast<Intrinsic>(intrinsic_name) + (XLenVT GPR:$rs1), + (reg_type VRM8:$vs2), + (XLenVT AVL:$atn)), + (!cast<Instruction>(inst_name) + (XLenVT GPR:$rs1), + (reg_type VRM8:$vs2), + GPR:$atn, log2sew, 1)>; + +class VPatXSfmmTileMove_V_T<string intrinsic_name, + string inst_name, + ValueType result_type, + int log2sew> : + Pat<(result_type (!cast<Intrinsic>(intrinsic_name) + (XLenVT GPR:$rs1), + (XLenVT AVL:$atn))), + (!cast<Instruction>(inst_name) + (XLenVT GPR:$rs1), + GPR:$atn, log2sew, 1)>; + +class VPatXSfmmVTDiscard<string intrinsic_name, + string inst_name> : + Pat<(!cast<Intrinsic>(intrinsic_name)), + (!cast<Instruction>(inst_name))>; + +foreach eew = [8, 16, 32, 64] in + def : VPatXSfmmTileStore<"int_riscv_sf_vste" # eew, "PseudoSF_VSTE" # eew, !logtwo(eew)>; + +foreach vti = [VI8M8, VI16M8, VI32M8, VI64M8, VF16M8, VF32M8, VF64M8, VBF16M8] in { + def : VPatXSfmmTileMove_T_V<"int_riscv_sf_vtmv_t_v", "PseudoSF_VTMV_T_V", vti.Vector, vti.Log2SEW>; + def : VPatXSfmmTileMove_V_T<"int_riscv_sf_vtmv_v_t", "PseudoSF_VTMV_V_T", vti.Vector, vti.Log2SEW>; +} + +def : VPatXSfmmVTDiscard<"int_riscv_sf_vtdiscard", "PseudoSF_VTDISCARD">; diff --git a/llvm/lib/Target/SPIRV/SPIRVAsmPrinter.cpp b/llvm/lib/Target/SPIRV/SPIRVAsmPrinter.cpp index b765fec..56a6168 100644 --- a/llvm/lib/Target/SPIRV/SPIRVAsmPrinter.cpp +++ b/llvm/lib/Target/SPIRV/SPIRVAsmPrinter.cpp @@ -139,8 +139,8 @@ void SPIRVAsmPrinter::emitEndOfAsmFile(Module &M) { // anymore. void SPIRVAsmPrinter::cleanUp(Module &M) { // Verifier disallows uses of intrinsic global variables. - for (StringRef GVName : {"llvm.global_ctors", "llvm.global_dtors", - "llvm.used", "llvm.compiler.used"}) { + for (StringRef GVName : + {"llvm.global_ctors", "llvm.global_dtors", "llvm.used"}) { if (GlobalVariable *GV = M.getNamedGlobal(GVName)) GV->setName(""); } diff --git a/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp b/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp index e16c8f0..c6c6182 100644 --- a/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp +++ b/llvm/lib/Target/SPIRV/SPIRVEmitIntrinsics.cpp @@ -17,6 +17,7 @@ #include "SPIRVTargetMachine.h" #include "SPIRVUtils.h" #include "llvm/ADT/DenseSet.h" +#include "llvm/ADT/StringSet.h" #include "llvm/IR/IRBuilder.h" #include "llvm/IR/InstIterator.h" #include "llvm/IR/InstVisitor.h" @@ -2028,9 +2029,13 @@ Instruction *SPIRVEmitIntrinsics::visitUnreachableInst(UnreachableInst &I) { void SPIRVEmitIntrinsics::processGlobalValue(GlobalVariable &GV, IRBuilder<> &B) { - // Skip special artifical variable llvm.global.annotations. - if (GV.getName() == "llvm.global.annotations") + // Skip special artificial variables. + static const StringSet<> ArtificialGlobals{"llvm.global.annotations", + "llvm.compiler.used"}; + + if (ArtificialGlobals.contains(GV.getName())) return; + Constant *Init = nullptr; if (hasInitializer(&GV)) { // Deduce element type and store results in Global Registry. diff --git a/llvm/lib/Target/SystemZ/SystemZ.h b/llvm/lib/Target/SystemZ/SystemZ.h index a0cf881..5a06ea3 100644 --- a/llvm/lib/Target/SystemZ/SystemZ.h +++ b/llvm/lib/Target/SystemZ/SystemZ.h @@ -24,6 +24,7 @@ class SystemZTargetMachine; namespace SystemZ { // Condition-code mask values. +const unsigned CCMASK_NONE = 0; const unsigned CCMASK_0 = 1 << 3; const unsigned CCMASK_1 = 1 << 2; const unsigned CCMASK_2 = 1 << 1; diff --git a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp index 3b7d11a..de28faf 100644 --- a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp +++ b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp @@ -15,6 +15,7 @@ #include "SystemZConstantPoolValue.h" #include "SystemZMachineFunctionInfo.h" #include "SystemZTargetMachine.h" +#include "llvm/ADT/SmallSet.h" #include "llvm/CodeGen/CallingConvLower.h" #include "llvm/CodeGen/ISDOpcodes.h" #include "llvm/CodeGen/MachineInstrBuilder.h" @@ -24,6 +25,7 @@ #include "llvm/IR/IntrinsicInst.h" #include "llvm/IR/Intrinsics.h" #include "llvm/IR/IntrinsicsS390.h" +#include "llvm/IR/PatternMatch.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/ErrorHandling.h" #include "llvm/Support/KnownBits.h" @@ -1514,6 +1516,9 @@ SystemZTargetLowering::getConstraintType(StringRef Constraint) const { default: break; } + } else if (Constraint.size() == 5 && Constraint.starts_with("{")) { + if (StringRef("{@cc}").compare(Constraint) == 0) + return C_Other; } return TargetLowering::getConstraintType(Constraint); } @@ -1707,6 +1712,10 @@ SystemZTargetLowering::getRegForInlineAsmConstraint( return parseRegisterNumber(Constraint, &SystemZ::VR128BitRegClass, SystemZMC::VR128Regs, 32); } + if (Constraint[1] == '@') { + if (StringRef("{@cc}").compare(Constraint) == 0) + return std::make_pair(0u, &SystemZ::GR32BitRegClass); + } } return TargetLowering::getRegForInlineAsmConstraint(TRI, Constraint, VT); } @@ -1737,6 +1746,38 @@ Register SystemZTargetLowering::getExceptionSelectorRegister( return Subtarget.isTargetXPLINK64() ? SystemZ::R2D : SystemZ::R7D; } +// Convert condition code in CCReg to an i32 value. +static SDValue getCCResult(SelectionDAG &DAG, SDValue CCReg) { + SDLoc DL(CCReg); + SDValue IPM = DAG.getNode(SystemZISD::IPM, DL, MVT::i32, CCReg); + return DAG.getNode(ISD::SRL, DL, MVT::i32, IPM, + DAG.getConstant(SystemZ::IPM_CC, DL, MVT::i32)); +} + +// Lower @cc targets via setcc. +SDValue SystemZTargetLowering::LowerAsmOutputForConstraint( + SDValue &Chain, SDValue &Glue, const SDLoc &DL, + const AsmOperandInfo &OpInfo, SelectionDAG &DAG) const { + if (StringRef("{@cc}").compare(OpInfo.ConstraintCode) != 0) + return SDValue(); + + // Check that return type is valid. + if (OpInfo.ConstraintVT.isVector() || !OpInfo.ConstraintVT.isInteger() || + OpInfo.ConstraintVT.getSizeInBits() < 8) + report_fatal_error("Glue output operand is of invalid type"); + + MachineFunction &MF = DAG.getMachineFunction(); + MachineRegisterInfo &MRI = MF.getRegInfo(); + MRI.addLiveIn(SystemZ::CC); + + if (Glue.getNode()) { + Glue = DAG.getCopyFromReg(Chain, DL, SystemZ::CC, MVT::i32, Glue); + Chain = Glue.getValue(1); + } else + Glue = DAG.getCopyFromReg(Chain, DL, SystemZ::CC, MVT::i32); + return getCCResult(DAG, Glue); +} + void SystemZTargetLowering::LowerAsmOperandForConstraint( SDValue Op, StringRef Constraint, std::vector<SDValue> &Ops, SelectionDAG &DAG) const { @@ -5300,14 +5341,6 @@ SDValue SystemZTargetLowering::lowerPREFETCH(SDValue Op, Node->getMemoryVT(), Node->getMemOperand()); } -// Convert condition code in CCReg to an i32 value. -static SDValue getCCResult(SelectionDAG &DAG, SDValue CCReg) { - SDLoc DL(CCReg); - SDValue IPM = DAG.getNode(SystemZISD::IPM, DL, MVT::i32, CCReg); - return DAG.getNode(ISD::SRL, DL, MVT::i32, IPM, - DAG.getConstant(SystemZ::IPM_CC, DL, MVT::i32)); -} - SDValue SystemZTargetLowering::lowerINTRINSIC_W_CHAIN(SDValue Op, SelectionDAG &DAG) const { @@ -8723,95 +8756,247 @@ SDValue SystemZTargetLowering::combineSETCC( return SDValue(); } -static bool combineCCMask(SDValue &CCReg, int &CCValid, int &CCMask) { +static std::pair<SDValue, int> findCCUse(const SDValue &Val) { + switch (Val.getOpcode()) { + default: + return std::make_pair(SDValue(), SystemZ::CCMASK_NONE); + case SystemZISD::IPM: + if (Val.getOperand(0).getOpcode() == SystemZISD::CLC || + Val.getOperand(0).getOpcode() == SystemZISD::STRCMP) + return std::make_pair(Val.getOperand(0), SystemZ::CCMASK_ICMP); + return std::make_pair(Val.getOperand(0), SystemZ::CCMASK_ANY); + case SystemZISD::SELECT_CCMASK: { + SDValue Op4CCReg = Val.getOperand(4); + if (Op4CCReg.getOpcode() == SystemZISD::ICMP || + Op4CCReg.getOpcode() == SystemZISD::TM) { + auto [OpCC, OpCCValid] = findCCUse(Op4CCReg.getOperand(0)); + if (OpCC != SDValue()) + return std::make_pair(OpCC, OpCCValid); + } + auto *CCValid = dyn_cast<ConstantSDNode>(Val.getOperand(2)); + if (!CCValid) + return std::make_pair(SDValue(), SystemZ::CCMASK_NONE); + int CCValidVal = CCValid->getZExtValue(); + return std::make_pair(Op4CCReg, CCValidVal); + } + case ISD::ADD: + case ISD::AND: + case ISD::OR: + case ISD::XOR: + case ISD::SHL: + case ISD::SRA: + case ISD::SRL: + auto [Op0CC, Op0CCValid] = findCCUse(Val.getOperand(0)); + if (Op0CC != SDValue()) + return std::make_pair(Op0CC, Op0CCValid); + return findCCUse(Val.getOperand(1)); + } +} + +static bool combineCCMask(SDValue &CCReg, int &CCValid, int &CCMask, + SelectionDAG &DAG); + +SmallVector<SDValue, 4> static simplifyAssumingCCVal(SDValue &Val, SDValue &CC, + SelectionDAG &DAG) { + SDLoc DL(Val); + auto Opcode = Val.getOpcode(); + switch (Opcode) { + default: + return {}; + case ISD::Constant: + return {Val, Val, Val, Val}; + case SystemZISD::IPM: { + SDValue IPMOp0 = Val.getOperand(0); + if (IPMOp0 != CC) + return {}; + SmallVector<SDValue, 4> ShiftedCCVals; + for (auto CC : {0, 1, 2, 3}) + ShiftedCCVals.emplace_back( + DAG.getConstant((CC << SystemZ::IPM_CC), DL, MVT::i32)); + return ShiftedCCVals; + } + case SystemZISD::SELECT_CCMASK: { + SDValue TrueVal = Val.getOperand(0), FalseVal = Val.getOperand(1); + auto *CCValid = dyn_cast<ConstantSDNode>(Val.getOperand(2)); + auto *CCMask = dyn_cast<ConstantSDNode>(Val.getOperand(3)); + if (!CCValid || !CCMask) + return {}; + + int CCValidVal = CCValid->getZExtValue(); + int CCMaskVal = CCMask->getZExtValue(); + const auto &&TrueSDVals = simplifyAssumingCCVal(TrueVal, CC, DAG); + const auto &&FalseSDVals = simplifyAssumingCCVal(FalseVal, CC, DAG); + if (TrueSDVals.empty() || FalseSDVals.empty()) + return {}; + SDValue Op4CCReg = Val.getOperand(4); + if (Op4CCReg != CC) + combineCCMask(Op4CCReg, CCValidVal, CCMaskVal, DAG); + if (Op4CCReg != CC) + return {}; + SmallVector<SDValue, 4> MergedSDVals; + for (auto &CCVal : {0, 1, 2, 3}) + MergedSDVals.emplace_back(((CCMaskVal & (1 << (3 - CCVal))) != 0) + ? TrueSDVals[CCVal] + : FalseSDVals[CCVal]); + return MergedSDVals; + } + case ISD::ADD: + case ISD::AND: + case ISD::OR: + case ISD::XOR: + case ISD::SRA: + // Avoid introducing CC spills (because ADD/AND/OR/XOR/SRA + // would clobber CC). + if (!Val.hasOneUse()) + return {}; + [[fallthrough]]; + case ISD::SHL: + case ISD::SRL: + SDValue Op0 = Val.getOperand(0), Op1 = Val.getOperand(1); + const auto &&Op0SDVals = simplifyAssumingCCVal(Op0, CC, DAG); + const auto &&Op1SDVals = simplifyAssumingCCVal(Op1, CC, DAG); + if (Op0SDVals.empty() || Op1SDVals.empty()) + return {}; + SmallVector<SDValue, 4> BinaryOpSDVals; + for (auto CCVal : {0, 1, 2, 3}) + BinaryOpSDVals.emplace_back(DAG.getNode( + Opcode, DL, Val.getValueType(), Op0SDVals[CCVal], Op1SDVals[CCVal])); + return BinaryOpSDVals; + } +} + +static bool combineCCMask(SDValue &CCReg, int &CCValid, int &CCMask, + SelectionDAG &DAG) { // We have a SELECT_CCMASK or BR_CCMASK comparing the condition code // set by the CCReg instruction using the CCValid / CCMask masks, - // If the CCReg instruction is itself a ICMP testing the condition + // If the CCReg instruction is itself a ICMP / TM testing the condition // code set by some other instruction, see whether we can directly // use that condition code. - - // Verify that we have an ICMP against some constant. - if (CCValid != SystemZ::CCMASK_ICMP) - return false; - auto *ICmp = CCReg.getNode(); - if (ICmp->getOpcode() != SystemZISD::ICMP) - return false; - auto *CompareLHS = ICmp->getOperand(0).getNode(); - auto *CompareRHS = dyn_cast<ConstantSDNode>(ICmp->getOperand(1)); - if (!CompareRHS) + auto *CCNode = CCReg.getNode(); + if (!CCNode) return false; - // Optimize the case where CompareLHS is a SELECT_CCMASK. - if (CompareLHS->getOpcode() == SystemZISD::SELECT_CCMASK) { - // Verify that we have an appropriate mask for a EQ or NE comparison. - bool Invert = false; - if (CCMask == SystemZ::CCMASK_CMP_NE) - Invert = !Invert; - else if (CCMask != SystemZ::CCMASK_CMP_EQ) + if (CCNode->getOpcode() == SystemZISD::TM) { + if (CCValid != SystemZ::CCMASK_TM) return false; - - // Verify that the ICMP compares against one of select values. - auto *TrueVal = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(0)); - if (!TrueVal) - return false; - auto *FalseVal = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(1)); - if (!FalseVal) + auto emulateTMCCMask = [](const SDValue &Op0Val, const SDValue &Op1Val) { + auto *Op0Node = dyn_cast<ConstantSDNode>(Op0Val.getNode()); + auto *Op1Node = dyn_cast<ConstantSDNode>(Op1Val.getNode()); + if (!Op0Node || !Op1Node) + return -1; + auto Op0APVal = Op0Node->getAPIntValue(); + auto Op1APVal = Op1Node->getAPIntValue(); + auto Result = Op0APVal & Op1APVal; + bool AllOnes = Result == Op1APVal; + bool AllZeros = Result == 0; + bool IsLeftMostBitSet = Result[Op1APVal.getActiveBits()] != 0; + return AllZeros ? 0 : AllOnes ? 3 : IsLeftMostBitSet ? 2 : 1; + }; + SDValue Op0 = CCNode->getOperand(0); + SDValue Op1 = CCNode->getOperand(1); + auto [Op0CC, Op0CCValid] = findCCUse(Op0); + if (Op0CC == SDValue()) return false; - if (CompareRHS->getAPIntValue() == FalseVal->getAPIntValue()) - Invert = !Invert; - else if (CompareRHS->getAPIntValue() != TrueVal->getAPIntValue()) + const auto &&Op0SDVals = simplifyAssumingCCVal(Op0, Op0CC, DAG); + const auto &&Op1SDVals = simplifyAssumingCCVal(Op1, Op0CC, DAG); + if (Op0SDVals.empty() || Op1SDVals.empty()) return false; - - // Compute the effective CC mask for the new branch or select. - auto *NewCCValid = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(2)); - auto *NewCCMask = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(3)); - if (!NewCCValid || !NewCCMask) - return false; - CCValid = NewCCValid->getZExtValue(); - CCMask = NewCCMask->getZExtValue(); - if (Invert) - CCMask ^= CCValid; - - // Return the updated CCReg link. - CCReg = CompareLHS->getOperand(4); + int NewCCMask = 0; + for (auto CC : {0, 1, 2, 3}) { + auto CCVal = emulateTMCCMask(Op0SDVals[CC], Op1SDVals[CC]); + if (CCVal < 0) + return false; + NewCCMask <<= 1; + NewCCMask |= (CCMask & (1 << (3 - CCVal))) != 0; + } + NewCCMask &= Op0CCValid; + CCReg = Op0CC; + CCMask = NewCCMask; + CCValid = Op0CCValid; return true; } + if (CCNode->getOpcode() != SystemZISD::ICMP || + CCValid != SystemZ::CCMASK_ICMP) + return false; - // Optimize the case where CompareRHS is (SRA (SHL (IPM))). - if (CompareLHS->getOpcode() == ISD::SRA) { - auto *SRACount = dyn_cast<ConstantSDNode>(CompareLHS->getOperand(1)); - if (!SRACount || SRACount->getZExtValue() != 30) - return false; - auto *SHL = CompareLHS->getOperand(0).getNode(); - if (SHL->getOpcode() != ISD::SHL) - return false; - auto *SHLCount = dyn_cast<ConstantSDNode>(SHL->getOperand(1)); - if (!SHLCount || SHLCount->getZExtValue() != 30 - SystemZ::IPM_CC) - return false; - auto *IPM = SHL->getOperand(0).getNode(); - if (IPM->getOpcode() != SystemZISD::IPM) - return false; - - // Avoid introducing CC spills (because SRA would clobber CC). - if (!CompareLHS->hasOneUse()) - return false; - // Verify that the ICMP compares against zero. - if (CompareRHS->getZExtValue() != 0) + SDValue CmpOp0 = CCNode->getOperand(0); + SDValue CmpOp1 = CCNode->getOperand(1); + SDValue CmpOp2 = CCNode->getOperand(2); + auto [Op0CC, Op0CCValid] = findCCUse(CmpOp0); + if (Op0CC != SDValue()) { + const auto &&Op0SDVals = simplifyAssumingCCVal(CmpOp0, Op0CC, DAG); + const auto &&Op1SDVals = simplifyAssumingCCVal(CmpOp1, Op0CC, DAG); + if (Op0SDVals.empty() || Op1SDVals.empty()) return false; - // Compute the effective CC mask for the new branch or select. - CCMask = SystemZ::reverseCCMask(CCMask); - - // Return the updated CCReg link. - CCReg = IPM->getOperand(0); + auto *CmpType = dyn_cast<ConstantSDNode>(CmpOp2); + auto CmpTypeVal = CmpType->getZExtValue(); + const auto compareCCSigned = [&CmpTypeVal](const SDValue &Op0Val, + const SDValue &Op1Val) { + auto *Op0Node = dyn_cast<ConstantSDNode>(Op0Val.getNode()); + auto *Op1Node = dyn_cast<ConstantSDNode>(Op1Val.getNode()); + if (!Op0Node || !Op1Node) + return -1; + auto Op0APVal = Op0Node->getAPIntValue(); + auto Op1APVal = Op1Node->getAPIntValue(); + if (CmpTypeVal == SystemZICMP::SignedOnly) + return Op0APVal == Op1APVal ? 0 : Op0APVal.slt(Op1APVal) ? 1 : 2; + return Op0APVal == Op1APVal ? 0 : Op0APVal.ult(Op1APVal) ? 1 : 2; + }; + int NewCCMask = 0; + for (auto CC : {0, 1, 2, 3}) { + auto CCVal = compareCCSigned(Op0SDVals[CC], Op1SDVals[CC]); + if (CCVal < 0) + return false; + NewCCMask <<= 1; + NewCCMask |= (CCMask & (1 << (3 - CCVal))) != 0; + } + NewCCMask &= Op0CCValid; + CCMask = NewCCMask; + CCReg = Op0CC; + CCValid = Op0CCValid; return true; } return false; } -SDValue SystemZTargetLowering::combineBR_CCMASK( - SDNode *N, DAGCombinerInfo &DCI) const { +// Merging versus split in multiple branches cost. +TargetLoweringBase::CondMergingParams +SystemZTargetLowering::getJumpConditionMergingParams(Instruction::BinaryOps Opc, + const Value *Lhs, + const Value *Rhs) const { + const auto isFlagOutOpCC = [](const Value *V) { + using namespace llvm::PatternMatch; + const Value *RHSVal; + const APInt *RHSC; + if (const auto *I = dyn_cast<Instruction>(V)) { + // PatternMatch.h provides concise tree-based pattern match of llvm IR. + if (match(I->getOperand(0), m_And(m_Value(RHSVal), m_APInt(RHSC))) || + match(I, m_Cmp(m_Value(RHSVal), m_APInt(RHSC)))) { + if (const auto *CB = dyn_cast<CallBase>(RHSVal)) { + if (CB->isInlineAsm()) { + const InlineAsm *IA = cast<InlineAsm>(CB->getCalledOperand()); + return IA && + IA->getConstraintString().find("{@cc}") != std::string::npos; + } + } + } + } + return false; + }; + // Pattern (ICmp %asm) or (ICmp (And %asm)). + // Cost of longest dependency chain (ICmp, And) is 2. CostThreshold or + // BaseCost can be set >=2. If cost of instruction <= CostThreshold + // conditionals will be merged or else conditionals will be split. + if (isFlagOutOpCC(Lhs) && isFlagOutOpCC(Rhs)) + return {3, 0, -1}; + // Default. + return {-1, -1, -1}; +} + +SDValue SystemZTargetLowering::combineBR_CCMASK(SDNode *N, + DAGCombinerInfo &DCI) const { SelectionDAG &DAG = DCI.DAG; // Combine BR_CCMASK (ICMP (SELECT_CCMASK)) into a single BR_CCMASK. @@ -8824,8 +9009,7 @@ SDValue SystemZTargetLowering::combineBR_CCMASK( int CCMaskVal = CCMask->getZExtValue(); SDValue Chain = N->getOperand(0); SDValue CCReg = N->getOperand(4); - - if (combineCCMask(CCReg, CCValidVal, CCMaskVal)) + if (combineCCMask(CCReg, CCValidVal, CCMaskVal, DAG)) return DAG.getNode(SystemZISD::BR_CCMASK, SDLoc(N), N->getValueType(0), Chain, DAG.getTargetConstant(CCValidVal, SDLoc(N), MVT::i32), @@ -8848,16 +9032,80 @@ SDValue SystemZTargetLowering::combineSELECT_CCMASK( int CCMaskVal = CCMask->getZExtValue(); SDValue CCReg = N->getOperand(4); - if (combineCCMask(CCReg, CCValidVal, CCMaskVal)) - return DAG.getNode(SystemZISD::SELECT_CCMASK, SDLoc(N), N->getValueType(0), - N->getOperand(0), N->getOperand(1), - DAG.getTargetConstant(CCValidVal, SDLoc(N), MVT::i32), - DAG.getTargetConstant(CCMaskVal, SDLoc(N), MVT::i32), - CCReg); + bool IsCombinedCCReg = combineCCMask(CCReg, CCValidVal, CCMaskVal, DAG); + + // Populate SDVals vector for each condition code ccval for given Val, which + // can again be another nested select_ccmask with the same CC. + const auto constructCCSDValsFromSELECT = [&CCReg](SDValue &Val) { + if (Val.getOpcode() == SystemZISD::SELECT_CCMASK) { + SmallVector<SDValue, 4> Res; + if (Val.getOperand(4) != CCReg) + return SmallVector<SDValue, 4>{}; + SDValue TrueVal = Val.getOperand(0), FalseVal = Val.getOperand(1); + auto *CCMask = dyn_cast<ConstantSDNode>(Val.getOperand(3)); + if (!CCMask) + return SmallVector<SDValue, 4>{}; + + int CCMaskVal = CCMask->getZExtValue(); + for (auto &CC : {0, 1, 2, 3}) + Res.emplace_back(((CCMaskVal & (1 << (3 - CC))) != 0) ? TrueVal + : FalseVal); + return Res; + } + return SmallVector<SDValue, 4>{Val, Val, Val, Val}; + }; + // Attempting to optimize TrueVal/FalseVal in outermost select_ccmask either + // with CCReg found by combineCCMask or original CCReg. + SDValue TrueVal = N->getOperand(0); + SDValue FalseVal = N->getOperand(1); + auto &&TrueSDVals = simplifyAssumingCCVal(TrueVal, CCReg, DAG); + auto &&FalseSDVals = simplifyAssumingCCVal(FalseVal, CCReg, DAG); + // TrueSDVals/FalseSDVals might be empty in case of non-constant + // TrueVal/FalseVal for select_ccmask, which can not be optimized further. + if (TrueSDVals.empty()) + TrueSDVals = constructCCSDValsFromSELECT(TrueVal); + if (FalseSDVals.empty()) + FalseSDVals = constructCCSDValsFromSELECT(FalseVal); + if (!TrueSDVals.empty() && !FalseSDVals.empty()) { + SmallSet<SDValue, 4> MergedSDValsSet; + // Ignoring CC values outside CCValiid. + for (auto CC : {0, 1, 2, 3}) { + if ((CCValidVal & ((1 << (3 - CC)))) != 0) + MergedSDValsSet.insert(((CCMaskVal & (1 << (3 - CC))) != 0) + ? TrueSDVals[CC] + : FalseSDVals[CC]); + } + if (MergedSDValsSet.size() == 1) + return *MergedSDValsSet.begin(); + if (MergedSDValsSet.size() == 2) { + auto BeginIt = MergedSDValsSet.begin(); + SDValue NewTrueVal = *BeginIt, NewFalseVal = *next(BeginIt); + if (NewTrueVal == FalseVal || NewFalseVal == TrueVal) + std::swap(NewTrueVal, NewFalseVal); + int NewCCMask = 0; + for (auto CC : {0, 1, 2, 3}) { + NewCCMask <<= 1; + NewCCMask |= ((CCMaskVal & (1 << (3 - CC))) != 0) + ? (TrueSDVals[CC] == NewTrueVal) + : (FalseSDVals[CC] == NewTrueVal); + } + CCMaskVal = NewCCMask; + CCMaskVal &= CCValidVal; + TrueVal = NewTrueVal; + FalseVal = NewFalseVal; + IsCombinedCCReg = true; + } + } + + if (IsCombinedCCReg) + return DAG.getNode( + SystemZISD::SELECT_CCMASK, SDLoc(N), N->getValueType(0), TrueVal, + FalseVal, DAG.getTargetConstant(CCValidVal, SDLoc(N), MVT::i32), + DAG.getTargetConstant(CCMaskVal, SDLoc(N), MVT::i32), CCReg); + return SDValue(); } - SDValue SystemZTargetLowering::combineGET_CCMASK( SDNode *N, DAGCombinerInfo &DCI) const { diff --git a/llvm/lib/Target/SystemZ/SystemZISelLowering.h b/llvm/lib/Target/SystemZ/SystemZISelLowering.h index f8706b7..d5b7603 100644 --- a/llvm/lib/Target/SystemZ/SystemZISelLowering.h +++ b/llvm/lib/Target/SystemZ/SystemZISelLowering.h @@ -533,6 +533,18 @@ public: } const char *getTargetNodeName(unsigned Opcode) const override; + + // This function currently returns cost for srl/ipm/cc sequence for merging. + CondMergingParams + getJumpConditionMergingParams(Instruction::BinaryOps Opc, const Value *Lhs, + const Value *Rhs) const override; + + // Handle Lowering flag assembly outputs. + SDValue LowerAsmOutputForConstraint(SDValue &Chain, SDValue &Flag, + const SDLoc &DL, + const AsmOperandInfo &Constraint, + SelectionDAG &DAG) const override; + std::pair<unsigned, const TargetRegisterClass *> getRegForInlineAsmConstraint(const TargetRegisterInfo *TRI, StringRef Constraint, MVT VT) const override; diff --git a/llvm/lib/Target/X86/X86.td b/llvm/lib/Target/X86/X86.td index 3af8b3e..2bf016a 100644 --- a/llvm/lib/Target/X86/X86.td +++ b/llvm/lib/Target/X86/X86.td @@ -1335,9 +1335,8 @@ def ProcessorFeatures { !listconcat(ARLFeatures, ARLSAdditionalFeatures); // Pantherlake - list<SubtargetFeature> PTLAdditionalFeatures = [FeaturePREFETCHI]; list<SubtargetFeature> PTLFeatures = - !listremove(!listconcat(ARLSFeatures, PTLAdditionalFeatures), [FeatureWIDEKL]); + !listremove(ARLSFeatures, [FeatureWIDEKL]); // Clearwaterforest diff --git a/llvm/lib/TargetParser/X86TargetParser.cpp b/llvm/lib/TargetParser/X86TargetParser.cpp index edca7c1..1932a3a 100644 --- a/llvm/lib/TargetParser/X86TargetParser.cpp +++ b/llvm/lib/TargetParser/X86TargetParser.cpp @@ -175,7 +175,7 @@ constexpr FeatureBitset FeaturesArrowlakeS = FeaturesArrowlake | FeatureAVXVNNIINT16 | FeatureSHA512 | FeatureSM3 | FeatureSM4; constexpr FeatureBitset FeaturesPantherlake = - (FeaturesArrowlakeS ^ FeatureWIDEKL) | FeaturePREFETCHI; + (FeaturesArrowlakeS ^ FeatureWIDEKL); constexpr FeatureBitset FeaturesClearwaterforest = (FeaturesSierraforest ^ FeatureWIDEKL) | FeatureAVXVNNIINT16 | FeatureSHA512 | FeatureSM3 | FeatureSM4 | FeaturePREFETCHI | FeatureUSERMSR; diff --git a/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp index 860f8f7..511bca4 100644 --- a/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp +++ b/llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp @@ -3337,7 +3337,7 @@ PHINode *FunctionStackPoisoner::createPHI(IRBuilder<> &IRB, Value *Cond, Value *ValueIfTrue, Instruction *ThenTerm, Value *ValueIfFalse) { - PHINode *PHI = IRB.CreatePHI(IntptrTy, 2); + PHINode *PHI = IRB.CreatePHI(ValueIfTrue->getType(), 2); BasicBlock *CondBlock = cast<Instruction>(Cond)->getParent(); PHI->addIncoming(ValueIfFalse, CondBlock); BasicBlock *ThenBlock = ThenTerm->getParent(); @@ -3360,7 +3360,7 @@ Value *FunctionStackPoisoner::createAllocaForLayout( assert((ClRealignStack & (ClRealignStack - 1)) == 0); uint64_t FrameAlignment = std::max(L.FrameAlignment, uint64_t(ClRealignStack)); Alloca->setAlignment(Align(FrameAlignment)); - return IRB.CreatePointerCast(Alloca, IntptrTy); + return Alloca; } void FunctionStackPoisoner::createDynamicAllocasInitStorage() { @@ -3572,10 +3572,12 @@ void FunctionStackPoisoner::processStaticAllocas() { DoDynamicAlloca &= !HasInlineAsm && !HasReturnsTwiceCall; DoStackMalloc &= !HasInlineAsm && !HasReturnsTwiceCall; + Type *PtrTy = F.getDataLayout().getAllocaPtrType(F.getContext()); Value *StaticAlloca = DoDynamicAlloca ? nullptr : createAllocaForLayout(IRB, L, false); - Value *FakeStack; + Value *FakeStackPtr; + Value *FakeStackInt; Value *LocalStackBase; Value *LocalStackBaseAlloca; uint8_t DIExprFlags = DIExpression::ApplyOffset; @@ -3603,20 +3605,21 @@ void FunctionStackPoisoner::processStaticAllocas() { RTCI.createRuntimeCall(IRBIf, AsanStackMallocFunc[StackMallocIdx], ConstantInt::get(IntptrTy, LocalStackSize)); IRB.SetInsertPoint(InsBefore); - FakeStack = createPHI(IRB, UseAfterReturnIsEnabled, FakeStackValue, Term, - ConstantInt::get(IntptrTy, 0)); + FakeStackInt = createPHI(IRB, UseAfterReturnIsEnabled, FakeStackValue, + Term, ConstantInt::get(IntptrTy, 0)); } else { // assert(ASan.UseAfterReturn == AsanDetectStackUseAfterReturnMode:Always) // void *FakeStack = __asan_stack_malloc_N(LocalStackSize); // void *LocalStackBase = (FakeStack) ? FakeStack : // alloca(LocalStackSize); StackMallocIdx = StackMallocSizeClass(LocalStackSize); - FakeStack = + FakeStackInt = RTCI.createRuntimeCall(IRB, AsanStackMallocFunc[StackMallocIdx], ConstantInt::get(IntptrTy, LocalStackSize)); } + FakeStackPtr = IRB.CreateIntToPtr(FakeStackInt, PtrTy); Value *NoFakeStack = - IRB.CreateICmpEQ(FakeStack, Constant::getNullValue(IntptrTy)); + IRB.CreateICmpEQ(FakeStackInt, Constant::getNullValue(IntptrTy)); Instruction *Term = SplitBlockAndInsertIfThen(NoFakeStack, InsBefore, false); IRBuilder<> IRBIf(Term); @@ -3624,67 +3627,53 @@ void FunctionStackPoisoner::processStaticAllocas() { DoDynamicAlloca ? createAllocaForLayout(IRBIf, L, true) : StaticAlloca; IRB.SetInsertPoint(InsBefore); - LocalStackBase = createPHI(IRB, NoFakeStack, AllocaValue, Term, FakeStack); + LocalStackBase = + createPHI(IRB, NoFakeStack, AllocaValue, Term, FakeStackPtr); IRB.CreateStore(LocalStackBase, LocalStackBaseAlloca); DIExprFlags |= DIExpression::DerefBefore; } else { // void *FakeStack = nullptr; // void *LocalStackBase = alloca(LocalStackSize); - FakeStack = ConstantInt::get(IntptrTy, 0); + FakeStackInt = Constant::getNullValue(IntptrTy); + FakeStackPtr = Constant::getNullValue(PtrTy); LocalStackBase = DoDynamicAlloca ? createAllocaForLayout(IRB, L, true) : StaticAlloca; LocalStackBaseAlloca = LocalStackBase; } - // It shouldn't matter whether we pass an `alloca` or a `ptrtoint` as the - // dbg.declare address opereand, but passing a `ptrtoint` seems to confuse - // later passes and can result in dropped variable coverage in debug info. - Value *LocalStackBaseAllocaPtr = - isa<PtrToIntInst>(LocalStackBaseAlloca) - ? cast<PtrToIntInst>(LocalStackBaseAlloca)->getPointerOperand() - : LocalStackBaseAlloca; - assert(isa<AllocaInst>(LocalStackBaseAllocaPtr) && - "Variable descriptions relative to ASan stack base will be dropped"); - // Replace Alloca instructions with base+offset. SmallVector<Value *> NewAllocaPtrs; for (const auto &Desc : SVD) { AllocaInst *AI = Desc.AI; - replaceDbgDeclare(AI, LocalStackBaseAllocaPtr, DIB, DIExprFlags, - Desc.Offset); - Value *NewAllocaPtr = IRB.CreateIntToPtr( - IRB.CreateAdd(LocalStackBase, ConstantInt::get(IntptrTy, Desc.Offset)), - AI->getType()); + replaceDbgDeclare(AI, LocalStackBaseAlloca, DIB, DIExprFlags, Desc.Offset); + Value *NewAllocaPtr = IRB.CreatePtrAdd( + LocalStackBase, ConstantInt::get(IntptrTy, Desc.Offset)); AI->replaceAllUsesWith(NewAllocaPtr); NewAllocaPtrs.push_back(NewAllocaPtr); } // The left-most redzone has enough space for at least 4 pointers. // Write the Magic value to redzone[0]. - Value *BasePlus0 = IRB.CreateIntToPtr(LocalStackBase, IntptrPtrTy); IRB.CreateStore(ConstantInt::get(IntptrTy, kCurrentStackFrameMagic), - BasePlus0); + LocalStackBase); // Write the frame description constant to redzone[1]. - Value *BasePlus1 = IRB.CreateIntToPtr( - IRB.CreateAdd(LocalStackBase, - ConstantInt::get(IntptrTy, ASan.LongSize / 8)), - IntptrPtrTy); + Value *BasePlus1 = IRB.CreatePtrAdd( + LocalStackBase, ConstantInt::get(IntptrTy, ASan.LongSize / 8)); GlobalVariable *StackDescriptionGlobal = createPrivateGlobalForString(*F.getParent(), DescriptionString, /*AllowMerging*/ true, genName("stack")); Value *Description = IRB.CreatePointerCast(StackDescriptionGlobal, IntptrTy); IRB.CreateStore(Description, BasePlus1); // Write the PC to redzone[2]. - Value *BasePlus2 = IRB.CreateIntToPtr( - IRB.CreateAdd(LocalStackBase, - ConstantInt::get(IntptrTy, 2 * ASan.LongSize / 8)), - IntptrPtrTy); + Value *BasePlus2 = IRB.CreatePtrAdd( + LocalStackBase, ConstantInt::get(IntptrTy, 2 * ASan.LongSize / 8)); IRB.CreateStore(IRB.CreatePointerCast(&F, IntptrTy), BasePlus2); const auto &ShadowAfterScope = GetShadowBytesAfterScope(SVD, L); // Poison the stack red zones at the entry. - Value *ShadowBase = ASan.memToShadow(LocalStackBase, IRB); + Value *ShadowBase = + ASan.memToShadow(IRB.CreatePtrToInt(LocalStackBase, IntptrTy), IRB); // As mask we must use most poisoned case: red zones and after scope. // As bytes we can use either the same or just red zones only. copyToShadow(ShadowAfterScope, ShadowAfterScope, IRB, ShadowBase); @@ -3723,7 +3712,7 @@ void FunctionStackPoisoner::processStaticAllocas() { IRBuilder<> IRBRet(Ret); // Mark the current frame as retired. IRBRet.CreateStore(ConstantInt::get(IntptrTy, kRetiredStackFrameMagic), - BasePlus0); + LocalStackBase); if (DoStackMalloc) { assert(StackMallocIdx >= 0); // if FakeStack != 0 // LocalStackBase == FakeStack @@ -3737,7 +3726,7 @@ void FunctionStackPoisoner::processStaticAllocas() { // else // <This is not a fake stack; unpoison the redzones> Value *Cmp = - IRBRet.CreateICmpNE(FakeStack, Constant::getNullValue(IntptrTy)); + IRBRet.CreateICmpNE(FakeStackInt, Constant::getNullValue(IntptrTy)); Instruction *ThenTerm, *ElseTerm; SplitBlockAndInsertIfThenElse(Cmp, Ret, &ThenTerm, &ElseTerm); @@ -3748,11 +3737,10 @@ void FunctionStackPoisoner::processStaticAllocas() { kAsanStackUseAfterReturnMagic); copyToShadow(ShadowAfterReturn, ShadowAfterReturn, IRBPoison, ShadowBase); - Value *SavedFlagPtrPtr = IRBPoison.CreateAdd( - FakeStack, + Value *SavedFlagPtrPtr = IRBPoison.CreatePtrAdd( + FakeStackPtr, ConstantInt::get(IntptrTy, ClassSize - ASan.LongSize / 8)); - Value *SavedFlagPtr = IRBPoison.CreateLoad( - IntptrTy, IRBPoison.CreateIntToPtr(SavedFlagPtrPtr, IntptrPtrTy)); + Value *SavedFlagPtr = IRBPoison.CreateLoad(IntptrTy, SavedFlagPtrPtr); IRBPoison.CreateStore( Constant::getNullValue(IRBPoison.getInt8Ty()), IRBPoison.CreateIntToPtr(SavedFlagPtr, IRBPoison.getPtrTy())); @@ -3760,7 +3748,7 @@ void FunctionStackPoisoner::processStaticAllocas() { // For larger frames call __asan_stack_free_*. RTCI.createRuntimeCall( IRBPoison, AsanStackFreeFunc[StackMallocIdx], - {FakeStack, ConstantInt::get(IntptrTy, LocalStackSize)}); + {FakeStackInt, ConstantInt::get(IntptrTy, LocalStackSize)}); } IRBuilder<> IRBElse(ElseTerm); diff --git a/llvm/test/Analysis/DependenceAnalysis/becount-couldnotcompute.ll b/llvm/test/Analysis/DependenceAnalysis/becount-couldnotcompute.ll new file mode 100644 index 0000000..49fbad3 --- /dev/null +++ b/llvm/test/Analysis/DependenceAnalysis/becount-couldnotcompute.ll @@ -0,0 +1,22 @@ +; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 6 +; RUN: opt < %s -disable-output "-passes=print<da>" -aa-pipeline=basic-aa 2>&1 | FileCheck %s + +; Test for function isKnownLessThan that calculates a back-edge taken count, +; which can return a CouldNotCompute SCEV. + +define void @test(i64 %conv, ptr %a) { +; CHECK-LABEL: 'test' +; CHECK-NEXT: Src: %ld = load i32, ptr %arrayidx12, align 4 --> Dst: %ld = load i32, ptr %arrayidx12, align 4 +; CHECK-NEXT: da analyze - none! +; +entry: + %sub = add i64 %conv, 1 + br label %loop + +loop: + %i = phi i64 [ %add26, %loop ], [ 0, %entry ] + %arrayidx12 = getelementptr i32, ptr %a, i64 %i + %ld = load i32, ptr %arrayidx12, align 4 + %add26 = add nsw i64 %sub, %i + br label %loop +} diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/knownbits-add.mir b/llvm/test/CodeGen/AArch64/GlobalISel/knownbits-add.mir new file mode 100644 index 0000000..824ada1 --- /dev/null +++ b/llvm/test/CodeGen/AArch64/GlobalISel/knownbits-add.mir @@ -0,0 +1,278 @@ +# NOTE: Assertions have been autogenerated by utils/update_givaluetracking_test_checks.py UTC_ARGS: --version 5 +# RUN: llc -mtriple=aarch64 -passes="print<gisel-value-tracking>" -filetype=null %s 2>&1 | FileCheck %s + +--- +name: Cst +body: | + bb.1: + ; CHECK-LABEL: name: @Cst + ; CHECK-NEXT: %0:_ KnownBits:00000010 SignBits:6 + ; CHECK-NEXT: %1:_ KnownBits:00011000 SignBits:3 + ; CHECK-NEXT: %2:_ KnownBits:00011010 SignBits:3 + %0:_(s8) = G_CONSTANT i8 2 + %1:_(s8) = G_CONSTANT i8 24 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: CstZero +body: | + bb.1: + ; CHECK-LABEL: name: @CstZero + ; CHECK-NEXT: %0:_ KnownBits:00000001 SignBits:7 + ; CHECK-NEXT: %1:_ KnownBits:11111111 SignBits:8 + ; CHECK-NEXT: %2:_ KnownBits:00000000 SignBits:8 + %0:_(s8) = G_CONSTANT i8 1 + %1:_(s8) = G_CONSTANT i8 255 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: CstNegOne +body: | + bb.1: + ; CHECK-LABEL: name: @CstNegOne + ; CHECK-NEXT: %0:_ KnownBits:00000000 SignBits:8 + ; CHECK-NEXT: %1:_ KnownBits:11111111 SignBits:8 + ; CHECK-NEXT: %2:_ KnownBits:11111111 SignBits:8 + %0:_(s8) = G_CONSTANT i8 0 + %1:_(s8) = G_CONSTANT i8 255 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: CstSeven +body: | + bb.1: + ; CHECK-LABEL: name: @CstSeven + ; CHECK-NEXT: %0:_ KnownBits:00001000 SignBits:4 + ; CHECK-NEXT: %1:_ KnownBits:11111111 SignBits:8 + ; CHECK-NEXT: %2:_ KnownBits:00000111 SignBits:5 + %0:_(s8) = G_CONSTANT i8 8 + %1:_(s8) = G_CONSTANT i8 255 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: CstNeg +body: | + bb.1: + ; CHECK-LABEL: name: @CstNeg + ; CHECK-NEXT: %0:_ KnownBits:11100000 SignBits:3 + ; CHECK-NEXT: %1:_ KnownBits:00000010 SignBits:6 + ; CHECK-NEXT: %2:_ KnownBits:11100010 SignBits:3 + %0:_(s8) = G_CONSTANT i8 224 + %1:_(s8) = G_CONSTANT i8 2 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: ScalarVar +body: | + bb.1: + ; CHECK-LABEL: name: @ScalarVar + ; CHECK-NEXT: %0:_ KnownBits:???????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:???????? SignBits:1 + ; CHECK-NEXT: %2:_ KnownBits:???????? SignBits:1 + %0:_(s8) = COPY $b0 + %1:_(s8) = COPY $b1 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: ScalarRhsEarlyOut +body: | + bb.1: + ; CHECK-LABEL: name: @ScalarRhsEarlyOut + ; CHECK-NEXT: %0:_ KnownBits:???????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:00000011 SignBits:6 + ; CHECK-NEXT: %2:_ KnownBits:???????? SignBits:1 + %0:_(s8) = COPY $b0 + %1:_(s8) = G_CONSTANT i8 3 + %2:_(s8) = G_ADD %0, %1 +... +--- +name: ScalarNonNegative +body: | + bb.1: + ; CHECK-LABEL: name: @ScalarNonNegative + ; CHECK-NEXT: %0:_ KnownBits:???????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:00001111 SignBits:4 + ; CHECK-NEXT: %2:_ KnownBits:0000???? SignBits:4 + ; CHECK-NEXT: %3:_ KnownBits:11111111 SignBits:8 + ; CHECK-NEXT: %4:_ KnownBits:???????? SignBits:4 + %0:_(s8) = COPY $b0 + %1:_(s8) = G_CONSTANT i8 15 + %2:_(s8) = G_AND %0, %1 + %3:_(s8) = G_CONSTANT i8 255 + %4:_(s8) = G_ADD %2, %3 +... +--- +name: ScalarLhsEarlyOut +body: | + bb.1: + ; CHECK-LABEL: name: @ScalarLhsEarlyOut + ; CHECK-NEXT: %0:_ KnownBits:???????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:00000011 SignBits:6 + ; CHECK-NEXT: %2:_ KnownBits:???????? SignBits:1 + %0:_(s8) = COPY $b0 + %1:_(s8) = G_CONSTANT i8 3 + %2:_(s8) = G_ADD %1, %0 +... +--- +name: ScalarPartKnown +body: | + bb.1: + ; CHECK-LABEL: name: @ScalarPartKnown + ; CHECK-NEXT: %0:_ KnownBits:???????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:00001111 SignBits:4 + ; CHECK-NEXT: %2:_ KnownBits:0000???? SignBits:4 + ; CHECK-NEXT: %3:_ KnownBits:00000101 SignBits:5 + ; CHECK-NEXT: %4:_ KnownBits:000????? SignBits:3 + %0:_(s8) = COPY $b0 + %1:_(s8) = G_CONSTANT i8 15 + %2:_(s8) = G_AND %0, %1 + %3:_(s8) = G_CONSTANT i8 5 + %4:_(s8) = G_ADD %2, %3 +... +--- +name: VectorCstZero +body: | + bb.1: + ; CHECK-LABEL: name: @VectorCstZero + ; CHECK-NEXT: %0:_ KnownBits:0000000000000001 SignBits:15 + ; CHECK-NEXT: %1:_ KnownBits:1111111111111111 SignBits:16 + ; CHECK-NEXT: %2:_ KnownBits:0000000000000001 SignBits:15 + ; CHECK-NEXT: %3:_ KnownBits:1111111111111111 SignBits:16 + ; CHECK-NEXT: %4:_ KnownBits:0000000000000000 SignBits:16 + %0:_(s16) = G_CONSTANT i16 1 + %1:_(s16) = G_CONSTANT i16 65535 + %2:_(<4 x s16>) = G_BUILD_VECTOR %0, %0, %0, %0 + %3:_(<4 x s16>) = G_BUILD_VECTOR %1, %1, %1, %1 + %4:_(<4 x s16>) = G_ADD %2, %3 +... +--- +name: VectorCstNegOne +body: | + bb.1: + ; CHECK-LABEL: name: @VectorCstNegOne + ; CHECK-NEXT: %0:_ KnownBits:0000000000000000 SignBits:16 + ; CHECK-NEXT: %1:_ KnownBits:1111111111111111 SignBits:16 + ; CHECK-NEXT: %2:_ KnownBits:0000000000000000 SignBits:16 + ; CHECK-NEXT: %3:_ KnownBits:1111111111111111 SignBits:16 + ; CHECK-NEXT: %4:_ KnownBits:1111111111111111 SignBits:16 + %0:_(s16) = G_CONSTANT i16 0 + %1:_(s16) = G_CONSTANT i16 65535 + %2:_(<4 x s16>) = G_BUILD_VECTOR %0, %0, %0, %0 + %3:_(<4 x s16>) = G_BUILD_VECTOR %1, %1, %1, %1 + %4:_(<4 x s16>) = G_ADD %2, %3 +... +--- +name: VectorVar +body: | + bb.1: + ; CHECK-LABEL: name: @VectorVar + ; CHECK-NEXT: %0:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %2:_ KnownBits:???????????????? SignBits:1 + %0:_(<4 x s16>) = COPY $d0 + %1:_(<4 x s16>) = COPY $d1 + %2:_(<4 x s16>) = G_ADD %0, %1 +... +--- +name: VectorRhsEarlyOut +body: | + bb.1: + ; CHECK-LABEL: name: @VectorRhsEarlyOut + ; CHECK-NEXT: %0:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:0000000000000011 SignBits:14 + ; CHECK-NEXT: %2:_ KnownBits:0000000000000011 SignBits:14 + ; CHECK-NEXT: %3:_ KnownBits:???????????????? SignBits:1 + %0:_(<4 x s16>) = COPY $d0 + %1:_(s16) = G_CONSTANT i16 3 + %2:_(<4 x s16>) = G_BUILD_VECTOR %1, %1, %1, %1 + %3:_(<4 x s16>) = G_ADD %2, %0 +... +--- +name: VectorNonNegative +body: | + bb.1: + ; CHECK-LABEL: name: @VectorNonNegative + ; CHECK-NEXT: %0:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:0000000011111111 SignBits:8 + ; CHECK-NEXT: %2:_ KnownBits:0000000011111111 SignBits:8 + ; CHECK-NEXT: %3:_ KnownBits:00000000???????? SignBits:8 + ; CHECK-NEXT: %4:_ KnownBits:1111111111111111 SignBits:16 + ; CHECK-NEXT: %5:_ KnownBits:1111111111111111 SignBits:16 + ; CHECK-NEXT: %6:_ KnownBits:???????????????? SignBits:8 + %0:_(<4 x s16>) = COPY $d0 + %1:_(s16) = G_CONSTANT i16 255 + %2:_(<4 x s16>) = G_BUILD_VECTOR %1, %1, %1, %1 + %3:_(<4 x s16>) = G_AND %0, %2 + %4:_(s16) = G_CONSTANT i16 65535 + %5:_(<4 x s16>) = G_BUILD_VECTOR %4, %4, %4, %4 + %6:_(<4 x s16>) = G_ADD %3, %5 +... +--- +name: VectorLhsEarlyOut +body: | + bb.1: + ; CHECK-LABEL: name: @VectorLhsEarlyOut + ; CHECK-NEXT: %0:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:0000000000000011 SignBits:14 + ; CHECK-NEXT: %2:_ KnownBits:0000000000000011 SignBits:14 + ; CHECK-NEXT: %3:_ KnownBits:???????????????? SignBits:1 + %0:_(<4 x s16>) = COPY $d0 + %1:_(s16) = G_CONSTANT i16 3 + %2:_(<4 x s16>) = G_BUILD_VECTOR %1, %1, %1, %1 + %3:_(<4 x s16>) = G_ADD %0, %2 +... +--- +name: VectorPartKnown +body: | + bb.1: + ; CHECK-LABEL: name: @VectorPartKnown + ; CHECK-NEXT: %0:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:0000000011111111 SignBits:8 + ; CHECK-NEXT: %2:_ KnownBits:0000000011111111 SignBits:8 + ; CHECK-NEXT: %3:_ KnownBits:00000000???????? SignBits:8 + ; CHECK-NEXT: %4:_ KnownBits:0000000000101010 SignBits:10 + ; CHECK-NEXT: %5:_ KnownBits:0000000001001010 SignBits:9 + ; CHECK-NEXT: %6:_ KnownBits:000000000??01010 SignBits:9 + ; CHECK-NEXT: %7:_ KnownBits:0000000????????? SignBits:7 + %0:_(<4 x s16>) = COPY $d0 + %1:_(s16) = G_CONSTANT i16 255 + %2:_(<4 x s16>) = G_BUILD_VECTOR %1, %1, %1, %1 + %3:_(<4 x s16>) = G_AND %0, %2 + %4:_(s16) = G_CONSTANT i16 42 + %5:_(s16) = G_CONSTANT i16 74 + %6:_(<4 x s16>) = G_BUILD_VECTOR %4, %5, %5, %4 + %7:_(<4 x s16>) = G_ADD %6, %3 +... +--- +name: VectorCst36 +body: | + bb.1: + ; CHECK-LABEL: name: @VectorCst36 + ; CHECK-NEXT: %0:_ KnownBits:0000000000000011 SignBits:14 + ; CHECK-NEXT: %1:_ KnownBits:0000000000000110 SignBits:13 + ; CHECK-NEXT: %2:_ KnownBits:0000000000000?1? SignBits:13 + ; CHECK-NEXT: %3:_ KnownBits:0000000000000?1? SignBits:13 + ; CHECK-NEXT: %4:_ KnownBits:000000000000???? SignBits:12 + %0:_(s16) = G_CONSTANT i16 3 + %1:_(s16) = G_CONSTANT i16 6 + %2:_(<4 x s16>) = G_BUILD_VECTOR %0, %1, %1, %0 + %3:_(<4 x s16>) = G_BUILD_VECTOR %0, %1, %1, %0 + %4:_(<4 x s16>) = G_ADD %2, %3 +... + +--- +name: VectorCst3unknown +body: | + bb.1: + ; CHECK-LABEL: name: @VectorCst3unknown + ; CHECK-NEXT: %0:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %1:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %2:_ KnownBits:0000000000000011 SignBits:14 + ; CHECK-NEXT: %3:_ KnownBits:???????????????? SignBits:1 + ; CHECK-NEXT: %4:_ KnownBits:???????????????? SignBits:1 + %0:_(<4 x s16>) = COPY $d0 + %1:_(s16) = COPY $h0 + %2:_(s16) = G_CONSTANT i16 3 + %3:_(<4 x s16>) = G_BUILD_VECTOR %1, %2, %2, %1 + %4:_(<4 x s16>) = G_ADD %0, %3 +... diff --git a/llvm/test/CodeGen/AArch64/combine-sdiv.ll b/llvm/test/CodeGen/AArch64/combine-sdiv.ll index dc88f94..cca190f 100644 --- a/llvm/test/CodeGen/AArch64/combine-sdiv.ll +++ b/llvm/test/CodeGen/AArch64/combine-sdiv.ll @@ -1774,3 +1774,88 @@ define i128 @combine_i128_sdiv_const100(i128 %x) { %1 = sdiv i128 %x, 100 ret i128 %1 } + +; The following only becomes an sdiv_by_one after type legalisation, after which +; the splatted scalar constant has a different type to the splat vector. This +; test verifies DAGCombiner does not care about this type difference. +define <16 x i16> @combine_vec_sdiv_by_one_obfuscated(<16 x i16> %x) "target-features"="+sve" { +; CHECK-SD-LABEL: combine_vec_sdiv_by_one_obfuscated: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: combine_vec_sdiv_by_one_obfuscated: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: movi v2.2d, #0000000000000000 +; CHECK-GI-NEXT: movi v3.8h, #1 +; CHECK-GI-NEXT: smov w8, v0.h[0] +; CHECK-GI-NEXT: mov v3.h[0], v2.h[0] +; CHECK-GI-NEXT: smov w9, v3.h[0] +; CHECK-GI-NEXT: smov w16, v3.h[7] +; CHECK-GI-NEXT: sdiv w14, w8, w9 +; CHECK-GI-NEXT: smov w8, v0.h[1] +; CHECK-GI-NEXT: smov w9, v3.h[1] +; CHECK-GI-NEXT: sdiv w15, w8, w9 +; CHECK-GI-NEXT: smov w8, v0.h[2] +; CHECK-GI-NEXT: smov w9, v3.h[2] +; CHECK-GI-NEXT: sdiv w13, w8, w9 +; CHECK-GI-NEXT: smov w8, v0.h[3] +; CHECK-GI-NEXT: smov w9, v3.h[3] +; CHECK-GI-NEXT: sdiv w12, w8, w9 +; CHECK-GI-NEXT: smov w8, v0.h[4] +; CHECK-GI-NEXT: smov w9, v3.h[4] +; CHECK-GI-NEXT: sdiv w11, w8, w9 +; CHECK-GI-NEXT: smov w8, v0.h[5] +; CHECK-GI-NEXT: smov w9, v3.h[5] +; CHECK-GI-NEXT: sdiv w10, w8, w9 +; CHECK-GI-NEXT: smov w8, v0.h[6] +; CHECK-GI-NEXT: smov w9, v3.h[6] +; CHECK-GI-NEXT: movi v3.8h, #1 +; CHECK-GI-NEXT: smov w17, v3.h[0] +; CHECK-GI-NEXT: smov w18, v3.h[1] +; CHECK-GI-NEXT: smov w0, v3.h[2] +; CHECK-GI-NEXT: smov w1, v3.h[3] +; CHECK-GI-NEXT: smov w2, v3.h[4] +; CHECK-GI-NEXT: smov w3, v3.h[5] +; CHECK-GI-NEXT: sdiv w8, w8, w9 +; CHECK-GI-NEXT: smov w9, v0.h[7] +; CHECK-GI-NEXT: fmov s0, w14 +; CHECK-GI-NEXT: mov v0.h[1], w15 +; CHECK-GI-NEXT: smov w15, v1.h[6] +; CHECK-GI-NEXT: mov v0.h[2], w13 +; CHECK-GI-NEXT: sdiv w9, w9, w16 +; CHECK-GI-NEXT: smov w16, v1.h[0] +; CHECK-GI-NEXT: mov v0.h[3], w12 +; CHECK-GI-NEXT: smov w12, v1.h[7] +; CHECK-GI-NEXT: mov v0.h[4], w11 +; CHECK-GI-NEXT: sdiv w16, w16, w17 +; CHECK-GI-NEXT: smov w17, v1.h[1] +; CHECK-GI-NEXT: mov v0.h[5], w10 +; CHECK-GI-NEXT: mov v0.h[6], w8 +; CHECK-GI-NEXT: sdiv w17, w17, w18 +; CHECK-GI-NEXT: smov w18, v1.h[2] +; CHECK-GI-NEXT: fmov s2, w16 +; CHECK-GI-NEXT: smov w16, v3.h[6] +; CHECK-GI-NEXT: mov v0.h[7], w9 +; CHECK-GI-NEXT: sdiv w18, w18, w0 +; CHECK-GI-NEXT: smov w0, v1.h[3] +; CHECK-GI-NEXT: mov v2.h[1], w17 +; CHECK-GI-NEXT: sdiv w0, w0, w1 +; CHECK-GI-NEXT: smov w1, v1.h[4] +; CHECK-GI-NEXT: mov v2.h[2], w18 +; CHECK-GI-NEXT: sdiv w1, w1, w2 +; CHECK-GI-NEXT: smov w2, v1.h[5] +; CHECK-GI-NEXT: mov v2.h[3], w0 +; CHECK-GI-NEXT: sdiv w14, w2, w3 +; CHECK-GI-NEXT: mov v2.h[4], w1 +; CHECK-GI-NEXT: sdiv w13, w15, w16 +; CHECK-GI-NEXT: smov w15, v3.h[7] +; CHECK-GI-NEXT: mov v2.h[5], w14 +; CHECK-GI-NEXT: sdiv w10, w12, w15 +; CHECK-GI-NEXT: mov v2.h[6], w13 +; CHECK-GI-NEXT: mov v2.h[7], w10 +; CHECK-GI-NEXT: mov v1.16b, v2.16b +; CHECK-GI-NEXT: ret + %zero_and_ones = shufflevector <16 x i16> zeroinitializer, <16 x i16> splat (i16 1), <16 x i32> <i32 0, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> + %div = sdiv <16 x i16> %x, %zero_and_ones + ret <16 x i16> %div +} diff --git a/llvm/test/CodeGen/AArch64/sme-za-exceptions.ll b/llvm/test/CodeGen/AArch64/sme-za-exceptions.ll index b6dee97e..b8d6c88 100644 --- a/llvm/test/CodeGen/AArch64/sme-za-exceptions.ll +++ b/llvm/test/CodeGen/AArch64/sme-za-exceptions.ll @@ -732,6 +732,247 @@ exit: ret void } +; This example corresponds to: +; +; __arm_agnostic("sme_za_state") void try_catch_agnostic_za_invoke() +; { +; try { +; agnostic_za_call(); +; } catch(...) { +; } +; } +; +; In this example we preserve all SME state enabled by PSTATE.ZA using +; `__arm_sme_save` before agnostic_za_call(). This is because on all normal +; returns from an agnostic ZA function ZA state should be preserved. That means +; we need to make sure ZA state is saved in case agnostic_za_call() throws, and +; we need to restore ZA state after unwinding to the catch block. + +define void @try_catch_agnostic_za_invoke() "aarch64_za_state_agnostic" personality ptr @__gxx_personality_v0 { +; CHECK-LABEL: try_catch_agnostic_za_invoke: +; CHECK: .Lfunc_begin5: +; CHECK-NEXT: .cfi_startproc +; CHECK-NEXT: .cfi_personality 156, DW.ref.__gxx_personality_v0 +; CHECK-NEXT: .cfi_lsda 28, .Lexception5 +; CHECK-NEXT: // %bb.0: // %entry +; CHECK-NEXT: stp x29, x30, [sp, #-32]! // 16-byte Folded Spill +; CHECK-NEXT: str x19, [sp, #16] // 8-byte Folded Spill +; CHECK-NEXT: mov x29, sp +; CHECK-NEXT: .cfi_def_cfa w29, 32 +; CHECK-NEXT: .cfi_offset w19, -16 +; CHECK-NEXT: .cfi_offset w30, -24 +; CHECK-NEXT: .cfi_offset w29, -32 +; CHECK-NEXT: bl __arm_sme_state_size +; CHECK-NEXT: sub sp, sp, x0 +; CHECK-NEXT: mov x19, sp +; CHECK-NEXT: .Ltmp15: // EH_LABEL +; CHECK-NEXT: mov x0, x19 +; CHECK-NEXT: bl __arm_sme_save +; CHECK-NEXT: bl agnostic_za_call +; CHECK-NEXT: .Ltmp16: // EH_LABEL +; CHECK-NEXT: .LBB5_1: // %exit +; CHECK-NEXT: mov x0, x19 +; CHECK-NEXT: bl __arm_sme_restore +; CHECK-NEXT: mov sp, x29 +; CHECK-NEXT: ldr x19, [sp, #16] // 8-byte Folded Reload +; CHECK-NEXT: ldp x29, x30, [sp], #32 // 16-byte Folded Reload +; CHECK-NEXT: ret +; CHECK-NEXT: .LBB5_2: // %catch +; CHECK-NEXT: .Ltmp17: // EH_LABEL +; CHECK-NEXT: bl __cxa_begin_catch +; CHECK-NEXT: bl __cxa_end_catch +; CHECK-NEXT: b .LBB5_1 +; +; CHECK-SDAG-LABEL: try_catch_agnostic_za_invoke: +; CHECK-SDAG: .Lfunc_begin5: +; CHECK-SDAG-NEXT: .cfi_startproc +; CHECK-SDAG-NEXT: .cfi_personality 156, DW.ref.__gxx_personality_v0 +; CHECK-SDAG-NEXT: .cfi_lsda 28, .Lexception5 +; CHECK-SDAG-NEXT: // %bb.0: // %entry +; CHECK-SDAG-NEXT: stp x29, x30, [sp, #-32]! // 16-byte Folded Spill +; CHECK-SDAG-NEXT: str x19, [sp, #16] // 8-byte Folded Spill +; CHECK-SDAG-NEXT: mov x29, sp +; CHECK-SDAG-NEXT: .cfi_def_cfa w29, 32 +; CHECK-SDAG-NEXT: .cfi_offset w19, -16 +; CHECK-SDAG-NEXT: .cfi_offset w30, -24 +; CHECK-SDAG-NEXT: .cfi_offset w29, -32 +; CHECK-SDAG-NEXT: bl __arm_sme_state_size +; CHECK-SDAG-NEXT: sub sp, sp, x0 +; CHECK-SDAG-NEXT: mov x19, sp +; CHECK-SDAG-NEXT: .Ltmp15: // EH_LABEL +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_save +; CHECK-SDAG-NEXT: bl agnostic_za_call +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_restore +; CHECK-SDAG-NEXT: .Ltmp16: // EH_LABEL +; CHECK-SDAG-NEXT: .LBB5_1: // %exit +; CHECK-SDAG-NEXT: mov sp, x29 +; CHECK-SDAG-NEXT: ldr x19, [sp, #16] // 8-byte Folded Reload +; CHECK-SDAG-NEXT: ldp x29, x30, [sp], #32 // 16-byte Folded Reload +; CHECK-SDAG-NEXT: ret +; CHECK-SDAG-NEXT: .LBB5_2: // %catch +; CHECK-SDAG-NEXT: .Ltmp17: // EH_LABEL +; CHECK-SDAG-NEXT: mov x1, x0 +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_restore +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_save +; CHECK-SDAG-NEXT: mov x0, x1 +; CHECK-SDAG-NEXT: bl __cxa_begin_catch +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_restore +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_save +; CHECK-SDAG-NEXT: bl __cxa_end_catch +; CHECK-SDAG-NEXT: mov x0, x19 +; CHECK-SDAG-NEXT: bl __arm_sme_restore +; CHECK-SDAG-NEXT: b .LBB5_1 +entry: + invoke void @agnostic_za_call() + to label %exit unwind label %catch + +catch: + %eh_info = landingpad { ptr, i32 } + catch ptr null + %exception_ptr = extractvalue { ptr, i32 } %eh_info, 0 + tail call ptr @__cxa_begin_catch(ptr %exception_ptr) + tail call void @__cxa_end_catch() + br label %exit + +exit: + ret void +} + +; This is the same `try_catch_agnostic_za_invoke`, but shows a lazy save would +; also need to be committed in a shared-ZA function calling an agnostic-ZA function. +define void @try_catch_inout_za_agnostic_za_callee() "aarch64_inout_za" personality ptr @__gxx_personality_v0 { +; CHECK-LABEL: try_catch_inout_za_agnostic_za_callee: +; CHECK: .Lfunc_begin6: +; CHECK-NEXT: .cfi_startproc +; CHECK-NEXT: .cfi_personality 156, DW.ref.__gxx_personality_v0 +; CHECK-NEXT: .cfi_lsda 28, .Lexception6 +; CHECK-NEXT: // %bb.0: // %entry +; CHECK-NEXT: stp x29, x30, [sp, #-16]! // 16-byte Folded Spill +; CHECK-NEXT: mov x29, sp +; CHECK-NEXT: sub sp, sp, #16 +; CHECK-NEXT: .cfi_def_cfa w29, 16 +; CHECK-NEXT: .cfi_offset w30, -8 +; CHECK-NEXT: .cfi_offset w29, -16 +; CHECK-NEXT: rdsvl x8, #1 +; CHECK-NEXT: mov x9, sp +; CHECK-NEXT: msub x9, x8, x8, x9 +; CHECK-NEXT: mov sp, x9 +; CHECK-NEXT: stp x9, x8, [x29, #-16] +; CHECK-NEXT: .Ltmp18: // EH_LABEL +; CHECK-NEXT: sub x8, x29, #16 +; CHECK-NEXT: msr TPIDR2_EL0, x8 +; CHECK-NEXT: bl agnostic_za_call +; CHECK-NEXT: .Ltmp19: // EH_LABEL +; CHECK-NEXT: .LBB6_1: // %exit +; CHECK-NEXT: smstart za +; CHECK-NEXT: mrs x8, TPIDR2_EL0 +; CHECK-NEXT: sub x0, x29, #16 +; CHECK-NEXT: cbnz x8, .LBB6_3 +; CHECK-NEXT: // %bb.2: // %exit +; CHECK-NEXT: bl __arm_tpidr2_restore +; CHECK-NEXT: .LBB6_3: // %exit +; CHECK-NEXT: msr TPIDR2_EL0, xzr +; CHECK-NEXT: mov sp, x29 +; CHECK-NEXT: ldp x29, x30, [sp], #16 // 16-byte Folded Reload +; CHECK-NEXT: ret +; CHECK-NEXT: .LBB6_4: // %catch +; CHECK-NEXT: .Ltmp20: // EH_LABEL +; CHECK-NEXT: bl __cxa_begin_catch +; CHECK-NEXT: bl __cxa_end_catch +; CHECK-NEXT: b .LBB6_1 +; +; CHECK-SDAG-LABEL: try_catch_inout_za_agnostic_za_callee: +; CHECK-SDAG: .Lfunc_begin6: +; CHECK-SDAG-NEXT: .cfi_startproc +; CHECK-SDAG-NEXT: .cfi_personality 156, DW.ref.__gxx_personality_v0 +; CHECK-SDAG-NEXT: .cfi_lsda 28, .Lexception6 +; CHECK-SDAG-NEXT: // %bb.0: // %entry +; CHECK-SDAG-NEXT: stp x29, x30, [sp, #-32]! // 16-byte Folded Spill +; CHECK-SDAG-NEXT: str x19, [sp, #16] // 8-byte Folded Spill +; CHECK-SDAG-NEXT: mov x29, sp +; CHECK-SDAG-NEXT: sub sp, sp, #16 +; CHECK-SDAG-NEXT: .cfi_def_cfa w29, 32 +; CHECK-SDAG-NEXT: .cfi_offset w19, -16 +; CHECK-SDAG-NEXT: .cfi_offset w30, -24 +; CHECK-SDAG-NEXT: .cfi_offset w29, -32 +; CHECK-SDAG-NEXT: rdsvl x8, #1 +; CHECK-SDAG-NEXT: mov x9, sp +; CHECK-SDAG-NEXT: msub x9, x8, x8, x9 +; CHECK-SDAG-NEXT: mov sp, x9 +; CHECK-SDAG-NEXT: stp x9, x8, [x29, #-16] +; CHECK-SDAG-NEXT: .Ltmp18: // EH_LABEL +; CHECK-SDAG-NEXT: sub x19, x29, #16 +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, x19 +; CHECK-SDAG-NEXT: bl agnostic_za_call +; CHECK-SDAG-NEXT: smstart za +; CHECK-SDAG-NEXT: mrs x8, TPIDR2_EL0 +; CHECK-SDAG-NEXT: sub x0, x29, #16 +; CHECK-SDAG-NEXT: cbnz x8, .LBB6_2 +; CHECK-SDAG-NEXT: // %bb.1: // %entry +; CHECK-SDAG-NEXT: bl __arm_tpidr2_restore +; CHECK-SDAG-NEXT: .LBB6_2: // %entry +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, xzr +; CHECK-SDAG-NEXT: .Ltmp19: // EH_LABEL +; CHECK-SDAG-NEXT: .LBB6_3: // %exit +; CHECK-SDAG-NEXT: mov sp, x29 +; CHECK-SDAG-NEXT: ldr x19, [sp, #16] // 8-byte Folded Reload +; CHECK-SDAG-NEXT: ldp x29, x30, [sp], #32 // 16-byte Folded Reload +; CHECK-SDAG-NEXT: ret +; CHECK-SDAG-NEXT: .LBB6_4: // %catch +; CHECK-SDAG-NEXT: .Ltmp20: // EH_LABEL +; CHECK-SDAG-NEXT: mov x1, x0 +; CHECK-SDAG-NEXT: smstart za +; CHECK-SDAG-NEXT: mrs x8, TPIDR2_EL0 +; CHECK-SDAG-NEXT: sub x0, x29, #16 +; CHECK-SDAG-NEXT: cbnz x8, .LBB6_6 +; CHECK-SDAG-NEXT: // %bb.5: // %catch +; CHECK-SDAG-NEXT: bl __arm_tpidr2_restore +; CHECK-SDAG-NEXT: .LBB6_6: // %catch +; CHECK-SDAG-NEXT: mov x0, x1 +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, xzr +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, x19 +; CHECK-SDAG-NEXT: bl __cxa_begin_catch +; CHECK-SDAG-NEXT: smstart za +; CHECK-SDAG-NEXT: mrs x8, TPIDR2_EL0 +; CHECK-SDAG-NEXT: sub x0, x29, #16 +; CHECK-SDAG-NEXT: cbnz x8, .LBB6_8 +; CHECK-SDAG-NEXT: // %bb.7: // %catch +; CHECK-SDAG-NEXT: bl __arm_tpidr2_restore +; CHECK-SDAG-NEXT: .LBB6_8: // %catch +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, xzr +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, x19 +; CHECK-SDAG-NEXT: bl __cxa_end_catch +; CHECK-SDAG-NEXT: smstart za +; CHECK-SDAG-NEXT: mrs x8, TPIDR2_EL0 +; CHECK-SDAG-NEXT: sub x0, x29, #16 +; CHECK-SDAG-NEXT: cbnz x8, .LBB6_10 +; CHECK-SDAG-NEXT: // %bb.9: // %catch +; CHECK-SDAG-NEXT: bl __arm_tpidr2_restore +; CHECK-SDAG-NEXT: .LBB6_10: // %catch +; CHECK-SDAG-NEXT: msr TPIDR2_EL0, xzr +; CHECK-SDAG-NEXT: b .LBB6_3 +entry: + invoke void @agnostic_za_call() + to label %exit unwind label %catch + +catch: + %eh_info = landingpad { ptr, i32 } + catch ptr null + %exception_ptr = extractvalue { ptr, i32 } %eh_info, 0 + tail call ptr @__cxa_begin_catch(ptr %exception_ptr) + tail call void @__cxa_end_catch() + br label %exit + +exit: + ret void +} + declare ptr @__cxa_allocate_exception(i64) declare void @__cxa_throw(ptr, ptr, ptr) declare ptr @__cxa_begin_catch(ptr) @@ -742,3 +983,4 @@ declare void @may_throw() declare void @shared_za_call() "aarch64_inout_za" declare void @noexcept_shared_za_call() "aarch64_inout_za" declare void @shared_zt0_call() "aarch64_inout_zt0" +declare void @agnostic_za_call() "aarch64_za_state_agnostic" diff --git a/llvm/test/CodeGen/PowerPC/check-zero-vector.ll b/llvm/test/CodeGen/PowerPC/compare-vector-with-zero.ll index 0f7e0c7..1325abf 100644 --- a/llvm/test/CodeGen/PowerPC/check-zero-vector.ll +++ b/llvm/test/CodeGen/PowerPC/compare-vector-with-zero.ll @@ -95,3 +95,80 @@ declare i4 @llvm.ctpop.i4(i4) #1 !6 = !{!"short", !7, i64 0} !7 = !{!"omnipotent char", !8, i64 0} !8 = !{!"Simple C/C++ TBAA"} + +; Function to lockdown changes for floating point vector comparisons +define range(i32 0, 5) i32 @cols_needed(ptr %colauths){ +; POWERPC_64LE-LABEL: cols_needed: +; POWERPC_64LE: # %bb.0: # %entry +; POWERPC_64LE-NEXT: lxv vs0, 0(r3) +; POWERPC_64LE-NEXT: xxlxor vs1, vs1, vs1 +; POWERPC_64LE-NEXT: li r4, 4 +; POWERPC_64LE-NEXT: li r3, 0 +; POWERPC_64LE-NEXT: xvcmpeqsp vs0, vs0, vs1 +; POWERPC_64LE-NEXT: xxlnor v2, vs0, vs0 +; POWERPC_64LE-NEXT: vextuwrx r4, r4, v2 +; POWERPC_64LE-NEXT: vextuwrx r3, r3, v2 +; POWERPC_64LE-NEXT: rlwinm r4, r4, 1, 30, 30 +; POWERPC_64LE-NEXT: sub r3, r4, r3 +; POWERPC_64LE-NEXT: mfvsrwz r4, v2 +; POWERPC_64LE-NEXT: rlwinm r4, r4, 2, 29, 29 +; POWERPC_64LE-NEXT: or r3, r3, r4 +; POWERPC_64LE-NEXT: li r4, 12 +; POWERPC_64LE-NEXT: vextuwrx r4, r4, v2 +; POWERPC_64LE-NEXT: slwi r4, r4, 3 +; POWERPC_64LE-NEXT: or r3, r3, r4 +; POWERPC_64LE-NEXT: clrlwi r3, r3, 28 +; POWERPC_64LE-NEXT: stb r3, -1(r1) +; POWERPC_64LE-NEXT: lbz r3, -1(r1) +; POWERPC_64LE-NEXT: popcntd r3, r3 +; POWERPC_64LE-NEXT: blr +; +; POWERPC_64-LABEL: cols_needed: +; POWERPC_64: # %bb.0: # %entry +; POWERPC_64-NEXT: lxv vs0, 0(r3) +; POWERPC_64-NEXT: xxlxor vs1, vs1, vs1 +; POWERPC_64-NEXT: li r4, 8 +; POWERPC_64-NEXT: xvcmpeqsp vs0, vs0, vs1 +; POWERPC_64-NEXT: xxlnor v2, vs0, vs0 +; POWERPC_64-NEXT: vextuwlx r4, r4, v2 +; POWERPC_64-NEXT: mfvsrwz r3, v2 +; POWERPC_64-NEXT: rlwinm r4, r4, 1, 30, 30 +; POWERPC_64-NEXT: rlwimi r4, r3, 2, 29, 29 +; POWERPC_64-NEXT: li r3, 0 +; POWERPC_64-NEXT: vextuwlx r3, r3, v2 +; POWERPC_64-NEXT: rlwimi r4, r3, 3, 0, 28 +; POWERPC_64-NEXT: li r3, 12 +; POWERPC_64-NEXT: vextuwlx r3, r3, v2 +; POWERPC_64-NEXT: sub r3, r4, r3 +; POWERPC_64-NEXT: clrlwi r3, r3, 28 +; POWERPC_64-NEXT: stb r3, -1(r1) +; POWERPC_64-NEXT: lbz r3, -1(r1) +; POWERPC_64-NEXT: popcntd r3, r3 +; POWERPC_64-NEXT: blr +; +; POWERPC_32-LABEL: cols_needed: +; POWERPC_32: # %bb.0: # %entry +; POWERPC_32-NEXT: lxv vs0, 0(r3) +; POWERPC_32-NEXT: xxlxor vs1, vs1, vs1 +; POWERPC_32-NEXT: xvcmpeqsp vs0, vs0, vs1 +; POWERPC_32-NEXT: xxlnor vs0, vs0, vs0 +; POWERPC_32-NEXT: stxv vs0, -32(r1) +; POWERPC_32-NEXT: lwz r3, -24(r1) +; POWERPC_32-NEXT: lwz r4, -28(r1) +; POWERPC_32-NEXT: rlwinm r3, r3, 1, 30, 30 +; POWERPC_32-NEXT: rlwimi r3, r4, 2, 29, 29 +; POWERPC_32-NEXT: lwz r4, -32(r1) +; POWERPC_32-NEXT: rlwimi r3, r4, 3, 0, 28 +; POWERPC_32-NEXT: lwz r4, -20(r1) +; POWERPC_32-NEXT: sub r3, r3, r4 +; POWERPC_32-NEXT: clrlwi r3, r3, 28 +; POWERPC_32-NEXT: popcntw r3, r3 +; POWERPC_32-NEXT: blr +entry: + %0 = load <4 x float>, ptr %colauths, align 4, !tbaa !5 + %1 = fcmp une <4 x float> %0, zeroinitializer + %2 = bitcast <4 x i1> %1 to i4 + %3 = tail call range(i4 0, 5) i4 @llvm.ctpop.i4(i4 %2) + %4 = zext nneg i4 %3 to i32 + ret i32 %4 +} diff --git a/llvm/test/CodeGen/RISCV/min-max.ll b/llvm/test/CodeGen/RISCV/min-max.ll index acde8ad..e7f6899 100644 --- a/llvm/test/CodeGen/RISCV/min-max.ll +++ b/llvm/test/CodeGen/RISCV/min-max.ll @@ -5,6 +5,12 @@ ; RUN: FileCheck %s --check-prefixes=ZBB,RV32ZBB ; RUN: llc < %s -mtriple=riscv64 -mattr=+zbb | \ ; RUN: FileCheck %s --check-prefixes=ZBB,RV64ZBB +; RUN: llc -mtriple=riscv32 -mattr=+experimental-xqcicm,+experimental-xqcics,+experimental-xqcicli,+zca,+short-forward-branch-opt,+conditional-cmv-fusion -verify-machineinstrs < %s | \ +; RUN: FileCheck %s --check-prefixes=XQCI +; RUN: llc < %s -mtriple=riscv32 -mattr=+short-forward-branch-opt | \ +; RUN: FileCheck %s --check-prefixes=RV32I-SFB +; RUN: llc < %s -mtriple=riscv64 -mattr=+short-forward-branch-opt | \ +; RUN: FileCheck %s --check-prefixes=RV64I-SFB ; Basic tests. @@ -23,6 +29,27 @@ define signext i8 @smax_i8(i8 signext %a, i8 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: max a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smax_i8: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvge a0, a1, a0, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_i8: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: blt a1, a0, .LBB0_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB0_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_i8: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a1, a0, .LBB0_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB0_2: +; RV64I-SFB-NEXT: ret %c = call i8 @llvm.smax.i8(i8 %a, i8 %b) ret i8 %c } @@ -42,6 +69,27 @@ define signext i16 @smax_i16(i16 signext %a, i16 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: max a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smax_i16: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvge a0, a1, a0, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_i16: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: blt a1, a0, .LBB1_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB1_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_i16: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a1, a0, .LBB1_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB1_2: +; RV64I-SFB-NEXT: ret %c = call i16 @llvm.smax.i16(i16 %a, i16 %b) ret i16 %c } @@ -61,6 +109,27 @@ define signext i32 @smax_i32(i32 signext %a, i32 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: max a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smax_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvge a0, a1, a0, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: blt a1, a0, .LBB2_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB2_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a1, a0, .LBB2_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB2_2: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smax.i32(i32 %a, i32 %b) ret i32 %c } @@ -112,6 +181,41 @@ define i64 @smax_i64(i64 %a, i64 %b) { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: max a0, a0, a1 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: smax_i64: +; XQCI: # %bb.0: +; XQCI-NEXT: sltu a4, a2, a0 +; XQCI-NEXT: slt a5, a3, a1 +; XQCI-NEXT: qc.mveq a5, a1, a3, a4 +; XQCI-NEXT: qc.mveqi a0, a5, 0, a2 +; XQCI-NEXT: qc.mveqi a1, a5, 0, a3 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_i64: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: sltu a4, a2, a0 +; RV32I-SFB-NEXT: slt a5, a3, a1 +; RV32I-SFB-NEXT: bne a1, a3, .LBB3_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a5, a4 +; RV32I-SFB-NEXT: .LBB3_2: +; RV32I-SFB-NEXT: bnez a5, .LBB3_4 +; RV32I-SFB-NEXT: # %bb.3: +; RV32I-SFB-NEXT: mv a0, a2 +; RV32I-SFB-NEXT: .LBB3_4: +; RV32I-SFB-NEXT: bnez a5, .LBB3_6 +; RV32I-SFB-NEXT: # %bb.5: +; RV32I-SFB-NEXT: mv a1, a3 +; RV32I-SFB-NEXT: .LBB3_6: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_i64: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a1, a0, .LBB3_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB3_2: +; RV64I-SFB-NEXT: ret %c = call i64 @llvm.smax.i64(i64 %a, i64 %b) ret i64 %c } @@ -131,6 +235,27 @@ define signext i8 @smin_i8(i8 signext %a, i8 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: min a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smin_i8: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvge a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_i8: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: blt a0, a1, .LBB4_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB4_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_i8: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a0, a1, .LBB4_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB4_2: +; RV64I-SFB-NEXT: ret %c = call i8 @llvm.smin.i8(i8 %a, i8 %b) ret i8 %c } @@ -150,6 +275,27 @@ define signext i16 @smin_i16(i16 signext %a, i16 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: min a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smin_i16: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvge a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_i16: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: blt a0, a1, .LBB5_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB5_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_i16: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a0, a1, .LBB5_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB5_2: +; RV64I-SFB-NEXT: ret %c = call i16 @llvm.smin.i16(i16 %a, i16 %b) ret i16 %c } @@ -169,6 +315,27 @@ define signext i32 @smin_i32(i32 signext %a, i32 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: min a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smin_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvge a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: blt a0, a1, .LBB6_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB6_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a0, a1, .LBB6_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB6_2: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smin.i32(i32 %a, i32 %b) ret i32 %c } @@ -220,6 +387,41 @@ define i64 @smin_i64(i64 %a, i64 %b) { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: min a0, a0, a1 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: smin_i64: +; XQCI: # %bb.0: +; XQCI-NEXT: sltu a4, a0, a2 +; XQCI-NEXT: slt a5, a1, a3 +; XQCI-NEXT: qc.mveq a5, a1, a3, a4 +; XQCI-NEXT: qc.mveqi a0, a5, 0, a2 +; XQCI-NEXT: qc.mveqi a1, a5, 0, a3 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_i64: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: sltu a4, a0, a2 +; RV32I-SFB-NEXT: slt a5, a1, a3 +; RV32I-SFB-NEXT: bne a1, a3, .LBB7_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a5, a4 +; RV32I-SFB-NEXT: .LBB7_2: +; RV32I-SFB-NEXT: bnez a5, .LBB7_4 +; RV32I-SFB-NEXT: # %bb.3: +; RV32I-SFB-NEXT: mv a0, a2 +; RV32I-SFB-NEXT: .LBB7_4: +; RV32I-SFB-NEXT: bnez a5, .LBB7_6 +; RV32I-SFB-NEXT: # %bb.5: +; RV32I-SFB-NEXT: mv a1, a3 +; RV32I-SFB-NEXT: .LBB7_6: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_i64: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: blt a0, a1, .LBB7_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB7_2: +; RV64I-SFB-NEXT: ret %c = call i64 @llvm.smin.i64(i64 %a, i64 %b) ret i64 %c } @@ -239,6 +441,27 @@ define i8 @umax_i8(i8 zeroext %a, i8 zeroext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: maxu a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: umax_i8: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvgeu a0, a1, a0, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_i8: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: bltu a1, a0, .LBB8_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB8_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_i8: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a1, a0, .LBB8_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB8_2: +; RV64I-SFB-NEXT: ret %c = call i8 @llvm.umax.i8(i8 %a, i8 %b) ret i8 %c } @@ -258,6 +481,27 @@ define i16 @umax_i16(i16 zeroext %a, i16 zeroext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: maxu a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: umax_i16: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvgeu a0, a1, a0, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_i16: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: bltu a1, a0, .LBB9_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB9_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_i16: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a1, a0, .LBB9_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB9_2: +; RV64I-SFB-NEXT: ret %c = call i16 @llvm.umax.i16(i16 %a, i16 %b) ret i16 %c } @@ -277,6 +521,27 @@ define signext i32 @umax_i32(i32 signext %a, i32 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: maxu a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: umax_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvgeu a0, a1, a0, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: bltu a1, a0, .LBB10_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB10_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a1, a0, .LBB10_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB10_2: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.umax.i32(i32 %a, i32 %b) ret i32 %c } @@ -328,6 +593,41 @@ define i64 @umax_i64(i64 %a, i64 %b) { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: maxu a0, a0, a1 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: umax_i64: +; XQCI: # %bb.0: +; XQCI-NEXT: sltu a4, a2, a0 +; XQCI-NEXT: sltu a5, a3, a1 +; XQCI-NEXT: qc.mveq a5, a1, a3, a4 +; XQCI-NEXT: qc.mveqi a0, a5, 0, a2 +; XQCI-NEXT: qc.mveqi a1, a5, 0, a3 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_i64: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: sltu a4, a2, a0 +; RV32I-SFB-NEXT: sltu a5, a3, a1 +; RV32I-SFB-NEXT: bne a1, a3, .LBB11_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a5, a4 +; RV32I-SFB-NEXT: .LBB11_2: +; RV32I-SFB-NEXT: bnez a5, .LBB11_4 +; RV32I-SFB-NEXT: # %bb.3: +; RV32I-SFB-NEXT: mv a0, a2 +; RV32I-SFB-NEXT: .LBB11_4: +; RV32I-SFB-NEXT: bnez a5, .LBB11_6 +; RV32I-SFB-NEXT: # %bb.5: +; RV32I-SFB-NEXT: mv a1, a3 +; RV32I-SFB-NEXT: .LBB11_6: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_i64: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a1, a0, .LBB11_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB11_2: +; RV64I-SFB-NEXT: ret %c = call i64 @llvm.umax.i64(i64 %a, i64 %b) ret i64 %c } @@ -347,6 +647,27 @@ define zeroext i8 @umin_i8(i8 zeroext %a, i8 zeroext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: minu a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: umin_i8: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvgeu a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umin_i8: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: bltu a0, a1, .LBB12_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB12_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umin_i8: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a0, a1, .LBB12_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB12_2: +; RV64I-SFB-NEXT: ret %c = call i8 @llvm.umin.i8(i8 %a, i8 %b) ret i8 %c } @@ -366,6 +687,27 @@ define zeroext i16 @umin_i16(i16 zeroext %a, i16 zeroext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: minu a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: umin_i16: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvgeu a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umin_i16: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: bltu a0, a1, .LBB13_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB13_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umin_i16: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a0, a1, .LBB13_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB13_2: +; RV64I-SFB-NEXT: ret %c = call i16 @llvm.umin.i16(i16 %a, i16 %b) ret i16 %c } @@ -385,6 +727,27 @@ define signext i32 @umin_i32(i32 signext %a, i32 signext %b) { ; ZBB: # %bb.0: ; ZBB-NEXT: minu a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: umin_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.mvgeu a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umin_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: bltu a0, a1, .LBB14_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB14_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umin_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a0, a1, .LBB14_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB14_2: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.umin.i32(i32 %a, i32 %b) ret i32 %c } @@ -436,6 +799,41 @@ define i64 @umin_i64(i64 %a, i64 %b) { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: minu a0, a0, a1 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: umin_i64: +; XQCI: # %bb.0: +; XQCI-NEXT: sltu a4, a0, a2 +; XQCI-NEXT: sltu a5, a1, a3 +; XQCI-NEXT: qc.mveq a5, a1, a3, a4 +; XQCI-NEXT: qc.mveqi a0, a5, 0, a2 +; XQCI-NEXT: qc.mveqi a1, a5, 0, a3 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umin_i64: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: sltu a4, a0, a2 +; RV32I-SFB-NEXT: sltu a5, a1, a3 +; RV32I-SFB-NEXT: bne a1, a3, .LBB15_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a5, a4 +; RV32I-SFB-NEXT: .LBB15_2: +; RV32I-SFB-NEXT: bnez a5, .LBB15_4 +; RV32I-SFB-NEXT: # %bb.3: +; RV32I-SFB-NEXT: mv a0, a2 +; RV32I-SFB-NEXT: .LBB15_4: +; RV32I-SFB-NEXT: bnez a5, .LBB15_6 +; RV32I-SFB-NEXT: # %bb.5: +; RV32I-SFB-NEXT: mv a1, a3 +; RV32I-SFB-NEXT: .LBB15_6: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umin_i64: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: bltu a0, a1, .LBB15_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB15_2: +; RV64I-SFB-NEXT: ret %c = call i64 @llvm.umin.i64(i64 %a, i64 %b) ret i64 %c } @@ -450,6 +848,18 @@ define signext i32 @smin_same_op_i32(i32 signext %a) { ; ZBB-LABEL: smin_same_op_i32: ; ZBB: # %bb.0: ; ZBB-NEXT: ret +; +; XQCI-LABEL: smin_same_op_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_same_op_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_same_op_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smin.i32(i32 %a, i32 %a) ret i32 %c } @@ -462,6 +872,18 @@ define signext i32 @smax_same_op_i32(i32 signext %a) { ; ZBB-LABEL: smax_same_op_i32: ; ZBB: # %bb.0: ; ZBB-NEXT: ret +; +; XQCI-LABEL: smax_same_op_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_same_op_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_same_op_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smax.i32(i32 %a, i32 %a) ret i32 %c } @@ -474,6 +896,18 @@ define signext i32 @umin_same_op_i32(i32 signext %a) { ; ZBB-LABEL: umin_same_op_i32: ; ZBB: # %bb.0: ; ZBB-NEXT: ret +; +; XQCI-LABEL: umin_same_op_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umin_same_op_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umin_same_op_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.umin.i32(i32 %a, i32 %a) ret i32 %c } @@ -486,6 +920,18 @@ define signext i32 @umax_same_op_i32(i32 signext %a) { ; ZBB-LABEL: umax_same_op_i32: ; ZBB: # %bb.0: ; ZBB-NEXT: ret +; +; XQCI-LABEL: umax_same_op_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_same_op_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_same_op_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.umax.i32(i32 %a, i32 %a) ret i32 %c } @@ -510,6 +956,19 @@ define signext i32 @smin_undef_i32() { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: li a0, 0 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: smin_undef_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_undef_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_undef_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a0, 0 +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smin.i32(i32 undef, i32 undef) ret i32 %c } @@ -532,6 +991,19 @@ define signext i32 @smax_undef_i32() { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: li a0, 0 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: smax_undef_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_undef_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_undef_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a0, 0 +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smax.i32(i32 undef, i32 undef) ret i32 %c } @@ -554,6 +1026,19 @@ define signext i32 @umin_undef_i32() { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: li a0, 0 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: umin_undef_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umin_undef_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umin_undef_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a0, 0 +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.umin.i32(i32 undef, i32 undef) ret i32 %c } @@ -576,6 +1061,19 @@ define signext i32 @umax_undef_i32() { ; RV64ZBB: # %bb.0: ; RV64ZBB-NEXT: li a0, 0 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: umax_undef_i32: +; XQCI: # %bb.0: +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_undef_i32: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_undef_i32: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a0, 0 +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.umax.i32(i32 undef, i32 undef) ret i32 %c } @@ -595,6 +1093,29 @@ define signext i32 @smax_i32_pos_constant(i32 signext %a) { ; ZBB-NEXT: li a1, 10 ; ZBB-NEXT: max a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smax_i32_pos_constant: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.lilti a0, a0, 11, 10 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_i32_pos_constant: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: li a1, 10 +; RV32I-SFB-NEXT: blt a1, a0, .LBB24_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB24_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_i32_pos_constant: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a1, 10 +; RV64I-SFB-NEXT: blt a1, a0, .LBB24_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB24_2: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smax.i32(i32 %a, i32 10) ret i32 %c } @@ -616,6 +1137,33 @@ define signext i32 @smax_i32_pos_constant_trailing_zeros(i32 signext %a) { ; ZBB-NEXT: li a1, 16 ; ZBB-NEXT: max a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smax_i32_pos_constant_trailing_zeros: +; XQCI: # %bb.0: +; XQCI-NEXT: andi a1, a0, -8 +; XQCI-NEXT: li a0, 16 +; XQCI-NEXT: qc.mvlt a0, a0, a1, a1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smax_i32_pos_constant_trailing_zeros: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: andi a1, a0, -8 +; RV32I-SFB-NEXT: li a0, 16 +; RV32I-SFB-NEXT: bge a0, a1, .LBB25_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB25_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smax_i32_pos_constant_trailing_zeros: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: andi a1, a0, -8 +; RV64I-SFB-NEXT: li a0, 16 +; RV64I-SFB-NEXT: bge a0, a1, .LBB25_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB25_2: +; RV64I-SFB-NEXT: ret %b = and i32 %a, -8 %c = call i32 @llvm.smax.i32(i32 %b, i32 16) %d = and i32 %c, -4 @@ -635,6 +1183,29 @@ define signext i32 @smin_i32_negone(i32 signext %a) { ; ZBB-NEXT: li a1, -1 ; ZBB-NEXT: min a0, a0, a1 ; ZBB-NEXT: ret +; +; XQCI-LABEL: smin_i32_negone: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.ligei a0, a0, 0, -1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_i32_negone: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: li a1, -1 +; RV32I-SFB-NEXT: bltz a0, .LBB26_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a1 +; RV32I-SFB-NEXT: .LBB26_2: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_i32_negone: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a1, -1 +; RV64I-SFB-NEXT: bltz a0, .LBB26_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB26_2: +; RV64I-SFB-NEXT: ret %c = call i32 @llvm.smin.i32(i32 %a, i32 -1) ret i32 %c } @@ -672,6 +1243,34 @@ define i64 @smin_i64_negone(i64 %a) { ; RV64ZBB-NEXT: li a1, -1 ; RV64ZBB-NEXT: min a0, a0, a1 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: smin_i64_negone: +; XQCI: # %bb.0: +; XQCI-NEXT: qc.ligei a0, a1, 0, -1 +; XQCI-NEXT: qc.ligei a1, a1, 0, -1 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: smin_i64_negone: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: li a2, -1 +; RV32I-SFB-NEXT: bltz a1, .LBB27_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a0, a2 +; RV32I-SFB-NEXT: .LBB27_2: +; RV32I-SFB-NEXT: bltz a1, .LBB27_4 +; RV32I-SFB-NEXT: # %bb.3: +; RV32I-SFB-NEXT: mv a1, a2 +; RV32I-SFB-NEXT: .LBB27_4: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: smin_i64_negone: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a1, -1 +; RV64I-SFB-NEXT: bltz a0, .LBB27_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB27_2: +; RV64I-SFB-NEXT: ret %c = call i64 @llvm.smin.i64(i64 %a, i64 -1) ret i64 %c } @@ -720,6 +1319,41 @@ define i64 @umax_i64_one(i64 %a, i64 %b) { ; RV64ZBB-NEXT: li a1, 1 ; RV64ZBB-NEXT: maxu a0, a0, a1 ; RV64ZBB-NEXT: ret +; +; XQCI-LABEL: umax_i64_one: +; XQCI: # %bb.0: +; XQCI-NEXT: mv a2, a1 +; XQCI-NEXT: qc.selectinei a2, 0, a0, 1 +; XQCI-NEXT: qc.liltui a0, a0, 2, 1 +; XQCI-NEXT: qc.mvnei a0, a1, 0, a2 +; XQCI-NEXT: ret +; +; RV32I-SFB-LABEL: umax_i64_one: +; RV32I-SFB: # %bb.0: +; RV32I-SFB-NEXT: li a2, 1 +; RV32I-SFB-NEXT: li a3, 1 +; RV32I-SFB-NEXT: beqz a1, .LBB28_2 +; RV32I-SFB-NEXT: # %bb.1: +; RV32I-SFB-NEXT: mv a3, a0 +; RV32I-SFB-NEXT: .LBB28_2: +; RV32I-SFB-NEXT: bnez a0, .LBB28_4 +; RV32I-SFB-NEXT: # %bb.3: +; RV32I-SFB-NEXT: mv a0, a2 +; RV32I-SFB-NEXT: .LBB28_4: +; RV32I-SFB-NEXT: beqz a1, .LBB28_6 +; RV32I-SFB-NEXT: # %bb.5: +; RV32I-SFB-NEXT: mv a0, a3 +; RV32I-SFB-NEXT: .LBB28_6: +; RV32I-SFB-NEXT: ret +; +; RV64I-SFB-LABEL: umax_i64_one: +; RV64I-SFB: # %bb.0: +; RV64I-SFB-NEXT: li a1, 1 +; RV64I-SFB-NEXT: bnez a0, .LBB28_2 +; RV64I-SFB-NEXT: # %bb.1: +; RV64I-SFB-NEXT: mv a0, a1 +; RV64I-SFB-NEXT: .LBB28_2: +; RV64I-SFB-NEXT: ret %c = call i64 @llvm.umax.i64(i64 %a, i64 1) ret i64 %c } diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive-O0-ATM-ATK.ll b/llvm/test/CodeGen/RISCV/rvv/sifive-O0-ATM-ATK.ll new file mode 100644 index 0000000..d9a49a1 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive-O0-ATM-ATK.ll @@ -0,0 +1,18 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -mtriple=riscv64 -mattr=+v -O0 -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK-RV64 + +define void @matmul() { +; CHECK-RV64-LABEL: matmul: +; CHECK-RV64: # %bb.0: # %entry +; CHECK-RV64-NEXT: li a0, 0 +; CHECK-RV64-NEXT: vsetvli zero, a0, 512 +; CHECK-RV64-NEXT: sf.vsettm zero, a0 +; CHECK-RV64-NEXT: sf.vtzero.t mt0 +; CHECK-RV64-NEXT: ret +entry: + call void @llvm.riscv.sf.vtzero.t.i64(i64 0, i64 0, i64 0, i64 3, i64 1) + ret void +} + +; Function Attrs: nocallback nofree nosync nounwind willreturn +declare void @llvm.riscv.sf.vtzero.t.i64(i64 immarg, i64, i64, i64 immarg, i64 immarg) #0 diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e4m3_e4m3.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e4m3_e4m3.ll new file mode 100644 index 0000000..9b9a849 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e4m3_e4m3.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.e4m3.e4m3.iXLen.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_e4m3_e4m3_w4_u8m8_u8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_e4m3_e4m3_w4_u8m8_u8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.e4m3.e4m3 mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.e4m3.e4m3.iXLen.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e4m3_e5m2.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e4m3_e5m2.ll new file mode 100644 index 0000000..b63974f --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e4m3_e5m2.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.e4m3.e5m2.iXLen.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_e4m3_e5m2_w4_u8m8_u8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_e4m3_e5m2_w4_u8m8_u8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.e4m3.e5m2 mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.e4m3.e5m2.iXLen.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e5m2_e4m3.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e5m2_e4m3.ll new file mode 100644 index 0000000..62d629b1 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e5m2_e4m3.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.e5m2.e4m3.iXLen.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_e5m2_e5m2_w4_u8m8_u8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_e5m2_e5m2_w4_u8m8_u8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.e5m2.e4m3 mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.e5m2.e4m3.iXLen.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e5m2_e5m2.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e5m2_e5m2.ll new file mode 100644 index 0000000..7a90c97 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_e5m2_e5m2.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.e5m2.e5m2.iXLen.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_e4m3_e5m2_w4_u8m8_u8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_e4m3_e5m2_w4_u8m8_u8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.e5m2.e5m2 mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.e5m2.e5m2.iXLen.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_f_f.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_f_f.ll new file mode 100644 index 0000000..29451c6 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_f_f.ll @@ -0,0 +1,52 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+xsfmm32a32f -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+xsfmm32a32f -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.f.f.iXLen.nxv32f16(iXLen, <vscale x 32 x half>, <vscale x 32 x half>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_f_f_w2_f16m8(iXLen %mtd, <vscale x 32 x half> %v1, <vscale x 32 x half> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_f_f_w2_f16m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e16, w2 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.f.f mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.f.f.iXLen.nxv32f16(iXLen 0, <vscale x 32 x half> %v1, <vscale x 32 x half> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 2) + ret void +} + +declare void @llvm.riscv.sf.mm.f.f.iXLen.nxv16f32(iXLen, <vscale x 16 x float>, <vscale x 16 x float>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_f_f_w1_f32m8(iXLen %mtd, <vscale x 16 x float> %v1, <vscale x 16 x float> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_f_f_w1_f32m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e32, w1 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.f.f mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.f.f.iXLen.nxv16f32(iXLen 0, <vscale x 16 x float> %v1, <vscale x 16 x float> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 1) + ret void +} + +declare void @llvm.riscv.sf.mm.f.f.iXLen.nxv8f64(iXLen, <vscale x 8 x double>, <vscale x 8 x double>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_f_f_w1_f64m8(iXLen %mtd, <vscale x 8 x double> %v1, <vscale x 8 x double> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_f_f_w1_f64m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e64, w1 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.f.f mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.f.f.iXLen.nxv8f64(iXLen 0, <vscale x 8 x double> %v1, <vscale x 8 x double> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 1) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_s_s.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_s_s.ll new file mode 100644 index 0000000..6a4b29f --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_s_s.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.s.s.iXLen.nxv64i8.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_s_s_w4_i8m8_i8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_s_s_w4_i8m8_i8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.s.s mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.s.s.iXLen.nxv64i8.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_s_u.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_s_u.ll new file mode 100644 index 0000000..79239b0 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_s_u.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.s.u.iXLen.nxv64i8.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_s_u_w4_i8m8_i8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_s_u_w4_i8m8_i8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.s.u mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.s.u.iXLen.nxv64i8.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_u_s.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_u_s.ll new file mode 100644 index 0000000..b0d039b --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_u_s.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.u.s.iXLen.nxv64i8.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_u_s_w4_i8m8_i8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_u_s_w4_i8m8_i8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.u.s mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.u.s.iXLen.nxv64i8.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_u_u.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_u_u.ll new file mode 100644 index 0000000..913c277 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_mm_u_u.ll @@ -0,0 +1,20 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+xsfmm32a8i \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.mm.u.u.iXLen.nxv64i8.nxv64i8(iXLen, <vscale x 64 x i8>, <vscale x 64 x i8>, iXLen, iXLen, iXLen, iXLen) + +define void @test_sf_mm_u_u_w4_i8m8_i8m8(iXLen %mtd, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk) { +; CHECK-LABEL: test_sf_mm_u_u_w4_i8m8_i8m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a1 +; CHECK-NEXT: sf.vsettk zero, a3 +; CHECK-NEXT: sf.mm.u.u mt0, v8, v16 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.mm.u.u.iXLen.nxv64i8.nxv64i8(iXLen 0, <vscale x 64 x i8> %v1, <vscale x 64 x i8> %v2, iXLen %tm, iXLen %tn, iXLen %tk, iXLen 4) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte16.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte16.ll new file mode 100644 index 0000000..8048dec --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte16.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vlte16.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vlte16(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vlte16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e16, w1 +; CHECK-NEXT: sf.vlte16 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vlte16.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte32.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte32.ll new file mode 100644 index 0000000..a526dc8 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte32.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vlte32.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vlte32(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vlte32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e32, w1 +; CHECK-NEXT: sf.vlte32 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vlte32.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte64.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte64.ll new file mode 100644 index 0000000..ed0c48a --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte64.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vlte64.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vlte64(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vlte64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e64, w1 +; CHECK-NEXT: sf.vlte64 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vlte64.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte8.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte8.ll new file mode 100644 index 0000000..67b3ed2 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vlte8.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vlte8.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vlte8(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vlte8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w1 +; CHECK-NEXT: sf.vlte8 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vlte8.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettk.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettk.ll new file mode 100644 index 0000000..4da37fa --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettk.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare iXLen @llvm.riscv.sf.vsettk.iXLen(iXLen, iXLen, iXLen) + +define iXLen @test_sf_vsettk(iXLen %tk) { +; CHECK-LABEL: test_sf_vsettk: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a1, zero, e16, w2 +; CHECK-NEXT: sf.vsettk a0, a0 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettk.iXLen(iXLen %tk, iXLen 1, iXLen 2) + ret iXLen %0 +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettm.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettm.ll new file mode 100644 index 0000000..143c26c --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettm.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare iXLen @llvm.riscv.sf.vsettm.iXLen(iXLen, iXLen, iXLen) + +define iXLen @test_sf_vsettm(iXLen %tm) { +; CHECK-LABEL: test_sf_vsettm: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a1, zero, e8, w4 +; CHECK-NEXT: sf.vsettm a0, a0 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettm.iXLen(iXLen %tm, iXLen 0, iXLen 3) + ret iXLen %0 +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettnt.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettnt.ll new file mode 100644 index 0000000..48fa1bc8 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vsettnt.ll @@ -0,0 +1,72 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen, iXLen, iXLen) + +define iXLen @test_sf_vsettnt_e8w1(iXLen %tn) { +; CHECK-LABEL: test_sf_vsettnt_e8w1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a0, a0, e8, w1 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen %tn, iXLen 0, iXLen 1) + ret iXLen %0 +} + +define iXLen @test_sf_vsettnt_e8w2(iXLen %tn) { +; CHECK-LABEL: test_sf_vsettnt_e8w2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a0, a0, e8, w2 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen %tn, iXLen 0, iXLen 2) + ret iXLen %0 +} + +define iXLen @test_sf_vsettnt_e8w4(iXLen %tn) { +; CHECK-LABEL: test_sf_vsettnt_e8w4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a0, a0, e8, w4 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen %tn, iXLen 0, iXLen 3) + ret iXLen %0 +} + +define iXLen @test_sf_vsettnt_e16w1(iXLen %tn) { +; CHECK-LABEL: test_sf_vsettnt_e16w1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a0, a0, e16, w1 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen %tn, iXLen 1, iXLen 1) + ret iXLen %0 +} + +define iXLen @test_sf_vsettnt_e16w2(iXLen %tn) { +; CHECK-LABEL: test_sf_vsettnt_e16w2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a0, a0, e16, w2 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen %tn, iXLen 1, iXLen 2) + ret iXLen %0 +} + +define iXLen @test_sf_vsettnt_e16w4(iXLen %tn) { +; CHECK-LABEL: test_sf_vsettnt_e16w4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt a0, a0, e16, w4 +; CHECK-NEXT: ret + entry: + %0 = call iXLen @llvm.riscv.sf.vsettnt.iXLen(iXLen %tn, iXLen 1, iXLen 3) + ret iXLen %0 +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste16.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste16.ll new file mode 100644 index 0000000..7a76151 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste16.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vste16.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vste16(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vste16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e16, w1 +; CHECK-NEXT: sf.vste16 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vste16.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste32.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste32.ll new file mode 100644 index 0000000..8ff6e6a --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste32.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vste32.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vste32(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vste32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e32, w1 +; CHECK-NEXT: sf.vste32 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vste32.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste64.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste64.ll new file mode 100644 index 0000000..53990e4 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste64.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vste64.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vste64(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vste64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e64, w1 +; CHECK-NEXT: sf.vste64 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vste64.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste8.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste8.ll new file mode 100644 index 0000000..09b7259 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vste8.ll @@ -0,0 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vste8.iXLen(iXLen, ptr, iXLen) + +define dso_local void @test_sf_vste8(iXLen %tss, ptr %base, iXLen %vl) { +; CHECK-LABEL: test_sf_vste8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a2, e8, w1 +; CHECK-NEXT: sf.vste8 a0, (a1) +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vste8.iXLen(iXLen %tss, ptr %base, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtdiscard.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtdiscard.ll new file mode 100644 index 0000000..394eb60 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtdiscard.ll @@ -0,0 +1,22 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vtdiscard() + +define dso_local void @test_sf_vtdiscard() { +; CHECK-LABEL: test_sf_vtdiscard: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vtdiscard +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtdiscard() + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtmv_t_v.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtmv_t_v.ll new file mode 100644 index 0000000..66c9d26 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtmv_t_v.ll @@ -0,0 +1,114 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vtmv.t.v.nxv32bf16.iXLen(iXLen, <vscale x 32 x bfloat>, iXLen) + +define void @test_sf_vtmv_t_v_bf16m8(iXLen %tss, <vscale x 32 x bfloat> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_bf16m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e16, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv32bf16.iXLen(iXLen %tss, <vscale x 32 x bfloat> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv32f16.iXLen(iXLen, <vscale x 32 x half>, iXLen) + +define void @test_sf_vtmv_t_v_f16(iXLen %tss, <vscale x 32 x half> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e16, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv32f16.iXLen(iXLen %tss, <vscale x 32 x half> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv16f32.iXLen(iXLen, <vscale x 16 x float>, iXLen) + +define void @test_sf_vtmv_t_v_f32(iXLen %tss, <vscale x 16 x float> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e32, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv16f32.iXLen(iXLen %tss, <vscale x 16 x float> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv8f64.iXLen(iXLen, <vscale x 8 x double>, iXLen) + +define void @test_sf_vtmv_t_v_f64(iXLen %tss, <vscale x 8 x double> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e64, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv8f64.iXLen(iXLen %tss, <vscale x 8 x double> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv64i8.iXLen(iXLen, <vscale x 64 x i8>, iXLen) + +define void @test_sf_vtmv_t_v_i8(iXLen %tss, <vscale x 64 x i8> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e8, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv64i8.iXLen(iXLen %tss, <vscale x 64 x i8> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv32i16.iXLen(iXLen, <vscale x 32 x i16>, iXLen) + +define void @test_sf_vtmv_t_v_i16(iXLen %tss, <vscale x 32 x i16> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e16, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv32i16.iXLen(iXLen %tss, <vscale x 32 x i16> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv16i32.iXLen(iXLen, <vscale x 16 x i32>, iXLen) + +define void @test_sf_vtmv_t_v_i32(iXLen %tss, <vscale x 16 x i32> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e32, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv16i32.iXLen(iXLen %tss, <vscale x 16 x i32> %src, iXLen %vl) + ret void +} + +declare void @llvm.riscv.sf.vtmv.t.v.nxv8i64.iXLen(iXLen, <vscale x 8 x i64>, iXLen) + +define void @test_sf_vtmv_t_v_i64(iXLen %tss, <vscale x 8 x i64> %src, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_t_v_i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e64, w1 +; CHECK-NEXT: sf.vtmv.t.v a0, v8 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtmv.t.v.nxv8i64.iXLen(iXLen %tss, <vscale x 8 x i64> %src, iXLen %vl) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtmv_v_t.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtmv_v_t.ll new file mode 100644 index 0000000..0dcc2ab --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtmv_v_t.ll @@ -0,0 +1,114 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare <vscale x 32 x bfloat> @llvm.riscv.sf.vtmv.v.t.nxv32bf16.iXLen(iXLen, iXLen) + +define <vscale x 32 x bfloat> @test_sf_vtmv_v_t_bf16m8(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_bf16m8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e16, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 32 x bfloat> @llvm.riscv.sf.vtmv.v.t.nxv32bf16.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 32 x bfloat> %0 +} + +declare <vscale x 32 x half> @llvm.riscv.sf.vtmv.v.t.nxv32f16.iXLen(iXLen, iXLen) + +define <vscale x 32 x half> @test_sf_vtmv_v_t_f16(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e16, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 32 x half> @llvm.riscv.sf.vtmv.v.t.nxv32f16.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 32 x half> %0 +} + +declare <vscale x 16 x float> @llvm.riscv.sf.vtmv.v.t.nxv16f32.iXLen(iXLen, iXLen) + +define <vscale x 16 x float> @test_sf_vtmv_v_t_f32(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e32, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 16 x float> @llvm.riscv.sf.vtmv.v.t.nxv16f32.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 16 x float> %0 +} + +declare <vscale x 8 x double> @llvm.riscv.sf.vtmv.v.t.nxv8f64.iXLen(iXLen, iXLen) + +define <vscale x 8 x double> @test_sf_vtmv_v_t_f64(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e64, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 8 x double> @llvm.riscv.sf.vtmv.v.t.nxv8f64.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 8 x double> %0 +} + +declare <vscale x 64 x i8> @llvm.riscv.sf.vtmv.v.t.nxv64i8.iXLen(iXLen, iXLen) + +define <vscale x 64 x i8> @test_sf_vtmv_v_t_i8(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e8, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 64 x i8> @llvm.riscv.sf.vtmv.v.t.nxv64i8.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 64 x i8> %0 +} + +declare <vscale x 32 x i16> @llvm.riscv.sf.vtmv.v.t.nxv32i16.iXLen(iXLen, iXLen) + +define <vscale x 32 x i16> @test_sf_vtmv_v_t_i16(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e16, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 32 x i16> @llvm.riscv.sf.vtmv.v.t.nxv32i16.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 32 x i16> %0 +} + +declare <vscale x 16 x i32> @llvm.riscv.sf.vtmv.v.t.nxv16i32.iXLen(iXLen, iXLen) + +define <vscale x 16 x i32> @test_sf_vtmv_v_t_i32(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e32, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 16 x i32> @llvm.riscv.sf.vtmv.v.t.nxv16i32.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 16 x i32> %0 +} + +declare <vscale x 8 x i64> @llvm.riscv.sf.vtmv.v.t.nxv8i64.iXLen(iXLen, iXLen) + +define <vscale x 8 x i64> @test_sf_vtmv_v_t_i64(iXLen %tss, iXLen %vl) { +; CHECK-LABEL: test_sf_vtmv_v_t_i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e64, w1 +; CHECK-NEXT: sf.vtmv.v.t v8, a0 +; CHECK-NEXT: ret + entry: + %0 = call <vscale x 8 x i64> @llvm.riscv.sf.vtmv.v.t.nxv8i64.iXLen(iXLen %tss, iXLen %vl) + ret <vscale x 8 x i64> %0 +} diff --git a/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtzero_t.ll b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtzero_t.ll new file mode 100644 index 0000000..bbccb02 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/sifive_sf_vtzero_t.ll @@ -0,0 +1,24 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -mattr=+zvfh -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+experimental-zvfbfmin -mattr=+xsfmmbase \ +; RUN: -mattr=+xsfmm32a -mattr=+xsfmm32a8f -mattr=+xsfmm32a4i -mattr=+xsfmm64a64f \ +; RUN: -verify-machineinstrs | FileCheck %s --check-prefixes=CHECK + +declare void @llvm.riscv.sf.vtzero.t.iXLen(iXLen, iXLen, iXLen, iXLen, iXLen) +define void @test_sf_vtzero_t(iXLen %tm, iXLen %tn) { +; CHECK-LABEL: test_sf_vtzero_t: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sf.vsettnt zero, a1, e8, w4 +; CHECK-NEXT: sf.vsettm zero, a0 +; CHECK-NEXT: sf.vtzero.t mt0 +; CHECK-NEXT: ret + entry: + call void @llvm.riscv.sf.vtzero.t.iXLen(iXLen 0, iXLen %tm, iXLen %tn, iXLen 3, iXLen 4) + ret void +} + diff --git a/llvm/test/CodeGen/SPIRV/llvm-compiler-used.ll b/llvm/test/CodeGen/SPIRV/llvm-compiler-used.ll new file mode 100644 index 0000000..ddc2585 --- /dev/null +++ b/llvm/test/CodeGen/SPIRV/llvm-compiler-used.ll @@ -0,0 +1,19 @@ +; RUN: llc -verify-machineinstrs -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s +; RUN: %if spirv-tools %{ llc -mtriple=spirv-unknown-unknown %s -o - -filetype=obj | spirv-val %} +; RUN: llc -verify-machineinstrs -mtriple=spirv-unknown-vulkan %s -o - | FileCheck %s +; RUN: %if spirv-tools %{ llc -mtriple=spirv-unknown-vulkan %s -o - -filetype=obj | spirv-val %} + +; Verify that llvm.compiler.used is not lowered. +; CHECK: OpName %{{[0-9]+}} "unused" +; CHECK-NOT: OpName %{{[0-9]+}} "llvm.compiler.used" + +; Check that the type of llvm.compiler.used is not emitted too. +; CHECK-NOT: OpTypeArray + +@unused = private addrspace(3) global i32 0 +@llvm.compiler.used = appending addrspace(2) global [1 x ptr addrspace (4)] [ptr addrspace(4) addrspacecast (ptr addrspace(3) @unused to ptr addrspace(4))] + +define spir_func void @foo() { +entry: + ret void +} diff --git a/llvm/test/CodeGen/SystemZ/htm-intrinsics.ll b/llvm/test/CodeGen/SystemZ/htm-intrinsics.ll index c6ee804..07fbed9 100644 --- a/llvm/test/CodeGen/SystemZ/htm-intrinsics.ll +++ b/llvm/test/CodeGen/SystemZ/htm-intrinsics.ll @@ -90,7 +90,7 @@ define i32 @test_tbegin_nofloat4(i32 %pad, ptr %ptr) { ; CHECK: tbegin 0, 65292 ; CHECK: ipm %r2 ; CHECK: srl %r2, 28 -; CHECK: ciblh %r2, 2, 0(%r14) +; CHECK: bnhr %r14 ; CHECK: mvhi 0(%r3), 0 ; CHECK: br %r14 %res = call i32 @llvm.s390.tbegin.nofloat(ptr null, i32 65292) @@ -219,7 +219,7 @@ define i32 @test_tend2(i32 %pad, ptr %ptr) { ; CHECK: tend ; CHECK: ipm %r2 ; CHECK: srl %r2, 28 -; CHECK: ciblh %r2, 2, 0(%r14) +; CHECK: bnhr %r14 ; CHECK: mvhi 0(%r3), 0 ; CHECK: br %r14 %res = call i32 @llvm.s390.tend() diff --git a/llvm/test/CodeGen/SystemZ/inline-asm-flag-output-01.ll b/llvm/test/CodeGen/SystemZ/inline-asm-flag-output-01.ll new file mode 100644 index 0000000..6b8746e --- /dev/null +++ b/llvm/test/CodeGen/SystemZ/inline-asm-flag-output-01.ll @@ -0,0 +1,738 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc < %s -verify-machineinstrs -mtriple=s390x-linux-gnu -O2 | FileCheck %s +; Test implementation of combining br_ccmask for flag output operand, and +; optimizing ipm sequence using conditional branches. + +declare void @dummy() + +; Check a case where the cc is used as an integer. +; Just (srl (ipm)) sequence without optimization. +define i32 @test(ptr %a) { +; CHECK-LABEL: test: +; CHECK: # %bb.0: +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: ipm %r2 +; CHECK-NEXT: srl %r2, 28 +; CHECK-NEXT: br %r14 + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + ret i32 %cc +} + +; Test-1(f1_0_*). Test all 14 valid combinations, where cc is being used for +; branching. + +; Check (cc == 0). +define void @f1_0_eq_0(ptr %a) { +; CHECK-LABEL: f1_0_eq_0: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jge dummy@PLT +; CHECK-NEXT: .LBB1_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp eq i32 %cc, 0 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc != 0). +define void @f1_0_ne_0(ptr %a) { +; CHECK-LABEL: f1_0_ne_0: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgne dummy@PLT +; CHECK-NEXT: .LBB2_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp ugt i32 %cc, 0 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 1). +define void @f1_0_eq_1(ptr %a) { +; CHECK-LABEL: f1_0_eq_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgl dummy@PLT +; CHECK-NEXT: .LBB3_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp eq i32 %cc, 1 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc != 1). +define void @f1_0_ne_1(ptr %a) { +; CHECK-LABEL: f1_0_ne_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnl dummy@PLT +; CHECK-NEXT: .LBB4_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp ne i32 %cc, 1 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 2). +define void @f1_0_eq_2(ptr %a) { +; CHECK-LABEL: f1_0_eq_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgh dummy@PLT +; CHECK-NEXT: .LBB5_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp eq i32 %cc, 2 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc != 2). +define void @f1_0_ne_2(ptr %a) { +; CHECK-LABEL: f1_0_ne_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnh dummy@PLT +; CHECK-NEXT: .LBB6_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp ne i32 %cc, 2 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 3). +define void @f1_0_eq_3(ptr %a) { +; CHECK-LABEL: f1_0_eq_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgo dummy@PLT +; CHECK-NEXT: .LBB7_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp eq i32 %cc, 3 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc != 3). +define void @f1_0_ne_3(ptr %a) { +; CHECK-LABEL: f1_0_ne_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgno dummy@PLT +; CHECK-NEXT: .LBB8_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp ult i32 %cc, 3 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 0|1). +define void @f1_0_01(ptr %a) { +; CHECK-LABEL: f1_0_01: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgle dummy@PLT +; CHECK-NEXT: .LBB9_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp ult i32 %cc, 2 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 0|2). +define void @f1_0_02(ptr %a) { +; CHECK-LABEL: f1_0_02: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jghe dummy@PLT +; CHECK-NEXT: .LBB10_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cmp = icmp eq i32 %and, 0 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 0|3). +define void @f1_0_03(ptr %a) { +; CHECK-LABEL: f1_0_03: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnlh dummy@PLT +; CHECK-NEXT: .LBB11_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp0 = icmp ne i32 %cc, 0 + %cmp3 = icmp ne i32 %cc, 3 + %cmp.inv = and i1 %cmp0, %cmp3 + br i1 %cmp.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 1|2). +define void @f1_0_12(ptr %a) { +; CHECK-LABEL: f1_0_12: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jglh dummy@PLT +; CHECK-NEXT: .LBB12_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq1 = icmp eq i32 %cc, 1 + %cmpeq2 = icmp eq i32 %cc, 2 + %cmp = or i1 %cmpeq1, %cmpeq2 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 1|3). +define void @f1_0_13(ptr %a) { +; CHECK-LABEL: f1_0_13: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnhe dummy@PLT +; CHECK-NEXT: .LBB13_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq1 = icmp eq i32 %cc, 1 + %cmpeq3 = icmp eq i32 %cc, 3 + %cmp = or i1 %cmpeq1, %cmpeq3 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check (cc == 2|3). +define void @f1_0_23(ptr %a) { +; CHECK-LABEL: f1_0_23: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnle dummy@PLT +; CHECK-NEXT: .LBB14_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmp = icmp ugt i32 %cc, 1 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Test-2(f1_1_*/f1_2_*/fl_3_*/f1_4_*). +; Test Mixed patterns involving Binary Ops. + +; Check 'add' for (cc != 0). +define void @f1_1_1(ptr %a) { +; CHECK-LABEL: f1_1_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgne dummy@PLT +; CHECK-NEXT: .LBB15_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cmp = icmp ult i32 %add, 3 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'add' for (cc == 1|2). +define void @f1_1_2(ptr %a) { +; CHECK-LABEL: f1_1_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jglh dummy@PLT +; CHECK-NEXT: .LBB16_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cmp = icmp ult i32 %add, 2 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'add' for (cc == 1|2). +define void @f1_1_3(ptr %a) { +; CHECK-LABEL: f1_1_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jglh dummy@PLT +; CHECK-NEXT: .LBB17_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -3 + %cmp.inv = icmp ult i32 %add, -2 + br i1 %cmp.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'and' with one operand cc and other select_ccmask(cc !=1). +define void @f1_2_1(ptr %a) { +; CHECK-LABEL: f1_2_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnl dummy@PLT +; CHECK-NEXT: .LBB18_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpne0 = icmp ne i32 %andcc, 0 + %cmpne3 = icmp ne i32 %cc, 3 + %cmp.inv = and i1 %cmpne3, %cmpne0 + br i1 %cmp.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'and' with both operands select_ccmask(cc != 2). +define void @f1_2_2(ptr %a) { +; CHECK-LABEL: f1_2_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnh dummy@PLT +; CHECK-NEXT: .LBB19_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %ugt1 = icmp samesign ugt i32 %cc, 1 + %cmpne3 = icmp ne i32 %cc, 3 + %and.cond.inv = and i1 %ugt1, %cmpne3 + br i1 %and.cond.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'and/tm' for (cc == 0|2). +define void @f1_2_3(ptr %a) { +; CHECK-LABEL: f1_2_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jghe dummy@PLT +; CHECK-NEXT: .LBB20_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cmp = icmp eq i32 %and, 0 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'and/tm' for (cc == 1|3). +define void @f1_2_4(ptr %a) { +; CHECK-LABEL: f1_2_4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnhe dummy@PLT +; CHECK-NEXT: .LBB21_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cmp = icmp eq i32 %and, 0 + br i1 %cmp, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'icmp' with one operand 'and' and other 'select_ccmask'(cc != 1). +define void @f1_2_5(ptr %a) { +; CHECK-LABEL: f1_2_5: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnl dummy@PLT +; CHECK-NEXT: .LBB22_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne3 = icmp ne i32 %cc, 3 + %cmp = xor i1 %cmpne3, %trunc + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check nested 'xor' cc with select_ccmask(cc != 1). +define void @f1_3_1(ptr %a) { +; CHECK-LABEL: f1_3_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnl dummy@PLT +; CHECK-NEXT: .LBB23_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq0 = icmp eq i32 %cc, 0 + %cmpeq2 = icmp eq i32 %cc, 2 + %xor = xor i1 %cmpeq0, %cmpeq2 + %cmpne3 = icmp ne i32 %cc, 3 + %cmp.inv = xor i1 %cmpne3, %xor + br i1 %cmp.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=1). +define void @f1_3_2(ptr %a) { +; CHECK-LABEL: f1_3_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnl dummy@PLT +; CHECK-NEXT: .LBB24_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpeq3 = icmp eq i32 %cc, 3 + %cmp.inv = xor i1 %cmpeq3, %trunc + br i1 %cmp.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=2). +define void @f1_3_3(ptr %a) { +; CHECK-LABEL: f1_3_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgnh dummy@PLT +; CHECK-NEXT: .LBB25_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne0 = icmp ne i32 %cc, 0 + %cmp.cond.inv = xor i1 %cmpne0, %trunc + br i1 %cmp.cond.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'or' with both operands are select_ccmask one with TM and other with +; ICMP(cc == 1). +define void @f1_4_1(ptr %a) { +; CHECK-LABEL: f1_4_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgl dummy@PLT +; CHECK-NEXT: .LBB26_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpeq0 = icmp eq i32 %andcc, 0 + %cmpeq3 = icmp eq i32 %cc, 3 + %cmp.cond.inv = or i1 %cmpeq3, %cmpeq0 + br i1 %cmp.cond.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'or' for (cc == 0|1). +define void @f1_4_2(ptr %a) { +; CHECK-LABEL: f1_4_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgle dummy@PLT +; CHECK-NEXT: .LBB27_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cmp.inv = icmp samesign ugt i32 %or, -3 + br i1 %cmp.inv, label %exit, label %branch +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + +; Check 'or' for (cc == 0|1). +define void @f1_4_3(ptr %a) { +; CHECK-LABEL: f1_4_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: jgle dummy@PLT +; CHECK-NEXT: .LBB28_1: # %exit +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cmp = icmp samesign ult i32 %or, -2 + br i1 %cmp, label %branch, label %exit +branch: + tail call void @dummy() + br label %exit +exit: + ret void +} + diff --git a/llvm/test/CodeGen/SystemZ/inline-asm-flag-output-02.ll b/llvm/test/CodeGen/SystemZ/inline-asm-flag-output-02.ll new file mode 100644 index 0000000..b9b9a4b --- /dev/null +++ b/llvm/test/CodeGen/SystemZ/inline-asm-flag-output-02.ll @@ -0,0 +1,1665 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc < %s -verify-machineinstrs -mtriple=s390x-linux-gnu -O2 | FileCheck %s +; Test implementation of combining select_ccmask for flag output operand and +; optimizing ipm sequence using conditional branches. + +; Test-1(f2_0_*): Both TrueVal and FalseVal non-const(14-valid CCMask). + +; Check (cc == 0). +define i64 @f2_0_eq_0(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_eq_0: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: ber %r14 +; CHECK-NEXT: .LBB0_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp eq i32 %cc, 0 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc != 0). +define i64 @f2_0_ne_0(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_ne_0: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bner %r14 +; CHECK-NEXT: .LBB1_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp ugt i32 %cc, 0 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc == 1). +define i64 @f2_0_eq_1(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_eq_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB2_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp eq i32 %cc, 1 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc != 1). +define i64 @f2_0_ne_1(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_ne_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB3_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp ne i32 %cc, 1 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc == 2). +define i64 @f2_0_eq_2(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_eq_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB4_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp eq i32 %cc, 2 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc != 2). +define i64 @f2_0_ne_2(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_ne_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnhr %r14 +; CHECK-NEXT: .LBB5_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp ne i32 %cc, 2 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc == 3). +define i64 @f2_0_eq_3(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_eq_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bor %r14 +; CHECK-NEXT: .LBB6_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp eq i32 %cc, 3 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc != 3). +define i64 @f2_0_ne_3(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_ne_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnor %r14 +; CHECK-NEXT: .LBB7_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp ult i32 %cc, 3 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc == 0|1). +define i64 @f2_0_01(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_01: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bler %r14 +; CHECK-NEXT: .LBB8_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp ult i32 %cc, 2 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc == 0|2). +define i64 @f2_0_02(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_02: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB9_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %and = and i32 %cc, 1 + %cond = icmp eq i32 %and, 0 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check (cc == 0|3). +define i64 @f2_0_03(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_0_03: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blhr %r14 +; CHECK-NEXT: .LBB10_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cmp0 = icmp ne i32 %cc, 0 + %cmp3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %cmp0, %cmp3 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check (cc == 1|2). +define i64 @f2_0_12(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_0_12: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlhr %r14 +; CHECK-NEXT: .LBB11_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %add = add nsw i32 %cc, -3 + %cond.inv = icmp ult i32 %add, -2 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check (cc == 1|3). +define i64 @f2_0_13(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_0_13: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB12_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %and = and i32 %cc, 1 + %cond.inv = icmp eq i32 %and, 0 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check (cc == 2|3). +define i64 @f2_0_23(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_0_23: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnler %r14 +; CHECK-NEXT: .LBB13_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %cond = icmp ugt i32 %cc, 1 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Test-2(f2_1_*/f2_2_*/f2_3_*/f2_4_*). +; Both TrueVal and FalseVal are non-const with mixed patterns involving +; Binary Ops. + +; Check 'add' for (cc != 0). +define i64 @f2_1_1(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_1_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bner %r14 +; CHECK-NEXT: .LBB14_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 3 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f2_1_2(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_1_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blhr %r14 +; CHECK-NEXT: .LBB15_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 2 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f2_1_3(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_1_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlhr %r14 +; CHECK-NEXT: .LBB16_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -3 + %cond.inv = icmp ult i32 %add, -2 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'and' with one operand cc and other select_ccmask(cc !=1). +define i64 @f2_2_1(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_2_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB17_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpne0 = icmp ne i32 %andcc, 0 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %cmpne3, %cmpne0 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'and' with both operands select_ccmask(cc != 2). +define i64 @f2_2_2(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_2_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB18_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %ugt1 = icmp samesign ugt i32 %cc, 1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %ugt1, %cmpne3 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'and/tm' for (cc == 0|2). +define i64 @f2_2_3(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_2_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB19_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond = icmp eq i32 %and, 0 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Check 'and/tm' for (cc == 1|3). +define i64 @f2_2_4(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_2_4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB20_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond.inv = icmp eq i32 %and, 0 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'icmp' with one operand 'and' and other 'select_ccmask'(cc != 1). +define i64 @f2_2_5(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_2_5: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB21_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond = xor i1 %cmpne3, %trunc + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + + +; Check nested 'xor' cc with select_ccmask(cc != 1). +define i64 @f2_3_1(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_3_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB22_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq0 = icmp eq i32 %cc, 0 + %cmpeq2 = icmp eq i32 %cc, 2 + %xor = xor i1 %cmpeq0, %cmpeq2 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = xor i1 %cmpne3, %xor + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=1). +define i64 @f2_3_2(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_3_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB23_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = xor i1 %cmpeq3, %trunc + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=2). +define i64 @f2_3_3(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_3_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB24_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne0 = icmp ne i32 %cc, 0 + %cond.inv = xor i1 %cmpne0, %trunc + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'or' with both operands select_ccmask with TM and ICMP(cc == 1). +define i64 @f2_4_1(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_4_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB25_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpeq0 = icmp eq i32 %andcc, 0 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = or i1 %cmpeq3, %cmpeq0 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f2_4_2(i64 %y, i64 %x, ptr %a) { +; CHECK-LABEL: f2_4_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnler %r14 +; CHECK-NEXT: .LBB26_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond.inv = icmp samesign ugt i32 %or, -3 + %res = select i1 %cond.inv, i64 %y, i64 %x + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f2_4_3(i64 %x, i64 %y, ptr %a) { +; CHECK-LABEL: f2_4_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r4), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bler %r14 +; CHECK-NEXT: .LBB27_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond = icmp samesign ult i32 %or, -2 + %res = select i1 %cond, i64 %x, i64 %y + ret i64 %res +} + +; Test-3(f3_1_*/f3_2_*/f3_3_*/f3_4_*). +; TrueVal is non-const and FalseVal is const with mixed patterns involving +; Binary Ops. + +; Check 'add' for (cc != 0). +define i64 @f3_1_1(i64 %x, ptr %a) { +; CHECK-LABEL: f3_1_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bner %r14 +; CHECK-NEXT: .LBB28_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 3 + %res = select i1 %cond, i64 %x, i64 5 + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f3_1_2(i64 %x, ptr %a) { +; CHECK-LABEL: f3_1_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blhr %r14 +; CHECK-NEXT: .LBB29_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 2 + %res = select i1 %cond, i64 %x, i64 5 + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f3_1_3(ptr %a, i64 %x) { +; CHECK-LABEL: f3_1_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bnlhr %r14 +; CHECK-NEXT: .LBB30_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -3 + %cond.inv = icmp ult i32 %add, -2 + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'and' with one operand cc and other select_ccmask(cc !=1). +define i64 @f3_2_1(ptr %a, i64 %x) { +; CHECK-LABEL: f3_2_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB31_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpne0 = icmp ne i32 %andcc, 0 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %cmpne3, %cmpne0 + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'and' with both operands select_ccmask(cc != 2). +define i64 @f3_2_2(ptr %a, i64 %x) { +; CHECK-LABEL: f3_2_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB32_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %ugt1 = icmp samesign ugt i32 %cc, 1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %ugt1, %cmpne3 + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'and/tm' for (cc == 0|2). +define i64 @f3_2_3(i64 %x, ptr %a) { +; CHECK-LABEL: f3_2_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB33_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond = icmp eq i32 %and, 0 + %res = select i1 %cond, i64 %x, i64 5 + ret i64 %res +} + +; Check 'and/tm' for (cc == 1|3). +define i64 @f3_2_4(ptr %a, i64 %x) { +; CHECK-LABEL: f3_2_4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB34_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond.inv = icmp eq i32 %and, 0 + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'icmp' with one operand 'and' and other 'select_ccmask'(cc != 1). +define i64 @f3_2_5(i64 %x, ptr %a) { +; CHECK-LABEL: f3_2_5: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB35_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond = xor i1 %cmpne3, %trunc + %res = select i1 %cond, i64 %x, i64 5 + ret i64 %res +} + + +; Check nested 'xor' cc with select_ccmask(cc != 1). +define i64 @f3_3_1(ptr %a, i64 %x) { +; CHECK-LABEL: f3_3_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB36_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq0 = icmp eq i32 %cc, 0 + %cmpeq2 = icmp eq i32 %cc, 2 + %xor = xor i1 %cmpeq0, %cmpeq2 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = xor i1 %cmpne3, %xor + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=1). +define i64 @f3_3_2(ptr %a, i64 %x) { +; CHECK-LABEL: f3_3_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB37_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = xor i1 %cmpeq3, %trunc + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=2). +define i64 @f3_3_3(ptr %a, i64 %x) { +; CHECK-LABEL: f3_3_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB38_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne0 = icmp ne i32 %cc, 0 + %cond.inv = xor i1 %cmpne0, %trunc + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'or' with both operands select_ccmask with TM and ICMP(cc == 1). +define i64 @f3_4_1(ptr %a, i64 %x) { +; CHECK-LABEL: f3_4_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB39_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpeq0 = icmp eq i32 %andcc, 0 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = or i1 %cmpeq3, %cmpeq0 + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f3_4_2(ptr %a, i64 %x) { +; CHECK-LABEL: f3_4_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bnler %r14 +; CHECK-NEXT: .LBB40_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond.inv = icmp samesign ugt i32 %or, -3 + %res = select i1 %cond.inv, i64 5, i64 %x + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f3_4_3(i64 %x, ptr %a) { +; CHECK-LABEL: f3_4_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bler %r14 +; CHECK-NEXT: .LBB41_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond = icmp samesign ult i32 %or, -2 + %res = select i1 %cond, i64 %x, i64 5 + ret i64 %res +} + + +; Test-4(f4_1_*/f4_2_*/f4_3_*/f4_4_*). +; TrueVal is const and FalseVal is non-const with mixed patterns involving +; Binary Ops. + +; Check 'add' for (cc != 0). +define i64 @f4_1_1(ptr %a, i64 %y) { +; CHECK-LABEL: f4_1_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bner %r14 +; CHECK-NEXT: .LBB42_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 3 + %res = select i1 %cond, i64 15, i64 %y + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f4_1_2(ptr %a, i64 %y) { +; CHECK-LABEL: f4_1_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: blhr %r14 +; CHECK-NEXT: .LBB43_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 2 + %res = select i1 %cond, i64 15, i64 %y + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f4_1_3(i64 %y, ptr %a) { +; CHECK-LABEL: f4_1_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlhr %r14 +; CHECK-NEXT: .LBB44_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -3 + %cond.inv = icmp ult i32 %add, -2 + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'and' with one operand cc and other select_ccmask(cc !=1). +define i64 @f4_2_1(i64 %y, ptr %a) { +; CHECK-LABEL: f4_2_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB45_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpne0 = icmp ne i32 %andcc, 0 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %cmpne3, %cmpne0 + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'and' with both operands select_ccmask(cc != 2). +define i64 @f4_2_2(i64 %y, ptr %a) { +; CHECK-LABEL: f4_2_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB46_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %ugt1 = icmp samesign ugt i32 %cc, 1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %ugt1, %cmpne3 + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'and/tm' for (cc == 0|2). +define i64 @f4_2_3(ptr %a, i64 %y) { +; CHECK-LABEL: f4_2_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB47_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond = icmp eq i32 %and, 0 + %res = select i1 %cond, i64 15, i64 %y + ret i64 %res +} + +; Check 'and/tm' for (cc == 1|3). +define i64 @f4_2_4(i64 %y, ptr %a) { +; CHECK-LABEL: f4_2_4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB48_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond.inv = icmp eq i32 %and, 0 + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'icmp' with one operand 'and' and other 'select_ccmask'(cc != 1). +define i64 @f4_2_5(ptr %a, i64 %y) { +; CHECK-LABEL: f4_2_5: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB49_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond = xor i1 %cmpne3, %trunc + %res = select i1 %cond, i64 15, i64 %y + ret i64 %res +} + + +; Check nested 'xor' cc with select_ccmask(cc != 1). +define i64 @f4_3_1(i64 %y, ptr %a) { +; CHECK-LABEL: f4_3_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB50_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq0 = icmp eq i32 %cc, 0 + %cmpeq2 = icmp eq i32 %cc, 2 + %xor = xor i1 %cmpeq0, %cmpeq2 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = xor i1 %cmpne3, %xor + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=1). +define i64 @f4_3_2(i64 %y, ptr %a) { +; CHECK-LABEL: f4_3_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB51_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = xor i1 %cmpeq3, %trunc + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=2). +define i64 @f4_3_3(i64 %y, ptr %a) { +; CHECK-LABEL: f4_3_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB52_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne0 = icmp ne i32 %cc, 0 + %cond.inv = xor i1 %cmpne0, %trunc + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'or' with both operands select_ccmask with TM and ICMP(cc == 1). +define i64 @f4_4_1(i64 %y,ptr %a) { +; CHECK-LABEL: f4_4_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB53_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpeq0 = icmp eq i32 %andcc, 0 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = or i1 %cmpeq3, %cmpeq0 + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f4_4_2(i64 %y, ptr %a) { +; CHECK-LABEL: f4_4_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r3), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: bnler %r14 +; CHECK-NEXT: .LBB54_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond.inv = icmp samesign ugt i32 %or, -3 + %res = select i1 %cond.inv, i64 %y, i64 15 + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f4_4_3(ptr %a, i64 %y) { +; CHECK-LABEL: f4_4_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bler %r14 +; CHECK-NEXT: .LBB55_1: # %entry +; CHECK-NEXT: lgr %r2, %r3 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond = icmp samesign ult i32 %or, -2 + %res = select i1 %cond, i64 15, i64 %y + ret i64 %res +} + +; Test-5(f5_1_*/f5_2_*/f5_3_*/f5_4_*). +; Both TrueVal and FalseVal are const with mixed patterns involving +; Binary Ops. + + +; Check 'add' for (cc != 0). +define i64 @f5_1_1(ptr %a) { +; CHECK-LABEL: f5_1_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bner %r14 +; CHECK-NEXT: .LBB56_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 3 + %res = select i1 %cond, i64 15, i64 5 + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f5_1_2(ptr %a) { +; CHECK-LABEL: f5_1_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: blhr %r14 +; CHECK-NEXT: .LBB57_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -1 + %cond = icmp ult i32 %add, 2 + %res = select i1 %cond, i64 15, i64 5 + ret i64 %res +} + +; Check 'add' for (cc == 1|2). +define i64 @f5_1_3(ptr %a) { +; CHECK-LABEL: f5_1_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bnlhr %r14 +; CHECK-NEXT: .LBB58_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %add = add nsw i32 %cc, -3 + %cond.inv = icmp ult i32 %add, -2 + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'and' with one operand cc and other select_ccmask(cc !=1). +define i64 @f5_2_1(ptr %a) { +; CHECK-LABEL: f5_2_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB59_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpne0 = icmp ne i32 %andcc, 0 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %cmpne3, %cmpne0 + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'and' with both operands select_ccmask(cc != 2). +define i64 @f5_2_2(ptr %a) { +; CHECK-LABEL: f5_2_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB60_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %ugt1 = icmp samesign ugt i32 %cc, 1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = and i1 %ugt1, %cmpne3 + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'and/tm' for (cc == 0|2). +define i64 @f5_2_3(ptr %a) { +; CHECK-LABEL: f5_2_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB61_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond = icmp eq i32 %and, 0 + %res = select i1 %cond, i64 15, i64 5 + ret i64 %res +} + +; Check 'and/tm' for (cc == 1|3). +define i64 @f5_2_4(ptr %a) { +; CHECK-LABEL: f5_2_4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB62_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %and = and i32 %cc, 1 + %cond.inv = icmp eq i32 %and, 0 + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'icmp' with one operand 'and' and other 'select_ccmask'(cc != 1). +define i64 @f5_2_5(ptr %a) { +; CHECK-LABEL: f5_2_5: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB63_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne3 = icmp ne i32 %cc, 3 + %cond = xor i1 %cmpne3, %trunc + %res = select i1 %cond, i64 15, i64 5 + ret i64 %res +} + + +; Check nested 'xor' cc with select_ccmask(cc != 1). +define i64 @f5_3_1(ptr %a) { +; CHECK-LABEL: f5_3_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB64_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %cmpeq0 = icmp eq i32 %cc, 0 + %cmpeq2 = icmp eq i32 %cc, 2 + %xor = xor i1 %cmpeq0, %cmpeq2 + %cmpne3 = icmp ne i32 %cc, 3 + %cond.inv = xor i1 %cmpne3, %xor + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=1). +define i64 @f5_3_2(ptr %a) { +; CHECK-LABEL: f5_3_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: blr %r14 +; CHECK-NEXT: .LBB65_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = xor i1 %cmpeq3, %trunc + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check branching on 'tm' and 'xor' with one operand cc and the other +; select_ccmask(cc !=2). +define i64 @f5_3_3(ptr %a) { +; CHECK-LABEL: f5_3_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bhr %r14 +; CHECK-NEXT: .LBB66_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %trunc = trunc i32 %cc to i1 + %cmpne0 = icmp ne i32 %cc, 0 + %cond.inv = xor i1 %cmpne0, %trunc + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'or' with both operands select_ccmask with TM and ICMP(cc == 1). +define i64 @f5_4_1(ptr %a) { +; CHECK-LABEL: f5_4_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bnlr %r14 +; CHECK-NEXT: .LBB67_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %andcc = and i32 %cc, 1 + %cmpeq0 = icmp eq i32 %andcc, 0 + %cmpeq3 = icmp eq i32 %cc, 3 + %cond.inv = or i1 %cmpeq3, %cmpeq0 + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f5_4_2(ptr %a) { +; CHECK-LABEL: f5_4_2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: bnler %r14 +; CHECK-NEXT: .LBB68_1: # %entry +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond.inv = icmp samesign ugt i32 %or, -3 + %res = select i1 %cond.inv, i64 5, i64 15 + ret i64 %res +} + +; Check 'or' for (cc == 0|1). +define i64 @f5_4_3(ptr %a) { +; CHECK-LABEL: f5_4_3: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bler %r14 +; CHECK-NEXT: .LBB69_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %tmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %tmp) + %or = or disjoint i32 %cc, -4 + %cond = icmp samesign ult i32 %or, -2 + %res = select i1 %cond, i64 15, i64 5 + ret i64 %res +} + +; Nested select_ccmask with TrueVal and FalseVal swapped with each other. +define i64 @f6_1(ptr %a) { +; CHECK-LABEL: f6_1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: #APP +; CHECK-NEXT: alsi 0(%r2), -1 +; CHECK-EMPTY: +; CHECK-NEXT: #NO_APP +; CHECK-NEXT: lghi %r2, 15 +; CHECK-NEXT: bher %r14 +; CHECK-NEXT: .LBB70_1: # %entry +; CHECK-NEXT: lghi %r2, 5 +; CHECK-NEXT: br %r14 +entry: + %cc = tail call i32 asm sideeffect "alsi $1,-1\0A", "={@cc},=*QS,*QS,~{memory}"(ptr elementtype(i32) %a, ptr elementtype(i32) %a) + %cmp = icmp ult i32 %cc, 4 + tail call void @llvm.assume(i1 %cmp) + %andcc = and i32 %cc, 1 + %cmpeq0 = icmp eq i32 %andcc, 0 + %cmpeq3 = icmp eq i32 %cc, 3 + %select = select i1 %cmpeq3, i64 5, i64 15 + %res = select i1 %cmpeq0, i64 %select, i64 5 + ret i64 %res +} + diff --git a/llvm/test/Instrumentation/AddressSanitizer/asan-funclet.ll b/llvm/test/Instrumentation/AddressSanitizer/asan-funclet.ll index ae8b2b3..2a7216f 100644 --- a/llvm/test/Instrumentation/AddressSanitizer/asan-funclet.ll +++ b/llvm/test/Instrumentation/AddressSanitizer/asan-funclet.ll @@ -23,7 +23,7 @@ declare i32 @dummyPersonality(...) define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr @__CxxFrameHandler3 { ; CHECK-INLINE-LABEL: define void @FuncletPersonality( -; CHECK-INLINE-SAME: ptr [[PTRPARAM:%.*]]) #[[ATTR3:[0-9]+]] personality ptr @__CxxFrameHandler3 { +; CHECK-INLINE-SAME: ptr [[PTRPARAM:%.*]]) #[[ATTR2:[0-9]+]] personality ptr @__CxxFrameHandler3 { ; CHECK-INLINE-NEXT: entry: ; CHECK-INLINE-NEXT: [[TMP0:%.*]] = alloca i64, align 32 ; CHECK-INLINE-NEXT: store i64 0, ptr [[TMP0]], align 8 @@ -37,33 +37,26 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: br label [[TMP6]] ; CHECK-INLINE: 6: ; CHECK-INLINE-NEXT: [[TMP7:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP5]], [[TMP4]] ] +; CHECK-INLINE-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP7]] to ptr ; CHECK-INLINE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[TMP7]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP8]], label [[TMP9:%.*]], label [[TMP11:%.*]] -; CHECK-INLINE: 9: +; CHECK-INLINE: 10: ; CHECK-INLINE-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 8544, align 32 -; CHECK-INLINE-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; CHECK-INLINE-NEXT: br label [[TMP11]] ; CHECK-INLINE: 11: -; CHECK-INLINE-NEXT: [[TMP12:%.*]] = phi i64 [ [[TMP7]], [[TMP6]] ], [ [[TMP10]], [[TMP9]] ] -; CHECK-INLINE-NEXT: store i64 [[TMP12]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 -; CHECK-INLINE-NEXT: [[TMP13:%.*]] = add i64 [[TMP12]], 32 -; CHECK-INLINE-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr -; CHECK-INLINE-NEXT: [[TMP15:%.*]] = add i64 [[TMP12]], 8480 -; CHECK-INLINE-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr -; CHECK-INLINE-NEXT: [[TMP17:%.*]] = add i64 [[TMP12]], 8496 -; CHECK-INLINE-NEXT: [[TMP18:%.*]] = inttoptr i64 [[TMP17]] to ptr -; CHECK-INLINE-NEXT: [[TMP19:%.*]] = add i64 [[TMP12]], 8512 -; CHECK-INLINE-NEXT: [[TMP20:%.*]] = inttoptr i64 [[TMP19]] to ptr -; CHECK-INLINE-NEXT: [[TMP21:%.*]] = add i64 [[TMP12]], 8528 -; CHECK-INLINE-NEXT: [[TMP22:%.*]] = inttoptr i64 [[TMP21]] to ptr -; CHECK-INLINE-NEXT: [[TMP23:%.*]] = inttoptr i64 [[TMP12]] to ptr +; CHECK-INLINE-NEXT: [[TMP23:%.*]] = phi ptr [ [[TMP10]], [[TMP6]] ], [ [[MYALLOCA]], [[TMP9]] ] +; CHECK-INLINE-NEXT: store ptr [[TMP23]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 +; CHECK-INLINE-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[TMP23]], i64 32 +; CHECK-INLINE-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP23]], i64 8480 +; CHECK-INLINE-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP23]], i64 8496 +; CHECK-INLINE-NEXT: [[TMP20:%.*]] = getelementptr i8, ptr [[TMP23]], i64 8512 +; CHECK-INLINE-NEXT: [[TMP22:%.*]] = getelementptr i8, ptr [[TMP23]], i64 8528 ; CHECK-INLINE-NEXT: store i64 1102416563, ptr [[TMP23]], align 8 -; CHECK-INLINE-NEXT: [[TMP24:%.*]] = add i64 [[TMP12]], 8 -; CHECK-INLINE-NEXT: [[TMP25:%.*]] = inttoptr i64 [[TMP24]] to ptr +; CHECK-INLINE-NEXT: [[TMP25:%.*]] = getelementptr i8, ptr [[TMP23]], i64 8 ; CHECK-INLINE-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP25]], align 8 -; CHECK-INLINE-NEXT: [[TMP26:%.*]] = add i64 [[TMP12]], 16 -; CHECK-INLINE-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP26]] to ptr +; CHECK-INLINE-NEXT: [[TMP27:%.*]] = getelementptr i8, ptr [[TMP23]], i64 16 ; CHECK-INLINE-NEXT: store i64 ptrtoint (ptr @FuncletPersonality to i64), ptr [[TMP27]], align 8 +; CHECK-INLINE-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[TMP23]] to i64 ; CHECK-INLINE-NEXT: [[TMP28:%.*]] = lshr i64 [[TMP12]], 3 ; CHECK-INLINE-NEXT: [[TMP29:%.*]] = add i64 [[TMP28]], [[TMP1]] ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_f1(i64 [[TMP29]], i64 4) @@ -87,21 +80,22 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_f3(i64 [[TMP38]], i64 1) ; CHECK-INLINE-NEXT: [[TMP39:%.*]] = add i64 [[TMP29]], 1066 ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_04(i64 [[TMP39]], i64 1) +; CHECK-INLINE-NEXT: [[TMP21:%.*]] = ptrtoint ptr [[TMP22]] to i64 ; CHECK-INLINE-NEXT: [[TMP40:%.*]] = lshr i64 [[TMP21]], 3 ; CHECK-INLINE-NEXT: [[TMP41:%.*]] = add i64 [[TMP40]], [[TMP1]] ; CHECK-INLINE-NEXT: [[TMP42:%.*]] = inttoptr i64 [[TMP41]] to ptr ; CHECK-INLINE-NEXT: [[TMP43:%.*]] = load i8, ptr [[TMP42]], align 1 ; CHECK-INLINE-NEXT: [[TMP44:%.*]] = icmp ne i8 [[TMP43]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP44]], label [[TMP45:%.*]], label [[TMP50:%.*]], !prof [[PROF1:![0-9]+]] -; CHECK-INLINE: 45: +; CHECK-INLINE: 39: ; CHECK-INLINE-NEXT: [[TMP46:%.*]] = and i64 [[TMP21]], 7 ; CHECK-INLINE-NEXT: [[TMP47:%.*]] = trunc i64 [[TMP46]] to i8 ; CHECK-INLINE-NEXT: [[TMP48:%.*]] = icmp sge i8 [[TMP47]], [[TMP43]] ; CHECK-INLINE-NEXT: br i1 [[TMP48]], label [[TMP49:%.*]], label [[TMP50]] -; CHECK-INLINE: 49: +; CHECK-INLINE: 43: ; CHECK-INLINE-NEXT: call void @__asan_report_store1(i64 [[TMP21]]) #[[ATTR7:[0-9]+]] ; CHECK-INLINE-NEXT: unreachable -; CHECK-INLINE: 50: +; CHECK-INLINE: 44: ; CHECK-INLINE-NEXT: store volatile i8 0, ptr [[TMP22]], align 1 ; CHECK-INLINE-NEXT: [[TMP51:%.*]] = add i64 [[TMP29]], 1066 ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_f8(i64 [[TMP51]], i64 1) @@ -125,10 +119,10 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: [[TMP65:%.*]] = load i8, ptr [[TMP64]], align 1 ; CHECK-INLINE-NEXT: [[TMP66:%.*]] = icmp ne i8 [[TMP65]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP66]], label [[TMP67:%.*]], label [[TMP68:%.*]] -; CHECK-INLINE: 67: +; CHECK-INLINE: 61: ; CHECK-INLINE-NEXT: call void @__asan_report_store8(i64 [[TMP59]]) #[[ATTR7]] ; CHECK-INLINE-NEXT: unreachable -; CHECK-INLINE: 68: +; CHECK-INLINE: 62: ; CHECK-INLINE-NEXT: store volatile i64 0, ptr [[TMP61]], align 8 ; CHECK-INLINE-NEXT: [[TMPCOPYI64:%.*]] = load i64, ptr [[TMP61]], align 8 ; CHECK-INLINE-NEXT: [[TMP69:%.*]] = and i64 [[TMPCOPYI64]], 31 @@ -150,15 +144,15 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: [[TMP83:%.*]] = load i8, ptr [[TMP82]], align 1 ; CHECK-INLINE-NEXT: [[TMP84:%.*]] = icmp ne i8 [[TMP83]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP84]], label [[TMP85:%.*]], label [[TMP90:%.*]], !prof [[PROF1]] -; CHECK-INLINE: 85: +; CHECK-INLINE: 79: ; CHECK-INLINE-NEXT: [[TMP86:%.*]] = and i64 [[TMP77]], 7 ; CHECK-INLINE-NEXT: [[TMP87:%.*]] = trunc i64 [[TMP86]] to i8 ; CHECK-INLINE-NEXT: [[TMP88:%.*]] = icmp sge i8 [[TMP87]], [[TMP83]] ; CHECK-INLINE-NEXT: br i1 [[TMP88]], label [[TMP89:%.*]], label [[TMP90]] -; CHECK-INLINE: 89: +; CHECK-INLINE: 83: ; CHECK-INLINE-NEXT: call void @__asan_report_store1(i64 [[TMP77]]) #[[ATTR7]] ; CHECK-INLINE-NEXT: unreachable -; CHECK-INLINE: 90: +; CHECK-INLINE: 84: ; CHECK-INLINE-NEXT: store volatile i8 0, ptr [[TMP79]], align 1 ; CHECK-INLINE-NEXT: invoke void @MayThrowFunc() ; CHECK-INLINE-NEXT: to label [[INVOKE_CONT:%.*]] unwind label [[EHCLEANUP:%.*]] @@ -170,15 +164,15 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: store i64 1172321806, ptr [[TMP23]], align 8 ; CHECK-INLINE-NEXT: [[TMP93:%.*]] = icmp ne i64 [[TMP7]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP93]], label [[TMP94:%.*]], label [[TMP95:%.*]] -; CHECK-INLINE: 94: +; CHECK-INLINE: 88: ; CHECK-INLINE-NEXT: call void @__asan_stack_free_8(i64 [[TMP7]], i64 8544) ; CHECK-INLINE-NEXT: br label [[TMP97:%.*]] -; CHECK-INLINE: 95: +; CHECK-INLINE: 89: ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP29]], i64 4) ; CHECK-INLINE-NEXT: [[TMP96:%.*]] = add i64 [[TMP29]], 1028 ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP96]], i64 40) ; CHECK-INLINE-NEXT: br label [[TMP97]] -; CHECK-INLINE: 97: +; CHECK-INLINE: 91: ; CHECK-INLINE-NEXT: ret void ; CHECK-INLINE: ehcleanup: ; CHECK-INLINE-NEXT: [[TMP98:%.*]] = cleanuppad within none [] @@ -189,23 +183,27 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: [[TMP102:%.*]] = load i8, ptr [[TMP101]], align 1 ; CHECK-INLINE-NEXT: [[TMP103:%.*]] = icmp ne i8 [[TMP102]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP103]], label [[TMP104:%.*]], label [[TMP109:%.*]], !prof [[PROF1]] -; CHECK-INLINE: 104: +; CHECK-INLINE: 98: ; CHECK-INLINE-NEXT: [[TMP105:%.*]] = and i64 [[TMP54]], 7 ; CHECK-INLINE-NEXT: [[TMP106:%.*]] = trunc i64 [[TMP105]] to i8 ; CHECK-INLINE-NEXT: [[TMP107:%.*]] = icmp sge i8 [[TMP106]], [[TMP102]] ; CHECK-INLINE-NEXT: br i1 [[TMP107]], label [[TMP108:%.*]], label [[TMP109]] -; CHECK-INLINE: 108: +; CHECK-INLINE: 102: ; CHECK-INLINE-NEXT: call void @__asan_report_store1(i64 [[TMP54]]) #[[ATTR7]] [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: unreachable -; CHECK-INLINE: 109: +; CHECK-INLINE: 103: ; CHECK-INLINE-NEXT: store volatile i8 0, ptr [[TMP56]], align 1 ; CHECK-INLINE-NEXT: call void @__asan_poison_stack_memory(i64 [[TMP54]], i64 4) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: call void @DeInit(ptr [[TMP14]]) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: [[TMP110:%.*]] = call ptr @__asan_memset(ptr [[TMP16]], i32 0, i64 4) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: [[TMP111:%.*]] = call ptr @__asan_memcpy(ptr [[TMP18]], ptr [[TMP16]], i64 4) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: [[TMP112:%.*]] = call ptr @__asan_memmove(ptr [[TMP20]], ptr [[TMP16]], i64 4) [ "funclet"(token [[TMP98]]) ] +; CHECK-INLINE-NEXT: [[TMP15:%.*]] = ptrtoint ptr [[TMP16]] to i64 +; CHECK-INLINE-NEXT: [[TMP17:%.*]] = ptrtoint ptr [[TMP18]] to i64 ; CHECK-INLINE-NEXT: call void @__sanitizer_ptr_cmp(i64 [[TMP15]], i64 [[TMP17]]) [ "funclet"(token [[TMP98]]) ] -; CHECK-INLINE-NEXT: call void @__sanitizer_ptr_sub(i64 [[TMP15]], i64 [[TMP17]]) [ "funclet"(token [[TMP98]]) ] +; CHECK-INLINE-NEXT: [[ADDR1:%.*]] = ptrtoint ptr [[TMP16]] to i64 +; CHECK-INLINE-NEXT: [[ADDR2:%.*]] = ptrtoint ptr [[TMP18]] to i64 +; CHECK-INLINE-NEXT: call void @__sanitizer_ptr_sub(i64 [[ADDR1]], i64 [[ADDR2]]) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: [[TMP113:%.*]] = ptrtoint ptr [[PTRPARAM]] to i64 ; CHECK-INLINE-NEXT: [[TMP114:%.*]] = add i64 [[TMP113]], 7 ; CHECK-INLINE-NEXT: [[TMP115:%.*]] = inttoptr i64 [[TMP114]] to ptr @@ -216,27 +214,27 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: [[TMP120:%.*]] = load i8, ptr [[TMP119]], align 1 ; CHECK-INLINE-NEXT: [[TMP121:%.*]] = icmp ne i8 [[TMP120]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP121]], label [[TMP122:%.*]], label [[TMP127:%.*]], !prof [[PROF1]] -; CHECK-INLINE: 122: +; CHECK-INLINE: 118: ; CHECK-INLINE-NEXT: [[TMP123:%.*]] = and i64 [[TMP116]], 7 ; CHECK-INLINE-NEXT: [[TMP124:%.*]] = trunc i64 [[TMP123]] to i8 ; CHECK-INLINE-NEXT: [[TMP125:%.*]] = icmp sge i8 [[TMP124]], [[TMP120]] ; CHECK-INLINE-NEXT: br i1 [[TMP125]], label [[TMP126:%.*]], label [[TMP127]] -; CHECK-INLINE: 126: +; CHECK-INLINE: 122: ; CHECK-INLINE-NEXT: call void @__asan_report_store_n(i64 [[TMP116]], i64 8) #[[ATTR7]] [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: unreachable -; CHECK-INLINE: 127: +; CHECK-INLINE: 123: ; CHECK-INLINE-NEXT: [[TMP128:%.*]] = lshr i64 [[TMP114]], 3 ; CHECK-INLINE-NEXT: [[TMP129:%.*]] = add i64 [[TMP128]], [[TMP1]] ; CHECK-INLINE-NEXT: [[TMP130:%.*]] = inttoptr i64 [[TMP129]] to ptr ; CHECK-INLINE-NEXT: [[TMP131:%.*]] = load i8, ptr [[TMP130]], align 1 ; CHECK-INLINE-NEXT: [[TMP132:%.*]] = icmp ne i8 [[TMP131]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP132]], label [[TMP133:%.*]], label [[EHEXIT:%.*]], !prof [[PROF1]] -; CHECK-INLINE: 133: +; CHECK-INLINE: 129: ; CHECK-INLINE-NEXT: [[TMP134:%.*]] = and i64 [[TMP114]], 7 ; CHECK-INLINE-NEXT: [[TMP135:%.*]] = trunc i64 [[TMP134]] to i8 ; CHECK-INLINE-NEXT: [[TMP136:%.*]] = icmp sge i8 [[TMP135]], [[TMP131]] ; CHECK-INLINE-NEXT: br i1 [[TMP136]], label [[TMP137:%.*]], label [[EHEXIT]] -; CHECK-INLINE: 137: +; CHECK-INLINE: 133: ; CHECK-INLINE-NEXT: call void @__asan_report_store_n(i64 [[TMP114]], i64 8) #[[ATTR7]] [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: unreachable ; CHECK-INLINE: ehexit: @@ -249,19 +247,19 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-INLINE-NEXT: store i64 1172321806, ptr [[TMP23]], align 8 ; CHECK-INLINE-NEXT: [[TMP142:%.*]] = icmp ne i64 [[TMP7]], 0 ; CHECK-INLINE-NEXT: br i1 [[TMP142]], label [[TMP143:%.*]], label [[TMP144:%.*]] -; CHECK-INLINE: 143: +; CHECK-INLINE: 139: ; CHECK-INLINE-NEXT: call void @__asan_stack_free_8(i64 [[TMP7]], i64 8544) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: br label [[TMP146:%.*]] -; CHECK-INLINE: 144: +; CHECK-INLINE: 140: ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP29]], i64 4) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: [[TMP145:%.*]] = add i64 [[TMP29]], 1028 ; CHECK-INLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP145]], i64 40) [ "funclet"(token [[TMP98]]) ] ; CHECK-INLINE-NEXT: br label [[TMP146]] -; CHECK-INLINE: 146: +; CHECK-INLINE: 142: ; CHECK-INLINE-NEXT: cleanupret from [[TMP98]] unwind to caller ; ; CHECK-OUTLINE-LABEL: define void @FuncletPersonality( -; CHECK-OUTLINE-SAME: ptr [[PTRPARAM:%.*]]) #[[ATTR3:[0-9]+]] personality ptr @__CxxFrameHandler3 { +; CHECK-OUTLINE-SAME: ptr [[PTRPARAM:%.*]]) #[[ATTR2:[0-9]+]] personality ptr @__CxxFrameHandler3 { ; CHECK-OUTLINE-NEXT: entry: ; CHECK-OUTLINE-NEXT: [[TMP0:%.*]] = alloca i64, align 32 ; CHECK-OUTLINE-NEXT: store i64 0, ptr [[TMP0]], align 8 @@ -275,37 +273,28 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-OUTLINE-NEXT: br label [[TMP6]] ; CHECK-OUTLINE: 6: ; CHECK-OUTLINE-NEXT: [[TMP7:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP5]], [[TMP4]] ] +; CHECK-OUTLINE-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP7]] to ptr ; CHECK-OUTLINE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[TMP7]], 0 ; CHECK-OUTLINE-NEXT: br i1 [[TMP8]], label [[TMP9:%.*]], label [[TMP11:%.*]] -; CHECK-OUTLINE: 9: +; CHECK-OUTLINE: 10: ; CHECK-OUTLINE-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 8608, align 32 -; CHECK-OUTLINE-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; CHECK-OUTLINE-NEXT: br label [[TMP11]] ; CHECK-OUTLINE: 11: -; CHECK-OUTLINE-NEXT: [[TMP12:%.*]] = phi i64 [ [[TMP7]], [[TMP6]] ], [ [[TMP10]], [[TMP9]] ] -; CHECK-OUTLINE-NEXT: store i64 [[TMP12]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 -; CHECK-OUTLINE-NEXT: [[TMP13:%.*]] = add i64 [[TMP12]], 32 -; CHECK-OUTLINE-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP15:%.*]] = add i64 [[TMP12]], 8480 -; CHECK-OUTLINE-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP17:%.*]] = add i64 [[TMP12]], 8496 -; CHECK-OUTLINE-NEXT: [[TMP18:%.*]] = inttoptr i64 [[TMP17]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP19:%.*]] = add i64 [[TMP12]], 8512 -; CHECK-OUTLINE-NEXT: [[TMP20:%.*]] = inttoptr i64 [[TMP19]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP21:%.*]] = add i64 [[TMP12]], 8528 -; CHECK-OUTLINE-NEXT: [[TMP22:%.*]] = inttoptr i64 [[TMP21]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP23:%.*]] = add i64 [[TMP12]], 8544 -; CHECK-OUTLINE-NEXT: [[TMP24:%.*]] = inttoptr i64 [[TMP23]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP25:%.*]] = add i64 [[TMP12]], 8560 -; CHECK-OUTLINE-NEXT: [[TMP26:%.*]] = inttoptr i64 [[TMP25]] to ptr -; CHECK-OUTLINE-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP12]] to ptr +; CHECK-OUTLINE-NEXT: [[TMP27:%.*]] = phi ptr [ [[TMP10]], [[TMP6]] ], [ [[MYALLOCA]], [[TMP9]] ] +; CHECK-OUTLINE-NEXT: store ptr [[TMP27]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 +; CHECK-OUTLINE-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[TMP27]], i64 32 +; CHECK-OUTLINE-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8480 +; CHECK-OUTLINE-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8496 +; CHECK-OUTLINE-NEXT: [[TMP20:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8512 +; CHECK-OUTLINE-NEXT: [[TMP22:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8528 +; CHECK-OUTLINE-NEXT: [[TMP24:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8544 +; CHECK-OUTLINE-NEXT: [[TMP26:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8560 ; CHECK-OUTLINE-NEXT: store i64 1102416563, ptr [[TMP27]], align 8 -; CHECK-OUTLINE-NEXT: [[TMP28:%.*]] = add i64 [[TMP12]], 8 -; CHECK-OUTLINE-NEXT: [[TMP29:%.*]] = inttoptr i64 [[TMP28]] to ptr +; CHECK-OUTLINE-NEXT: [[TMP29:%.*]] = getelementptr i8, ptr [[TMP27]], i64 8 ; CHECK-OUTLINE-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP29]], align 8 -; CHECK-OUTLINE-NEXT: [[TMP30:%.*]] = add i64 [[TMP12]], 16 -; CHECK-OUTLINE-NEXT: [[TMP31:%.*]] = inttoptr i64 [[TMP30]] to ptr +; CHECK-OUTLINE-NEXT: [[TMP31:%.*]] = getelementptr i8, ptr [[TMP27]], i64 16 ; CHECK-OUTLINE-NEXT: store i64 ptrtoint (ptr @FuncletPersonality to i64), ptr [[TMP31]], align 8 +; CHECK-OUTLINE-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[TMP27]] to i64 ; CHECK-OUTLINE-NEXT: [[TMP32:%.*]] = lshr i64 [[TMP12]], 3 ; CHECK-OUTLINE-NEXT: [[TMP33:%.*]] = add i64 [[TMP32]], [[TMP1]] ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_f1(i64 [[TMP33]], i64 4) @@ -335,10 +324,12 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_f3(i64 [[TMP45]], i64 5) ; CHECK-OUTLINE-NEXT: [[TMP46:%.*]] = add i64 [[TMP33]], 1066 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_04(i64 [[TMP46]], i64 1) +; CHECK-OUTLINE-NEXT: [[TMP21:%.*]] = ptrtoint ptr [[TMP22]] to i64 ; CHECK-OUTLINE-NEXT: call void @__asan_store1(i64 [[TMP21]]) ; CHECK-OUTLINE-NEXT: store volatile i8 0, ptr [[TMP22]], align 1 ; CHECK-OUTLINE-NEXT: [[TMP47:%.*]] = add i64 [[TMP33]], 1066 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_f8(i64 [[TMP47]], i64 1) +; CHECK-OUTLINE-NEXT: [[TMP25:%.*]] = ptrtoint ptr [[TMP26]] to i64 ; CHECK-OUTLINE-NEXT: call void @__asan_store8(i64 [[TMP25]]) ; CHECK-OUTLINE-NEXT: store volatile i64 0, ptr [[TMP26]], align 8 ; CHECK-OUTLINE-NEXT: [[TMPCOPYI64:%.*]] = load i64, ptr [[TMP26]], align 8 @@ -367,22 +358,23 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-OUTLINE-NEXT: store i64 1172321806, ptr [[TMP27]], align 8 ; CHECK-OUTLINE-NEXT: [[TMP61:%.*]] = icmp ne i64 [[TMP7]], 0 ; CHECK-OUTLINE-NEXT: br i1 [[TMP61]], label [[TMP62:%.*]], label [[TMP63:%.*]] -; CHECK-OUTLINE: 62: +; CHECK-OUTLINE: 55: ; CHECK-OUTLINE-NEXT: call void @__asan_stack_free_8(i64 [[TMP7]], i64 8608) ; CHECK-OUTLINE-NEXT: br label [[TMP66:%.*]] -; CHECK-OUTLINE: 63: +; CHECK-OUTLINE: 56: ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP33]], i64 4) ; CHECK-OUTLINE-NEXT: [[TMP64:%.*]] = add i64 [[TMP33]], 1028 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP64]], i64 42) ; CHECK-OUTLINE-NEXT: [[TMP65:%.*]] = add i64 [[TMP33]], 1071 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP65]], i64 5) ; CHECK-OUTLINE-NEXT: br label [[TMP66]] -; CHECK-OUTLINE: 66: +; CHECK-OUTLINE: 59: ; CHECK-OUTLINE-NEXT: ret void ; CHECK-OUTLINE: ehcleanup: ; CHECK-OUTLINE-NEXT: [[TMP67:%.*]] = cleanuppad within none [] ; CHECK-OUTLINE-NEXT: [[TMP68:%.*]] = add i64 [[TMP33]], 1068 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_04(i64 [[TMP68]], i64 1) [ "funclet"(token [[TMP67]]) ] +; CHECK-OUTLINE-NEXT: [[TMP23:%.*]] = ptrtoint ptr [[TMP24]] to i64 ; CHECK-OUTLINE-NEXT: call void @__asan_store1(i64 [[TMP23]]) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: store volatile i8 0, ptr [[TMP24]], align 1 ; CHECK-OUTLINE-NEXT: [[TMP69:%.*]] = add i64 [[TMP33]], 1068 @@ -391,8 +383,12 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-OUTLINE-NEXT: [[TMP70:%.*]] = call ptr @__asan_memset(ptr [[TMP16]], i32 0, i64 4) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: [[TMP71:%.*]] = call ptr @__asan_memcpy(ptr [[TMP18]], ptr [[TMP16]], i64 4) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: [[TMP72:%.*]] = call ptr @__asan_memmove(ptr [[TMP20]], ptr [[TMP16]], i64 4) [ "funclet"(token [[TMP67]]) ] +; CHECK-OUTLINE-NEXT: [[TMP15:%.*]] = ptrtoint ptr [[TMP16]] to i64 +; CHECK-OUTLINE-NEXT: [[TMP17:%.*]] = ptrtoint ptr [[TMP18]] to i64 ; CHECK-OUTLINE-NEXT: call void @__sanitizer_ptr_cmp(i64 [[TMP15]], i64 [[TMP17]]) [ "funclet"(token [[TMP67]]) ] -; CHECK-OUTLINE-NEXT: call void @__sanitizer_ptr_sub(i64 [[TMP15]], i64 [[TMP17]]) [ "funclet"(token [[TMP67]]) ] +; CHECK-OUTLINE-NEXT: [[ADDR1:%.*]] = ptrtoint ptr [[TMP16]] to i64 +; CHECK-OUTLINE-NEXT: [[ADDR2:%.*]] = ptrtoint ptr [[TMP18]] to i64 +; CHECK-OUTLINE-NEXT: call void @__sanitizer_ptr_sub(i64 [[ADDR1]], i64 [[ADDR2]]) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: [[TMP73:%.*]] = ptrtoint ptr [[PTRPARAM]] to i64 ; CHECK-OUTLINE-NEXT: call void @__asan_storeN(i64 [[TMP73]], i64 8) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: store i64 0, ptr [[PTRPARAM]], align 1 @@ -404,17 +400,17 @@ define void @FuncletPersonality(ptr %ptrParam) sanitize_address personality ptr ; CHECK-OUTLINE-NEXT: store i64 1172321806, ptr [[TMP27]], align 8 ; CHECK-OUTLINE-NEXT: [[TMP78:%.*]] = icmp ne i64 [[TMP7]], 0 ; CHECK-OUTLINE-NEXT: br i1 [[TMP78]], label [[TMP79:%.*]], label [[TMP80:%.*]] -; CHECK-OUTLINE: 79: +; CHECK-OUTLINE: 75: ; CHECK-OUTLINE-NEXT: call void @__asan_stack_free_8(i64 [[TMP7]], i64 8608) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: br label [[TMP83:%.*]] -; CHECK-OUTLINE: 80: +; CHECK-OUTLINE: 76: ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP33]], i64 4) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: [[TMP81:%.*]] = add i64 [[TMP33]], 1028 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP81]], i64 42) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: [[TMP82:%.*]] = add i64 [[TMP33]], 1071 ; CHECK-OUTLINE-NEXT: call void @__asan_set_shadow_00(i64 [[TMP82]], i64 5) [ "funclet"(token [[TMP67]]) ] ; CHECK-OUTLINE-NEXT: br label [[TMP83]] -; CHECK-OUTLINE: 83: +; CHECK-OUTLINE: 79: ; CHECK-OUTLINE-NEXT: cleanupret from [[TMP67]] unwind to caller ; @@ -487,7 +483,7 @@ nopredecessor: ; Non-Windows personality, ensure no funclet gets attached to asan runtime call. define void @OtherPersonality(ptr %ptrParam) sanitize_address personality ptr @dummyPersonality { ; CHECK-LABEL: define void @OtherPersonality( -; CHECK-SAME: ptr [[PTRPARAM:%.*]]) #[[ATTR3:[0-9]+]] personality ptr @dummyPersonality { +; CHECK-SAME: ptr [[PTRPARAM:%.*]]) #[[ATTR2:[0-9]+]] personality ptr @dummyPersonality { ; CHECK-NEXT: entry: ; CHECK-NEXT: [[TMP0:%.*]] = load i64, ptr @__asan_shadow_memory_dynamic_address, align 8 ; CHECK-NEXT: [[ASAN_LOCAL_STACK_BASE:%.*]] = alloca i64, align 8 @@ -499,25 +495,22 @@ define void @OtherPersonality(ptr %ptrParam) sanitize_address personality ptr @d ; CHECK-NEXT: br label [[TMP5]] ; CHECK: 5: ; CHECK-NEXT: [[TMP6:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP4]], [[TMP3]] ] +; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP6]] to ptr ; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[TMP6]], 0 ; CHECK-NEXT: br i1 [[TMP7]], label [[TMP8:%.*]], label [[TMP10:%.*]] -; CHECK: 8: +; CHECK: 9: ; CHECK-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 64, align 32 -; CHECK-NEXT: [[TMP9:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; CHECK-NEXT: br label [[TMP10]] ; CHECK: 10: -; CHECK-NEXT: [[TMP11:%.*]] = phi i64 [ [[TMP6]], [[TMP5]] ], [ [[TMP9]], [[TMP8]] ] -; CHECK-NEXT: store i64 [[TMP11]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 -; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP11]], 32 -; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr -; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP11]] to ptr +; CHECK-NEXT: [[TMP14:%.*]] = phi ptr [ [[TMP9]], [[TMP5]] ], [ [[MYALLOCA]], [[TMP8]] ] +; CHECK-NEXT: store ptr [[TMP14]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 +; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[TMP14]], i64 32 ; CHECK-NEXT: store i64 1102416563, ptr [[TMP14]], align 8 -; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP11]], 8 -; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr +; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i8, ptr [[TMP14]], i64 8 ; CHECK-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack.1 to i64), ptr [[TMP16]], align 8 -; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP11]], 16 -; CHECK-NEXT: [[TMP18:%.*]] = inttoptr i64 [[TMP17]] to ptr +; CHECK-NEXT: [[TMP18:%.*]] = getelementptr i8, ptr [[TMP14]], i64 16 ; CHECK-NEXT: store i64 ptrtoint (ptr @OtherPersonality to i64), ptr [[TMP18]], align 8 +; CHECK-NEXT: [[TMP11:%.*]] = ptrtoint ptr [[TMP14]] to i64 ; CHECK-NEXT: [[TMP19:%.*]] = lshr i64 [[TMP11]], 3 ; CHECK-NEXT: [[TMP20:%.*]] = add i64 [[TMP19]], [[TMP0]] ; CHECK-NEXT: [[TMP21:%.*]] = add i64 [[TMP20]], 0 @@ -532,14 +525,14 @@ define void @OtherPersonality(ptr %ptrParam) sanitize_address personality ptr @d ; CHECK-NEXT: store i64 1172321806, ptr [[TMP14]], align 8 ; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP6]], 0 ; CHECK-NEXT: br i1 [[TMP24]], label [[TMP25:%.*]], label [[TMP26:%.*]] -; CHECK: 25: +; CHECK: 22: ; CHECK-NEXT: call void @__asan_stack_free_0(i64 [[TMP6]], i64 64) ; CHECK-NEXT: br label [[TMP28:%.*]] -; CHECK: 26: +; CHECK: 23: ; CHECK-NEXT: [[TMP27:%.*]] = add i64 [[TMP20]], 0 ; CHECK-NEXT: call void @__asan_set_shadow_00(i64 [[TMP27]], i64 8) ; CHECK-NEXT: br label [[TMP28]] -; CHECK: 28: +; CHECK: 25: ; CHECK-NEXT: ret void ; CHECK: ehcleanup: ; CHECK-NEXT: [[TMP29:%.*]] = cleanuppad within none [] @@ -547,14 +540,14 @@ define void @OtherPersonality(ptr %ptrParam) sanitize_address personality ptr @d ; CHECK-NEXT: store i64 1172321806, ptr [[TMP14]], align 8 ; CHECK-NEXT: [[TMP31:%.*]] = icmp ne i64 [[TMP6]], 0 ; CHECK-NEXT: br i1 [[TMP31]], label [[TMP32:%.*]], label [[TMP33:%.*]] -; CHECK: 32: +; CHECK: 29: ; CHECK-NEXT: call void @__asan_stack_free_0(i64 [[TMP6]], i64 64) ; CHECK-NEXT: br label [[TMP35:%.*]] -; CHECK: 33: +; CHECK: 30: ; CHECK-NEXT: [[TMP34:%.*]] = add i64 [[TMP20]], 0 ; CHECK-NEXT: call void @__asan_set_shadow_00(i64 [[TMP34]], i64 8) ; CHECK-NEXT: br label [[TMP35]] -; CHECK: 35: +; CHECK: 32: ; CHECK-NEXT: cleanupret from [[TMP29]] unwind to caller ; entry: diff --git a/llvm/test/Instrumentation/AddressSanitizer/fake-stack.ll b/llvm/test/Instrumentation/AddressSanitizer/fake-stack.ll index 3cccabb1..1b00cd8 100644 --- a/llvm/test/Instrumentation/AddressSanitizer/fake-stack.ll +++ b/llvm/test/Instrumentation/AddressSanitizer/fake-stack.ll @@ -20,24 +20,20 @@ define void @Simple() uwtable sanitize_address { ; NEVER-LABEL: @Simple( ; NEVER-NEXT: entry: ; NEVER-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 64, align 32 -; NEVER-NEXT: [[TMP0:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 -; NEVER-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], 32 -; NEVER-NEXT: [[TMP2:%.*]] = inttoptr i64 [[TMP1]] to ptr -; NEVER-NEXT: [[TMP3:%.*]] = inttoptr i64 [[TMP0]] to ptr -; NEVER-NEXT: store i64 1102416563, ptr [[TMP3]], align 8 -; NEVER-NEXT: [[TMP4:%.*]] = add i64 [[TMP0]], 8 -; NEVER-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr +; NEVER-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 32 +; NEVER-NEXT: store i64 1102416563, ptr [[MYALLOCA]], align 8 +; NEVER-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 8 ; NEVER-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP5]], align 8 -; NEVER-NEXT: [[TMP6:%.*]] = add i64 [[TMP0]], 16 -; NEVER-NEXT: [[TMP7:%.*]] = inttoptr i64 [[TMP6]] to ptr +; NEVER-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 16 ; NEVER-NEXT: store i64 ptrtoint (ptr @Simple to i64), ptr [[TMP7]], align 8 +; NEVER-NEXT: [[TMP0:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; NEVER-NEXT: [[TMP8:%.*]] = lshr i64 [[TMP0]], 3 ; NEVER-NEXT: [[TMP9:%.*]] = add i64 [[TMP8]], 2147450880 ; NEVER-NEXT: [[TMP10:%.*]] = add i64 [[TMP9]], 0 ; NEVER-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr ; NEVER-NEXT: store i64 -868083113472691727, ptr [[TMP11]], align 1 ; NEVER-NEXT: call void @Foo(ptr [[TMP2]]) -; NEVER-NEXT: store i64 1172321806, ptr [[TMP3]], align 8 +; NEVER-NEXT: store i64 1172321806, ptr [[MYALLOCA]], align 8 ; NEVER-NEXT: [[TMP12:%.*]] = add i64 [[TMP9]], 0 ; NEVER-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr ; NEVER-NEXT: store i64 0, ptr [[TMP13]], align 1 @@ -54,25 +50,22 @@ define void @Simple() uwtable sanitize_address { ; RUNTIME-NEXT: br label [[TMP4]] ; RUNTIME: 4: ; RUNTIME-NEXT: [[TMP5:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[TMP3]], [[TMP2]] ] +; RUNTIME-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP5]] to ptr ; RUNTIME-NEXT: [[TMP6:%.*]] = icmp eq i64 [[TMP5]], 0 ; RUNTIME-NEXT: br i1 [[TMP6]], label [[TMP7:%.*]], label [[TMP9:%.*]] -; RUNTIME: 7: +; RUNTIME: 8: ; RUNTIME-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 64, align 32 -; RUNTIME-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; RUNTIME-NEXT: br label [[TMP9]] ; RUNTIME: 9: -; RUNTIME-NEXT: [[TMP10:%.*]] = phi i64 [ [[TMP5]], [[TMP4]] ], [ [[TMP8]], [[TMP7]] ] -; RUNTIME-NEXT: store i64 [[TMP10]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 -; RUNTIME-NEXT: [[TMP11:%.*]] = add i64 [[TMP10]], 32 -; RUNTIME-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr -; RUNTIME-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP10]] to ptr +; RUNTIME-NEXT: [[TMP13:%.*]] = phi ptr [ [[TMP8]], [[TMP4]] ], [ [[MYALLOCA]], [[TMP7]] ] +; RUNTIME-NEXT: store ptr [[TMP13]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 +; RUNTIME-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP13]], i64 32 ; RUNTIME-NEXT: store i64 1102416563, ptr [[TMP13]], align 8 -; RUNTIME-NEXT: [[TMP14:%.*]] = add i64 [[TMP10]], 8 -; RUNTIME-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr +; RUNTIME-NEXT: [[TMP15:%.*]] = getelementptr i8, ptr [[TMP13]], i64 8 ; RUNTIME-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP15]], align 8 -; RUNTIME-NEXT: [[TMP16:%.*]] = add i64 [[TMP10]], 16 -; RUNTIME-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr +; RUNTIME-NEXT: [[TMP17:%.*]] = getelementptr i8, ptr [[TMP13]], i64 16 ; RUNTIME-NEXT: store i64 ptrtoint (ptr @Simple to i64), ptr [[TMP17]], align 8 +; RUNTIME-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[TMP13]] to i64 ; RUNTIME-NEXT: [[TMP18:%.*]] = lshr i64 [[TMP10]], 3 ; RUNTIME-NEXT: [[TMP19:%.*]] = add i64 [[TMP18]], 2147450880 ; RUNTIME-NEXT: [[TMP20:%.*]] = add i64 [[TMP19]], 0 @@ -82,47 +75,43 @@ define void @Simple() uwtable sanitize_address { ; RUNTIME-NEXT: store i64 1172321806, ptr [[TMP13]], align 8 ; RUNTIME-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP5]], 0 ; RUNTIME-NEXT: br i1 [[TMP22]], label [[TMP23:%.*]], label [[TMP30:%.*]] -; RUNTIME: 23: +; RUNTIME: 20: ; RUNTIME-NEXT: [[TMP24:%.*]] = add i64 [[TMP19]], 0 ; RUNTIME-NEXT: [[TMP25:%.*]] = inttoptr i64 [[TMP24]] to ptr ; RUNTIME-NEXT: store i64 -723401728380766731, ptr [[TMP25]], align 1 -; RUNTIME-NEXT: [[TMP26:%.*]] = add i64 [[TMP5]], 56 -; RUNTIME-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP26]] to ptr +; RUNTIME-NEXT: [[TMP27:%.*]] = getelementptr i8, ptr [[TMP8]], i64 56 ; RUNTIME-NEXT: [[TMP28:%.*]] = load i64, ptr [[TMP27]], align 8 ; RUNTIME-NEXT: [[TMP29:%.*]] = inttoptr i64 [[TMP28]] to ptr ; RUNTIME-NEXT: store i8 0, ptr [[TMP29]], align 1 ; RUNTIME-NEXT: br label [[TMP33:%.*]] -; RUNTIME: 30: +; RUNTIME: 26: ; RUNTIME-NEXT: [[TMP31:%.*]] = add i64 [[TMP19]], 0 ; RUNTIME-NEXT: [[TMP32:%.*]] = inttoptr i64 [[TMP31]] to ptr ; RUNTIME-NEXT: store i64 0, ptr [[TMP32]], align 1 ; RUNTIME-NEXT: br label [[TMP33]] -; RUNTIME: 33: +; RUNTIME: 29: ; RUNTIME-NEXT: ret void ; ; ALWAYS-LABEL: @Simple( ; ALWAYS-NEXT: entry: ; ALWAYS-NEXT: [[ASAN_LOCAL_STACK_BASE:%.*]] = alloca i64, align 8 ; ALWAYS-NEXT: [[TMP0:%.*]] = call i64 @__asan_stack_malloc_always_0(i64 64) +; ALWAYS-NEXT: [[TMP3:%.*]] = inttoptr i64 [[TMP0]] to ptr ; ALWAYS-NEXT: [[TMP1:%.*]] = icmp eq i64 [[TMP0]], 0 ; ALWAYS-NEXT: br i1 [[TMP1]], label [[TMP2:%.*]], label [[TMP4:%.*]] -; ALWAYS: 2: +; ALWAYS: 3: ; ALWAYS-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 64, align 32 -; ALWAYS-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; ALWAYS-NEXT: br label [[TMP4]] ; ALWAYS: 4: -; ALWAYS-NEXT: [[TMP5:%.*]] = phi i64 [ [[TMP0]], [[ENTRY:%.*]] ], [ [[TMP3]], [[TMP2]] ] -; ALWAYS-NEXT: store i64 [[TMP5]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 -; ALWAYS-NEXT: [[TMP6:%.*]] = add i64 [[TMP5]], 32 -; ALWAYS-NEXT: [[TMP7:%.*]] = inttoptr i64 [[TMP6]] to ptr -; ALWAYS-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP5]] to ptr +; ALWAYS-NEXT: [[TMP8:%.*]] = phi ptr [ [[TMP3]], [[ENTRY:%.*]] ], [ [[MYALLOCA]], [[TMP2]] ] +; ALWAYS-NEXT: store ptr [[TMP8]], ptr [[ASAN_LOCAL_STACK_BASE]], align 8 +; ALWAYS-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[TMP8]], i64 32 ; ALWAYS-NEXT: store i64 1102416563, ptr [[TMP8]], align 8 -; ALWAYS-NEXT: [[TMP9:%.*]] = add i64 [[TMP5]], 8 -; ALWAYS-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr +; ALWAYS-NEXT: [[TMP10:%.*]] = getelementptr i8, ptr [[TMP8]], i64 8 ; ALWAYS-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP10]], align 8 -; ALWAYS-NEXT: [[TMP11:%.*]] = add i64 [[TMP5]], 16 -; ALWAYS-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr +; ALWAYS-NEXT: [[TMP12:%.*]] = getelementptr i8, ptr [[TMP8]], i64 16 ; ALWAYS-NEXT: store i64 ptrtoint (ptr @Simple to i64), ptr [[TMP12]], align 8 +; ALWAYS-NEXT: [[TMP5:%.*]] = ptrtoint ptr [[TMP8]] to i64 ; ALWAYS-NEXT: [[TMP13:%.*]] = lshr i64 [[TMP5]], 3 ; ALWAYS-NEXT: [[TMP14:%.*]] = add i64 [[TMP13]], 2147450880 ; ALWAYS-NEXT: [[TMP15:%.*]] = add i64 [[TMP14]], 0 @@ -132,22 +121,21 @@ define void @Simple() uwtable sanitize_address { ; ALWAYS-NEXT: store i64 1172321806, ptr [[TMP8]], align 8 ; ALWAYS-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP0]], 0 ; ALWAYS-NEXT: br i1 [[TMP17]], label [[TMP18:%.*]], label [[TMP25:%.*]] -; ALWAYS: 18: +; ALWAYS: 15: ; ALWAYS-NEXT: [[TMP19:%.*]] = add i64 [[TMP14]], 0 ; ALWAYS-NEXT: [[TMP20:%.*]] = inttoptr i64 [[TMP19]] to ptr ; ALWAYS-NEXT: store i64 -723401728380766731, ptr [[TMP20]], align 1 -; ALWAYS-NEXT: [[TMP21:%.*]] = add i64 [[TMP0]], 56 -; ALWAYS-NEXT: [[TMP22:%.*]] = inttoptr i64 [[TMP21]] to ptr +; ALWAYS-NEXT: [[TMP22:%.*]] = getelementptr i8, ptr [[TMP3]], i64 56 ; ALWAYS-NEXT: [[TMP23:%.*]] = load i64, ptr [[TMP22]], align 8 ; ALWAYS-NEXT: [[TMP24:%.*]] = inttoptr i64 [[TMP23]] to ptr ; ALWAYS-NEXT: store i8 0, ptr [[TMP24]], align 1 ; ALWAYS-NEXT: br label [[TMP28:%.*]] -; ALWAYS: 25: +; ALWAYS: 21: ; ALWAYS-NEXT: [[TMP26:%.*]] = add i64 [[TMP14]], 0 ; ALWAYS-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP26]] to ptr ; ALWAYS-NEXT: store i64 0, ptr [[TMP27]], align 1 ; ALWAYS-NEXT: br label [[TMP28]] -; ALWAYS: 28: +; ALWAYS: 24: ; ALWAYS-NEXT: ret void ; entry: @@ -160,17 +148,13 @@ define void @Huge() uwtable sanitize_address { ; CHECK-LABEL: @Huge( ; CHECK-NEXT: entry: ; CHECK-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 100288, align 32 -; CHECK-NEXT: [[TMP0:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 -; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], 32 -; CHECK-NEXT: [[TMP2:%.*]] = inttoptr i64 [[TMP1]] to ptr -; CHECK-NEXT: [[TMP3:%.*]] = inttoptr i64 [[TMP0]] to ptr -; CHECK-NEXT: store i64 1102416563, ptr [[TMP3]], align 8 -; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP0]], 8 -; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 32 +; CHECK-NEXT: store i64 1102416563, ptr [[MYALLOCA]], align 8 +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 8 ; CHECK-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack.1 to i64), ptr [[TMP5]], align 8 -; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP0]], 16 -; CHECK-NEXT: [[TMP7:%.*]] = inttoptr i64 [[TMP6]] to ptr +; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 16 ; CHECK-NEXT: store i64 ptrtoint (ptr @Huge to i64), ptr [[TMP7]], align 8 +; CHECK-NEXT: [[TMP0:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; CHECK-NEXT: [[TMP8:%.*]] = lshr i64 [[TMP0]], 3 ; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[TMP8]], 2147450880 ; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP9]], 0 @@ -189,7 +173,7 @@ define void @Huge() uwtable sanitize_address { ; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr ; CHECK-NEXT: store i64 -868082074056920077, ptr [[TMP19]], align 1 ; CHECK-NEXT: call void @Foo(ptr [[TMP2]]) -; CHECK-NEXT: store i64 1172321806, ptr [[TMP3]], align 8 +; CHECK-NEXT: store i64 1172321806, ptr [[MYALLOCA]], align 8 ; CHECK-NEXT: [[TMP20:%.*]] = add i64 [[TMP9]], 0 ; CHECK-NEXT: [[TMP21:%.*]] = inttoptr i64 [[TMP20]] to ptr ; CHECK-NEXT: store i32 0, ptr [[TMP21]], align 1 diff --git a/llvm/test/Instrumentation/AddressSanitizer/lifetime.ll b/llvm/test/Instrumentation/AddressSanitizer/lifetime.ll index d1e0180..82e114e 100644 --- a/llvm/test/Instrumentation/AddressSanitizer/lifetime.ll +++ b/llvm/test/Instrumentation/AddressSanitizer/lifetime.ll @@ -16,17 +16,13 @@ define void @lifetime() sanitize_address { ; CHECK-DEFAULT-NEXT: [[TMP1:%.*]] = alloca i64, align 32 ; CHECK-DEFAULT-NEXT: store i64 0, ptr [[TMP1]], align 8 ; CHECK-DEFAULT-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 64, align 32 -; CHECK-DEFAULT-NEXT: [[TMP2:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 -; CHECK-DEFAULT-NEXT: [[TMP3:%.*]] = add i64 [[TMP2]], 32 -; CHECK-DEFAULT-NEXT: [[TMP4:%.*]] = inttoptr i64 [[TMP3]] to ptr -; CHECK-DEFAULT-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP2]] to ptr -; CHECK-DEFAULT-NEXT: store i64 1102416563, ptr [[TMP5]], align 8 -; CHECK-DEFAULT-NEXT: [[TMP6:%.*]] = add i64 [[TMP2]], 8 -; CHECK-DEFAULT-NEXT: [[TMP7:%.*]] = inttoptr i64 [[TMP6]] to ptr +; CHECK-DEFAULT-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 32 +; CHECK-DEFAULT-NEXT: store i64 1102416563, ptr [[MYALLOCA]], align 8 +; CHECK-DEFAULT-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 8 ; CHECK-DEFAULT-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP7]], align 8 -; CHECK-DEFAULT-NEXT: [[TMP8:%.*]] = add i64 [[TMP2]], 16 -; CHECK-DEFAULT-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr +; CHECK-DEFAULT-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 16 ; CHECK-DEFAULT-NEXT: store i64 ptrtoint (ptr @lifetime to i64), ptr [[TMP9]], align 8 +; CHECK-DEFAULT-NEXT: [[TMP2:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; CHECK-DEFAULT-NEXT: [[TMP10:%.*]] = lshr i64 [[TMP2]], 3 ; CHECK-DEFAULT-NEXT: [[TMP11:%.*]] = add i64 [[TMP10]], 2147450880 ; CHECK-DEFAULT-NEXT: [[TMP12:%.*]] = add i64 [[TMP11]], 0 @@ -41,16 +37,16 @@ define void @lifetime() sanitize_address { ; CHECK-DEFAULT-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr ; CHECK-DEFAULT-NEXT: [[TMP20:%.*]] = load i8, ptr [[TMP19]], align 1 ; CHECK-DEFAULT-NEXT: [[TMP21:%.*]] = icmp ne i8 [[TMP20]], 0 -; CHECK-DEFAULT-NEXT: br i1 [[TMP21]], label %[[BB22:.*]], label %[[BB27:.*]], !prof [[PROF1:![0-9]+]] -; CHECK-DEFAULT: [[BB22]]: +; CHECK-DEFAULT-NEXT: br i1 [[TMP21]], label %[[BB18:.*]], label %[[BB23:.*]], !prof [[PROF1:![0-9]+]] +; CHECK-DEFAULT: [[BB18]]: ; CHECK-DEFAULT-NEXT: [[TMP23:%.*]] = and i64 [[TMP16]], 7 ; CHECK-DEFAULT-NEXT: [[TMP24:%.*]] = trunc i64 [[TMP23]] to i8 ; CHECK-DEFAULT-NEXT: [[TMP25:%.*]] = icmp sge i8 [[TMP24]], [[TMP20]] -; CHECK-DEFAULT-NEXT: br i1 [[TMP25]], label %[[BB26:.*]], label %[[BB27]] -; CHECK-DEFAULT: [[BB26]]: +; CHECK-DEFAULT-NEXT: br i1 [[TMP25]], label %[[BB22:.*]], label %[[BB23]] +; CHECK-DEFAULT: [[BB22]]: ; CHECK-DEFAULT-NEXT: call void @__asan_report_store1(i64 [[TMP16]]) #[[ATTR4:[0-9]+]] ; CHECK-DEFAULT-NEXT: unreachable -; CHECK-DEFAULT: [[BB27]]: +; CHECK-DEFAULT: [[BB23]]: ; CHECK-DEFAULT-NEXT: store volatile i8 0, ptr [[TMP4]], align 1 ; CHECK-DEFAULT-NEXT: [[TMP28:%.*]] = add i64 [[TMP11]], 4 ; CHECK-DEFAULT-NEXT: [[TMP29:%.*]] = inttoptr i64 [[TMP28]] to ptr @@ -73,16 +69,16 @@ define void @lifetime() sanitize_address { ; CHECK-DEFAULT-NEXT: [[TMP41:%.*]] = inttoptr i64 [[TMP40]] to ptr ; CHECK-DEFAULT-NEXT: [[TMP42:%.*]] = load i8, ptr [[TMP41]], align 1 ; CHECK-DEFAULT-NEXT: [[TMP43:%.*]] = icmp ne i8 [[TMP42]], 0 -; CHECK-DEFAULT-NEXT: br i1 [[TMP43]], label %[[BB44:.*]], label %[[BB49:.*]], !prof [[PROF1]] -; CHECK-DEFAULT: [[BB44]]: +; CHECK-DEFAULT-NEXT: br i1 [[TMP43]], label %[[BB40:.*]], label %[[BB45:.*]], !prof [[PROF1]] +; CHECK-DEFAULT: [[BB40]]: ; CHECK-DEFAULT-NEXT: [[TMP45:%.*]] = and i64 [[TMP38]], 7 ; CHECK-DEFAULT-NEXT: [[TMP46:%.*]] = trunc i64 [[TMP45]] to i8 ; CHECK-DEFAULT-NEXT: [[TMP47:%.*]] = icmp sge i8 [[TMP46]], [[TMP42]] -; CHECK-DEFAULT-NEXT: br i1 [[TMP47]], label %[[BB48:.*]], label %[[BB49]] -; CHECK-DEFAULT: [[BB48]]: +; CHECK-DEFAULT-NEXT: br i1 [[TMP47]], label %[[BB44:.*]], label %[[BB45]] +; CHECK-DEFAULT: [[BB44]]: ; CHECK-DEFAULT-NEXT: call void @__asan_report_store1(i64 [[TMP38]]) #[[ATTR4]] ; CHECK-DEFAULT-NEXT: unreachable -; CHECK-DEFAULT: [[BB49]]: +; CHECK-DEFAULT: [[BB45]]: ; CHECK-DEFAULT-NEXT: store volatile i8 0, ptr [[TMP36]], align 1 ; CHECK-DEFAULT-NEXT: [[TMP50:%.*]] = ptrtoint ptr [[TMP36]] to i64 ; CHECK-DEFAULT-NEXT: call void @__asan_poison_stack_memory(i64 [[TMP50]], i64 40) @@ -95,16 +91,16 @@ define void @lifetime() sanitize_address { ; CHECK-DEFAULT-NEXT: [[TMP56:%.*]] = inttoptr i64 [[TMP55]] to ptr ; CHECK-DEFAULT-NEXT: [[TMP57:%.*]] = load i8, ptr [[TMP56]], align 1 ; CHECK-DEFAULT-NEXT: [[TMP58:%.*]] = icmp ne i8 [[TMP57]], 0 -; CHECK-DEFAULT-NEXT: br i1 [[TMP58]], label %[[BB59:.*]], label %[[BB64:.*]], !prof [[PROF1]] -; CHECK-DEFAULT: [[BB59]]: +; CHECK-DEFAULT-NEXT: br i1 [[TMP58]], label %[[BB55:.*]], label %[[BB60:.*]], !prof [[PROF1]] +; CHECK-DEFAULT: [[BB55]]: ; CHECK-DEFAULT-NEXT: [[TMP60:%.*]] = and i64 [[TMP53]], 7 ; CHECK-DEFAULT-NEXT: [[TMP61:%.*]] = trunc i64 [[TMP60]] to i8 ; CHECK-DEFAULT-NEXT: [[TMP62:%.*]] = icmp sge i8 [[TMP61]], [[TMP57]] -; CHECK-DEFAULT-NEXT: br i1 [[TMP62]], label %[[BB63:.*]], label %[[BB64]] -; CHECK-DEFAULT: [[BB63]]: +; CHECK-DEFAULT-NEXT: br i1 [[TMP62]], label %[[BB59:.*]], label %[[BB60]] +; CHECK-DEFAULT: [[BB59]]: ; CHECK-DEFAULT-NEXT: call void @__asan_report_store1(i64 [[TMP53]]) #[[ATTR4]] ; CHECK-DEFAULT-NEXT: unreachable -; CHECK-DEFAULT: [[BB64]]: +; CHECK-DEFAULT: [[BB60]]: ; CHECK-DEFAULT-NEXT: store volatile i8 0, ptr [[TMP4]], align 1 ; CHECK-DEFAULT-NEXT: [[TMP65:%.*]] = add i64 [[TMP11]], 4 ; CHECK-DEFAULT-NEXT: [[TMP66:%.*]] = inttoptr i64 [[TMP65]] to ptr @@ -112,7 +108,7 @@ define void @lifetime() sanitize_address { ; CHECK-DEFAULT-NEXT: [[TMP67:%.*]] = ptrtoint ptr [[TMP1]] to i64 ; CHECK-DEFAULT-NEXT: [[TMP68:%.*]] = load i64, ptr [[TMP1]], align 8 ; CHECK-DEFAULT-NEXT: call void @__asan_allocas_unpoison(i64 [[TMP68]], i64 [[TMP67]]) -; CHECK-DEFAULT-NEXT: store i64 1172321806, ptr [[TMP5]], align 8 +; CHECK-DEFAULT-NEXT: store i64 1172321806, ptr [[MYALLOCA]], align 8 ; CHECK-DEFAULT-NEXT: [[TMP69:%.*]] = add i64 [[TMP11]], 0 ; CHECK-DEFAULT-NEXT: [[TMP70:%.*]] = inttoptr i64 [[TMP69]] to ptr ; CHECK-DEFAULT-NEXT: store i64 0, ptr [[TMP70]], align 1 @@ -121,17 +117,13 @@ define void @lifetime() sanitize_address { ; CHECK-NO-DYNAMIC-LABEL: define void @lifetime( ; CHECK-NO-DYNAMIC-SAME: ) #[[ATTR0:[0-9]+]] { ; CHECK-NO-DYNAMIC-NEXT: [[MYALLOCA:%.*]] = alloca i8, i64 64, align 32 -; CHECK-NO-DYNAMIC-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 -; CHECK-NO-DYNAMIC-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], 32 -; CHECK-NO-DYNAMIC-NEXT: [[TMP3:%.*]] = inttoptr i64 [[TMP2]] to ptr -; CHECK-NO-DYNAMIC-NEXT: [[TMP4:%.*]] = inttoptr i64 [[TMP1]] to ptr -; CHECK-NO-DYNAMIC-NEXT: store i64 1102416563, ptr [[TMP4]], align 8 -; CHECK-NO-DYNAMIC-NEXT: [[TMP5:%.*]] = add i64 [[TMP1]], 8 -; CHECK-NO-DYNAMIC-NEXT: [[TMP6:%.*]] = inttoptr i64 [[TMP5]] to ptr +; CHECK-NO-DYNAMIC-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 32 +; CHECK-NO-DYNAMIC-NEXT: store i64 1102416563, ptr [[MYALLOCA]], align 8 +; CHECK-NO-DYNAMIC-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 8 ; CHECK-NO-DYNAMIC-NEXT: store i64 ptrtoint (ptr @___asan_gen_stack to i64), ptr [[TMP6]], align 8 -; CHECK-NO-DYNAMIC-NEXT: [[TMP7:%.*]] = add i64 [[TMP1]], 16 -; CHECK-NO-DYNAMIC-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr +; CHECK-NO-DYNAMIC-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[MYALLOCA]], i64 16 ; CHECK-NO-DYNAMIC-NEXT: store i64 ptrtoint (ptr @lifetime to i64), ptr [[TMP8]], align 8 +; CHECK-NO-DYNAMIC-NEXT: [[TMP1:%.*]] = ptrtoint ptr [[MYALLOCA]] to i64 ; CHECK-NO-DYNAMIC-NEXT: [[TMP9:%.*]] = lshr i64 [[TMP1]], 3 ; CHECK-NO-DYNAMIC-NEXT: [[TMP10:%.*]] = add i64 [[TMP9]], 2147450880 ; CHECK-NO-DYNAMIC-NEXT: [[TMP11:%.*]] = add i64 [[TMP10]], 0 @@ -146,16 +138,16 @@ define void @lifetime() sanitize_address { ; CHECK-NO-DYNAMIC-NEXT: [[TMP18:%.*]] = inttoptr i64 [[TMP17]] to ptr ; CHECK-NO-DYNAMIC-NEXT: [[TMP19:%.*]] = load i8, ptr [[TMP18]], align 1 ; CHECK-NO-DYNAMIC-NEXT: [[TMP20:%.*]] = icmp ne i8 [[TMP19]], 0 -; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP20]], label %[[BB21:.*]], label %[[BB26:.*]], !prof [[PROF1:![0-9]+]] -; CHECK-NO-DYNAMIC: [[BB21]]: +; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP20]], label %[[BB17:.*]], label %[[BB22:.*]], !prof [[PROF1:![0-9]+]] +; CHECK-NO-DYNAMIC: [[BB17]]: ; CHECK-NO-DYNAMIC-NEXT: [[TMP22:%.*]] = and i64 [[TMP15]], 7 ; CHECK-NO-DYNAMIC-NEXT: [[TMP23:%.*]] = trunc i64 [[TMP22]] to i8 ; CHECK-NO-DYNAMIC-NEXT: [[TMP24:%.*]] = icmp sge i8 [[TMP23]], [[TMP19]] -; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP24]], label %[[BB25:.*]], label %[[BB26]] -; CHECK-NO-DYNAMIC: [[BB25]]: +; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP24]], label %[[BB21:.*]], label %[[BB22]] +; CHECK-NO-DYNAMIC: [[BB21]]: ; CHECK-NO-DYNAMIC-NEXT: call void @__asan_report_store1(i64 [[TMP15]]) #[[ATTR4:[0-9]+]] ; CHECK-NO-DYNAMIC-NEXT: unreachable -; CHECK-NO-DYNAMIC: [[BB26]]: +; CHECK-NO-DYNAMIC: [[BB22]]: ; CHECK-NO-DYNAMIC-NEXT: store volatile i8 0, ptr [[TMP3]], align 1 ; CHECK-NO-DYNAMIC-NEXT: [[TMP27:%.*]] = add i64 [[TMP10]], 4 ; CHECK-NO-DYNAMIC-NEXT: [[TMP28:%.*]] = inttoptr i64 [[TMP27]] to ptr @@ -171,16 +163,16 @@ define void @lifetime() sanitize_address { ; CHECK-NO-DYNAMIC-NEXT: [[TMP34:%.*]] = inttoptr i64 [[TMP33]] to ptr ; CHECK-NO-DYNAMIC-NEXT: [[TMP35:%.*]] = load i8, ptr [[TMP34]], align 1 ; CHECK-NO-DYNAMIC-NEXT: [[TMP36:%.*]] = icmp ne i8 [[TMP35]], 0 -; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP36]], label %[[BB37:.*]], label %[[BB42:.*]], !prof [[PROF1]] -; CHECK-NO-DYNAMIC: [[BB37]]: +; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP36]], label %[[BB33:.*]], label %[[BB38:.*]], !prof [[PROF1]] +; CHECK-NO-DYNAMIC: [[BB33]]: ; CHECK-NO-DYNAMIC-NEXT: [[TMP38:%.*]] = and i64 [[TMP31]], 7 ; CHECK-NO-DYNAMIC-NEXT: [[TMP39:%.*]] = trunc i64 [[TMP38]] to i8 ; CHECK-NO-DYNAMIC-NEXT: [[TMP40:%.*]] = icmp sge i8 [[TMP39]], [[TMP35]] -; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP40]], label %[[BB41:.*]], label %[[BB42]] -; CHECK-NO-DYNAMIC: [[BB41]]: +; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP40]], label %[[BB37:.*]], label %[[BB38]] +; CHECK-NO-DYNAMIC: [[BB37]]: ; CHECK-NO-DYNAMIC-NEXT: call void @__asan_report_store1(i64 [[TMP31]]) #[[ATTR4]] ; CHECK-NO-DYNAMIC-NEXT: unreachable -; CHECK-NO-DYNAMIC: [[BB42]]: +; CHECK-NO-DYNAMIC: [[BB38]]: ; CHECK-NO-DYNAMIC-NEXT: store volatile i8 0, ptr [[ARR]], align 1 ; CHECK-NO-DYNAMIC-NEXT: call void @llvm.lifetime.end.p0(ptr [[ARR]]) ; CHECK-NO-DYNAMIC-NEXT: [[TMP43:%.*]] = add i64 [[TMP10]], 4 @@ -192,21 +184,21 @@ define void @lifetime() sanitize_address { ; CHECK-NO-DYNAMIC-NEXT: [[TMP48:%.*]] = inttoptr i64 [[TMP47]] to ptr ; CHECK-NO-DYNAMIC-NEXT: [[TMP49:%.*]] = load i8, ptr [[TMP48]], align 1 ; CHECK-NO-DYNAMIC-NEXT: [[TMP50:%.*]] = icmp ne i8 [[TMP49]], 0 -; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP50]], label %[[BB51:.*]], label %[[BB56:.*]], !prof [[PROF1]] -; CHECK-NO-DYNAMIC: [[BB51]]: +; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP50]], label %[[BB47:.*]], label %[[BB52:.*]], !prof [[PROF1]] +; CHECK-NO-DYNAMIC: [[BB47]]: ; CHECK-NO-DYNAMIC-NEXT: [[TMP52:%.*]] = and i64 [[TMP45]], 7 ; CHECK-NO-DYNAMIC-NEXT: [[TMP53:%.*]] = trunc i64 [[TMP52]] to i8 ; CHECK-NO-DYNAMIC-NEXT: [[TMP54:%.*]] = icmp sge i8 [[TMP53]], [[TMP49]] -; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP54]], label %[[BB55:.*]], label %[[BB56]] -; CHECK-NO-DYNAMIC: [[BB55]]: +; CHECK-NO-DYNAMIC-NEXT: br i1 [[TMP54]], label %[[BB51:.*]], label %[[BB52]] +; CHECK-NO-DYNAMIC: [[BB51]]: ; CHECK-NO-DYNAMIC-NEXT: call void @__asan_report_store1(i64 [[TMP45]]) #[[ATTR4]] ; CHECK-NO-DYNAMIC-NEXT: unreachable -; CHECK-NO-DYNAMIC: [[BB56]]: +; CHECK-NO-DYNAMIC: [[BB52]]: ; CHECK-NO-DYNAMIC-NEXT: store volatile i8 0, ptr [[TMP3]], align 1 ; CHECK-NO-DYNAMIC-NEXT: [[TMP57:%.*]] = add i64 [[TMP10]], 4 ; CHECK-NO-DYNAMIC-NEXT: [[TMP58:%.*]] = inttoptr i64 [[TMP57]] to ptr ; CHECK-NO-DYNAMIC-NEXT: store i8 -8, ptr [[TMP58]], align 1 -; CHECK-NO-DYNAMIC-NEXT: store i64 1172321806, ptr [[TMP4]], align 8 +; CHECK-NO-DYNAMIC-NEXT: store i64 1172321806, ptr [[MYALLOCA]], align 8 ; CHECK-NO-DYNAMIC-NEXT: [[TMP59:%.*]] = add i64 [[TMP10]], 0 ; CHECK-NO-DYNAMIC-NEXT: [[TMP60:%.*]] = inttoptr i64 [[TMP59]] to ptr ; CHECK-NO-DYNAMIC-NEXT: store i64 0, ptr [[TMP60]], align 1 diff --git a/llvm/test/Instrumentation/AddressSanitizer/local_stack_base.ll b/llvm/test/Instrumentation/AddressSanitizer/local_stack_base.ll index afa46e44..8b0940d 100644 --- a/llvm/test/Instrumentation/AddressSanitizer/local_stack_base.ll +++ b/llvm/test/Instrumentation/AddressSanitizer/local_stack_base.ll @@ -16,9 +16,8 @@ entry: call void @llvm.dbg.declare(metadata ptr %i.addr, metadata !12, metadata !DIExpression()), !dbg !13 ; CHECK: %asan_local_stack_base = alloca i64 - ; CHECK: %[[ALLOCA:.*]] = ptrtoint ptr %MyAlloca to i64 - ; CHECK: %[[PHI:.*]] = phi i64 {{.*}} %[[ALLOCA]], - ; CHECK: store i64 %[[PHI]], ptr %asan_local_stack_base + ; CHECK: %[[PHI:.*]] = phi ptr {{.*}} %MyAlloca + ; CHECK: store ptr %[[PHI]], ptr %asan_local_stack_base ; CHECK: #dbg_declare(ptr %asan_local_stack_base, [[VAR_I:![0-9]+]], !DIExpression(DW_OP_deref, DW_OP_plus_uconst, 32), [[LOC_I:![0-9]+]] %0 = load i32, ptr %i.addr, align 4, !dbg !14 %add = add nsw i32 %0, 2, !dbg !15 diff --git a/llvm/test/Instrumentation/AddressSanitizer/stack-poisoning-byval-args.ll b/llvm/test/Instrumentation/AddressSanitizer/stack-poisoning-byval-args.ll index d85f217..7e14987 100644 --- a/llvm/test/Instrumentation/AddressSanitizer/stack-poisoning-byval-args.ll +++ b/llvm/test/Instrumentation/AddressSanitizer/stack-poisoning-byval-args.ll @@ -19,8 +19,7 @@ entry: ; CHECK-LABEL: foo ; CHECK: call i64 @__asan_stack_malloc ; CHECK: alloca i8, i64 {{.*}} align 64 -; CHECK: [[copyPtr:%[^ \t]+]] = inttoptr i64 %{{[^ \t]+}} to ptr -; CHECK: call void @llvm.memcpy{{[^%]+}}[[copyPtr]]{{[^%]+}} align 64 %a,{{[^,]+}}, +; CHECK: call void @llvm.memcpy{{[^%]+}}[[copyPtr:%[0-9]+]]{{[^%]+}} align 64 %a,{{[^,]+}}, ; CHECK: call i32 @bar(ptr [[copyPtr]]) ; CHECK: ret void @@ -38,8 +37,7 @@ entry: ; CHECK-LABEL: baz ; CHECK: call i64 @__asan_stack_malloc ; CHECK: alloca i8, i64 {{.*}} align 32 -; CHECK: [[copyPtr:%[^ \t]+]] = inttoptr i64 %{{[^ \t]+}} to ptr -; CHECK: call void @llvm.memcpy{{[^%]+}}[[copyPtr]]{{[^%]+}} align 4 %0,{{[^,]+}} +; CHECK: call void @llvm.memcpy{{[^%]+}}[[copyPtr:%[0-9]+]]{{[^%]+}} align 4 %0,{{[^,]+}} ; CHECK: call i32 @bar(ptr [[copyPtr]]) ; CHECK: ret void diff --git a/llvm/test/Instrumentation/AddressSanitizer/stack_dynamic_alloca.ll b/llvm/test/Instrumentation/AddressSanitizer/stack_dynamic_alloca.ll index d56cd34..c8478c8 100644 --- a/llvm/test/Instrumentation/AddressSanitizer/stack_dynamic_alloca.ll +++ b/llvm/test/Instrumentation/AddressSanitizer/stack_dynamic_alloca.ll @@ -19,15 +19,16 @@ entry: ; CHECK-RUNTIME: [[FAKE_STACK_BB:^[0-9]+]]: ; CHECK-RUNTIME: [[FAKE_STACK:%[0-9]+]] = phi i64 [ 0, %entry ], [ [[FAKE_STACK_RT]], %[[UAR_ENABLED_BB]] ] +; CHECK-RUNTIME: [[FAKE_STACK_PTR:%[0-9]+]] = inttoptr i64 [[FAKE_STACK]] to ptr +; CHECK-ALWAYS: [[FAKE_STACK_PTR:%[0-9]+]] = inttoptr i64 [[FAKE_STACK_RT]] to ptr ; CHECK-RUNTIME: icmp eq i64 [[FAKE_STACK]], 0 ; CHECK-ALWAYS: icmp eq i64 [[FAKE_STACK_RT]], 0 ; CHECK: [[NO_FAKE_STACK_BB:^[0-9]+]]: ; CHECK: %MyAlloca = alloca i8, i64 -; CHECK: [[ALLOCA:%[0-9]+]] = ptrtoint ptr %MyAlloca -; CHECK-RUNTIME: phi i64 [ [[FAKE_STACK]], %[[FAKE_STACK_BB]] ], [ [[ALLOCA]], %[[NO_FAKE_STACK_BB]] ] -; CHECK-ALWAYS: phi i64 [ [[FAKE_STACK_RT]], %entry ], [ [[ALLOCA]], %[[NO_FAKE_STACK_BB]] ] +; CHECK-RUNTIME: phi ptr [ [[FAKE_STACK_PTR]], %[[FAKE_STACK_BB]] ], [ %MyAlloca, %[[NO_FAKE_STACK_BB]] ] +; CHECK-ALWAYS: phi ptr [ [[FAKE_STACK_PTR]], %entry ], [ %MyAlloca, %[[NO_FAKE_STACK_BB]] ] ; CHECK: ret void diff --git a/llvm/unittests/ExecutionEngine/Orc/ExecutorAddressTest.cpp b/llvm/unittests/ExecutionEngine/Orc/ExecutorAddressTest.cpp index ae9db14..9a37980 100644 --- a/llvm/unittests/ExecutionEngine/Orc/ExecutorAddressTest.cpp +++ b/llvm/unittests/ExecutionEngine/Orc/ExecutorAddressTest.cpp @@ -97,10 +97,16 @@ TEST(ExecutorAddrTest, AddrRanges) { EXPECT_FALSE(R1.contains(A0)); EXPECT_FALSE(R1.contains(A2)); + EXPECT_TRUE(R3.contains(R0)); // True for singleton range at start. + EXPECT_TRUE(R3.contains(R1)); // True for singleton range at end. + EXPECT_FALSE(R3.contains(R2)); // False for non-overlaping singleton range. + EXPECT_FALSE(R3.contains(R4)); // False for overlapping, uncontained range. + EXPECT_FALSE(R1.overlaps(R0)); EXPECT_FALSE(R1.overlaps(R2)); EXPECT_TRUE(R1.overlaps(R3)); EXPECT_TRUE(R1.overlaps(R4)); + EXPECT_TRUE(R3.overlaps(R4)); EXPECT_LE(R0, R0); EXPECT_LT(R0, R1); diff --git a/llvm/unittests/IR/ConstantsTest.cpp b/llvm/unittests/IR/ConstantsTest.cpp index 54c7ddd..6376165 100644 --- a/llvm/unittests/IR/ConstantsTest.cpp +++ b/llvm/unittests/IR/ConstantsTest.cpp @@ -564,13 +564,17 @@ TEST(ConstantsTest, FoldGlobalVariablePtr) { Global->setAlignment(Align(4)); - ConstantInt *TheConstant(ConstantInt::get(IntType, 2)); + ConstantInt *TheConstant = ConstantInt::get(IntType, 2); - Constant *TheConstantExpr(ConstantExpr::getPtrToInt(Global.get(), IntType)); + Constant *PtrToInt = ConstantExpr::getPtrToInt(Global.get(), IntType); + ASSERT_TRUE( + ConstantFoldBinaryInstruction(Instruction::And, PtrToInt, TheConstant) + ->isNullValue()); - ASSERT_TRUE(ConstantFoldBinaryInstruction(Instruction::And, TheConstantExpr, - TheConstant) - ->isNullValue()); + Constant *PtrToAddr = ConstantExpr::getPtrToAddr(Global.get(), IntType); + ASSERT_TRUE( + ConstantFoldBinaryInstruction(Instruction::And, PtrToAddr, TheConstant) + ->isNullValue()); } // Check that containsUndefOrPoisonElement and containsPoisonElement is working diff --git a/llvm/unittests/Target/AArch64/AArch64SelectionDAGTest.cpp b/llvm/unittests/Target/AArch64/AArch64SelectionDAGTest.cpp index 5ac4c53..809960d 100644 --- a/llvm/unittests/Target/AArch64/AArch64SelectionDAGTest.cpp +++ b/llvm/unittests/Target/AArch64/AArch64SelectionDAGTest.cpp @@ -228,6 +228,114 @@ TEST_F(AArch64SelectionDAGTest, ComputeNumSignBits_SUB) { EXPECT_EQ(DAG->ComputeNumSignBits(Op), 2u); } +TEST_F(AArch64SelectionDAGTest, ComputeNumSignBits_ADD) { + SDLoc Loc; + auto IntVT = EVT::getIntegerVT(Context, 8); + auto Nneg1 = DAG->getConstant(0xFF, Loc, IntVT); + auto N0 = DAG->getConstant(0x00, Loc, IntVT); + auto N1 = DAG->getConstant(0x01, Loc, IntVT); + auto N5 = DAG->getConstant(0x05, Loc, IntVT); + auto N8 = DAG->getConstant(0x08, Loc, IntVT); + auto Nsign1 = DAG->getConstant(0x55, Loc, IntVT); + auto UnknownOp = DAG->getRegister(0, IntVT); + auto Mask = DAG->getConstant(0x1e, Loc, IntVT); + auto Nsign3 = DAG->getNode(ISD::AND, Loc, IntVT, Mask, UnknownOp); + // RHS early out + // Nsign1 = 01010101 + // Nsign3 = 000????0 + auto OpRhsEo = DAG->getNode(ISD::ADD, Loc, IntVT, Nsign3, Nsign1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpRhsEo), 1u); + + // ADD 0 -1 + // N0 = 00000000 + // Nneg1 = 11111111 + auto OpNegZero = DAG->getNode(ISD::ADD, Loc, IntVT, N0, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpNegZero), 8u); + + // ADD 1 -1 + // N1 = 00000001 + // Nneg1 = 11111111 + auto OpNegOne = DAG->getNode(ISD::ADD, Loc, IntVT, N1, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpNegOne), 8u); + + // ADD 8 -1 + // N8 = 00001000 + // Nneg1 = 11111111 + auto OpSeven = DAG->getNode(ISD::ADD, Loc, IntVT, N8, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpSeven), 5u); + + // Non negative + // Nsign3 = 000????0 + // Nneg1 = 11111111 + auto OpNonNeg = DAG->getNode(ISD::ADD, Loc, IntVT, Nsign3, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpNonNeg), 3u); + + // LHS early out + // Nsign1 = 01010101 + // Nsign3 = 000????0 + auto OpLhsEo = DAG->getNode(ISD::ADD, Loc, IntVT, Nsign1, Nsign3); + EXPECT_EQ(DAG->ComputeNumSignBits(OpLhsEo), 1u); + + // Nsign3 = 000????0 + // N5 = 00000101 + auto Op = DAG->getNode(ISD::ADD, Loc, IntVT, Nsign3, N5); + EXPECT_EQ(DAG->ComputeNumSignBits(Op), 2u); +} + +TEST_F(AArch64SelectionDAGTest, ComputeNumSignBits_ADDC) { + SDLoc Loc; + auto IntVT = EVT::getIntegerVT(Context, 8); + auto Nneg1 = DAG->getConstant(0xFF, Loc, IntVT); + auto N0 = DAG->getConstant(0x00, Loc, IntVT); + auto N1 = DAG->getConstant(0x01, Loc, IntVT); + auto N5 = DAG->getConstant(0x05, Loc, IntVT); + auto N8 = DAG->getConstant(0x08, Loc, IntVT); + auto Nsign1 = DAG->getConstant(0x55, Loc, IntVT); + auto UnknownOp = DAG->getRegister(0, IntVT); + auto Mask = DAG->getConstant(0x1e, Loc, IntVT); + auto Nsign3 = DAG->getNode(ISD::AND, Loc, IntVT, Mask, UnknownOp); + // RHS early out + // Nsign1 = 01010101 + // Nsign3 = 000????0 + auto OpRhsEo = DAG->getNode(ISD::ADDC, Loc, IntVT, Nsign3, Nsign1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpRhsEo), 1u); + + // ADD 0 -1 + // N0 = 00000000 + // Nneg1 = 11111111 + auto OpNegZero = DAG->getNode(ISD::ADDC, Loc, IntVT, N0, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpNegZero), 8u); + + // ADD 1 -1 + // N1 = 00000001 + // Nneg1 = 11111111 + auto OpNegOne = DAG->getNode(ISD::ADDC, Loc, IntVT, N1, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpNegOne), 8u); + + // ADD 8 -1 + // N8 = 00001000 + // Nneg1 = 11111111 + auto OpSeven = DAG->getNode(ISD::ADDC, Loc, IntVT, N8, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpSeven), 4u); + + // Non negative + // Nsign3 = 000????0 + // Nneg1 = 11111111 + auto OpNonNeg = DAG->getNode(ISD::ADDC, Loc, IntVT, Nsign3, Nneg1); + EXPECT_EQ(DAG->ComputeNumSignBits(OpNonNeg), 3u); + + // LHS early out + // Nsign1 = 01010101 + // Nsign3 = 000????0 + auto OpLhsEo = DAG->getNode(ISD::ADDC, Loc, IntVT, Nsign1, Nsign3); + EXPECT_EQ(DAG->ComputeNumSignBits(OpLhsEo), 1u); + + // Nsign3 = 000????0 + // N5 = 00000101 + auto Op = DAG->getNode(ISD::ADDC, Loc, IntVT, Nsign3, N5); + EXPECT_EQ(DAG->ComputeNumSignBits(Op), 2u); +} + TEST_F(AArch64SelectionDAGTest, SimplifyDemandedVectorElts_EXTRACT_SUBVECTOR) { TargetLowering TL(*TM); |