Index: llvm/trunk/lib/Target/RISCV/RISCVISelLowering.h =================================================================== --- llvm/trunk/lib/Target/RISCV/RISCVISelLowering.h +++ llvm/trunk/lib/Target/RISCV/RISCVISelLowering.h @@ -41,7 +41,14 @@ // at instruction selection time. DIVW, DIVUW, - REMUW + REMUW, + // FPR32<->GPR transfer operations for RV64. Needed as an i32<->f32 bitcast + // is not legal on RV64. FMV_W_X_RV64 matches the semantics of the FMV.W.X. + // FMV_X_ANYEXTW_RV64 is similar to FMV.X.W but has an any-extended result. + // This is a more convenient semantic for producing dagcombines that remove + // unnecessary GPR->FPR->GPR moves. + FMV_W_X_RV64, + FMV_X_ANYEXTW_RV64 }; } Index: llvm/trunk/lib/Target/RISCV/RISCVISelLowering.cpp =================================================================== --- llvm/trunk/lib/Target/RISCV/RISCVISelLowering.cpp +++ llvm/trunk/lib/Target/RISCV/RISCVISelLowering.cpp @@ -137,6 +137,9 @@ setOperationAction(Op, MVT::f32, Expand); } + if (Subtarget.hasStdExtF() && Subtarget.is64Bit()) + setOperationAction(ISD::BITCAST, MVT::i32, Custom); + if (Subtarget.hasStdExtD()) { setOperationAction(ISD::FMINNUM, MVT::f64, Legal); setOperationAction(ISD::FMAXNUM, MVT::f64, Legal); @@ -338,6 +341,17 @@ return lowerFRAMEADDR(Op, DAG); case ISD::RETURNADDR: return lowerRETURNADDR(Op, DAG); + case ISD::BITCAST: { + assert(Subtarget.is64Bit() && Subtarget.hasStdExtF() && + "Unexpected custom legalisation"); + SDLoc DL(Op); + SDValue Op0 = Op.getOperand(0); + if (Op.getValueType() != MVT::f32 || Op0.getValueType() != MVT::i32) + return SDValue(); + SDValue NewOp0 = DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op0); + SDValue FPConv = DAG.getNode(RISCVISD::FMV_W_X_RV64, DL, MVT::f32, NewOp0); + return FPConv; + } } } @@ -579,6 +593,18 @@ return; Results.push_back(customLegalizeToWOp(N, DAG)); break; + case ISD::BITCAST: { + assert(N->getValueType(0) == MVT::i32 && Subtarget.is64Bit() && + Subtarget.hasStdExtF() && "Unexpected custom legalisation"); + SDLoc DL(N); + SDValue Op0 = N->getOperand(0); + if (Op0.getValueType() != MVT::f32) + return; + SDValue FPConv = + DAG.getNode(RISCVISD::FMV_X_ANYEXTW_RV64, DL, MVT::i64, Op0); + Results.push_back(DAG.getNode(ISD::TRUNCATE, DL, MVT::i32, FPConv)); + break; + } } } @@ -633,6 +659,38 @@ return SDValue(); break; } + case RISCVISD::FMV_X_ANYEXTW_RV64: { + SDLoc DL(N); + SDValue Op0 = N->getOperand(0); + // If the input to FMV_X_ANYEXTW_RV64 is just FMV_W_X_RV64 then the + // conversion is unnecessary and can be replaced with an ANY_EXTEND + // of the FMV_W_X_RV64 operand. + if (Op0->getOpcode() == RISCVISD::FMV_W_X_RV64) { + SDValue AExtOp = + DAG.getNode(ISD::ANY_EXTEND, DL, MVT::i64, Op0.getOperand(0)); + return DCI.CombineTo(N, AExtOp); + } + + // This is a target-specific version of a DAGCombine performed in + // DAGCombiner::visitBITCAST. It performs the equivalent of: + // fold (bitconvert (fneg x)) -> (xor (bitconvert x), signbit) + // fold (bitconvert (fabs x)) -> (and (bitconvert x), (not signbit)) + if (!(Op0.getOpcode() == ISD::FNEG || Op0.getOpcode() == ISD::FABS) || + !Op0.getNode()->hasOneUse()) + break; + SDValue NewFMV = DAG.getNode(RISCVISD::FMV_X_ANYEXTW_RV64, DL, MVT::i64, + Op0.getOperand(0)); + APInt SignBit = APInt::getSignMask(32).sext(64); + if (Op0.getOpcode() == ISD::FNEG) { + return DCI.CombineTo(N, + DAG.getNode(ISD::XOR, DL, MVT::i64, NewFMV, + DAG.getConstant(SignBit, DL, MVT::i64))); + } + assert(Op0.getOpcode() == ISD::FABS); + return DCI.CombineTo(N, + DAG.getNode(ISD::AND, DL, MVT::i64, NewFMV, + DAG.getConstant(~SignBit, DL, MVT::i64))); + } } return SDValue(); @@ -874,7 +932,7 @@ assert(XLen == 32 || XLen == 64); MVT XLenVT = XLen == 32 ? MVT::i32 : MVT::i64; if (ValVT == MVT::f32) { - LocVT = MVT::i32; + LocVT = XLenVT; LocInfo = CCValAssign::BCvt; } @@ -1047,6 +1105,10 @@ case CCValAssign::Full: break; case CCValAssign::BCvt: + if (VA.getLocVT() == MVT::i64 && VA.getValVT() == MVT::f32) { + Val = DAG.getNode(RISCVISD::FMV_W_X_RV64, DL, MVT::f32, Val); + break; + } Val = DAG.getNode(ISD::BITCAST, DL, VA.getValVT(), Val); break; } @@ -1082,6 +1144,10 @@ case CCValAssign::Full: break; case CCValAssign::BCvt: + if (VA.getLocVT() == MVT::i64 && VA.getValVT() == MVT::f32) { + Val = DAG.getNode(RISCVISD::FMV_X_ANYEXTW_RV64, DL, MVT::i64, Val); + break; + } Val = DAG.getNode(ISD::BITCAST, DL, LocVT, Val); break; } @@ -1108,9 +1174,12 @@ llvm_unreachable("Unexpected CCValAssign::LocInfo"); case CCValAssign::Full: case CCValAssign::Indirect: + case CCValAssign::BCvt: ExtType = ISD::NON_EXTLOAD; break; } + if (ValVT == MVT::f32) + LocVT = MVT::f32; Val = DAG.getExtLoad( ExtType, DL, LocVT, Chain, FIN, MachinePointerInfo::getFixedStack(DAG.getMachineFunction(), FI), ValVT); @@ -1759,6 +1828,10 @@ return "RISCVISD::DIVUW"; case RISCVISD::REMUW: return "RISCVISD::REMUW"; + case RISCVISD::FMV_W_X_RV64: + return "RISCVISD::FMV_W_X_RV64"; + case RISCVISD::FMV_X_ANYEXTW_RV64: + return "RISCVISD::FMV_X_ANYEXTW_RV64"; } return nullptr; } Index: llvm/trunk/lib/Target/RISCV/RISCVInstrInfoF.td =================================================================== --- llvm/trunk/lib/Target/RISCV/RISCVInstrInfoF.td +++ llvm/trunk/lib/Target/RISCV/RISCVInstrInfoF.td @@ -12,6 +12,20 @@ //===----------------------------------------------------------------------===// //===----------------------------------------------------------------------===// +// RISC-V specific DAG Nodes. +//===----------------------------------------------------------------------===// + +def SDT_RISCVFMV_W_X_RV64 + : SDTypeProfile<1, 1, [SDTCisVT<0, f32>, SDTCisVT<1, i64>]>; +def SDT_RISCVFMV_X_ANYEXTW_RV64 + : SDTypeProfile<1, 1, [SDTCisVT<0, i64>, SDTCisVT<1, f32>]>; + +def riscv_fmv_w_x_rv64 + : SDNode<"RISCVISD::FMV_W_X_RV64", SDT_RISCVFMV_W_X_RV64>; +def riscv_fmv_x_anyextw_rv64 + : SDNode<"RISCVISD::FMV_X_ANYEXTW_RV64", SDT_RISCVFMV_X_ANYEXTW_RV64>; + +//===----------------------------------------------------------------------===// // Operand and SDNode transformation definitions. //===----------------------------------------------------------------------===// @@ -333,3 +347,37 @@ def : Pat<(sint_to_fp GPR:$rs1), (FCVT_S_W $rs1, 0b111)>; def : Pat<(uint_to_fp GPR:$rs1), (FCVT_S_WU $rs1, 0b111)>; } // Predicates = [HasStdExtF, IsRV32] + +let Predicates = [HasStdExtF, IsRV32] in { +// FP->[u]int. Round-to-zero must be used +def : Pat<(fp_to_sint FPR32:$rs1), (FCVT_W_S $rs1, 0b001)>; +def : Pat<(fp_to_uint FPR32:$rs1), (FCVT_WU_S $rs1, 0b001)>; + +// [u]int->fp. Match GCC and default to using dynamic rounding mode. +def : Pat<(sint_to_fp GPR:$rs1), (FCVT_S_W $rs1, 0b111)>; +def : Pat<(uint_to_fp GPR:$rs1), (FCVT_S_WU $rs1, 0b111)>; +} // Predicates = [HasStdExtF, IsRV32] + +let Predicates = [HasStdExtF, IsRV64] in { +def : Pat<(riscv_fmv_w_x_rv64 GPR:$src), (FMV_W_X GPR:$src)>; +def : Pat<(riscv_fmv_x_anyextw_rv64 FPR32:$src), (FMV_X_W FPR32:$src)>; +def : Pat<(sexti32 (riscv_fmv_x_anyextw_rv64 FPR32:$src)), + (FMV_X_W FPR32:$src)>; + +// FP->[u]int32 is mostly handled by the FP->[u]int64 patterns. This is safe +// because fpto[u|s]i produces poison if the value can't fit into the target. +// We match the single case below because fcvt.wu.s sign-extends its result so +// is cheaper than fcvt.lu.s+sext.w. +def : Pat<(sext_inreg (assertzexti32 (fp_to_uint FPR32:$rs1)), i32), + (FCVT_WU_S $rs1, 0b001)>; + +// FP->[u]int64 +def : Pat<(fp_to_sint FPR32:$rs1), (FCVT_L_S $rs1, 0b001)>; +def : Pat<(fp_to_uint FPR32:$rs1), (FCVT_LU_S $rs1, 0b001)>; + +// [u]int->fp. Match GCC and default to using dynamic rounding mode. +def : Pat<(sint_to_fp (sext_inreg GPR:$rs1, i32)), (FCVT_S_W $rs1, 0b111)>; +def : Pat<(uint_to_fp (zexti32 GPR:$rs1)), (FCVT_S_WU $rs1, 0b111)>; +def : Pat<(sint_to_fp GPR:$rs1), (FCVT_S_L $rs1, 0b111)>; +def : Pat<(uint_to_fp GPR:$rs1), (FCVT_S_LU $rs1, 0b111)>; +} // Predicates = [HasStdExtF, IsRV64] Index: llvm/trunk/test/CodeGen/RISCV/float-arith.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-arith.ll +++ llvm/trunk/test/CodeGen/RISCV/float-arith.ll @@ -1,6 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s ; These tests are each targeted at a particular RISC-V FPU instruction. Most ; other files in this folder exercise LLVM IR instructions that don't directly @@ -14,6 +16,14 @@ ; RV32IF-NEXT: fadd.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fadd_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fadd float %a, %b ret float %1 } @@ -26,6 +36,14 @@ ; RV32IF-NEXT: fsub.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fsub_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fsub.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fsub float %a, %b ret float %1 } @@ -38,6 +56,14 @@ ; RV32IF-NEXT: fmul.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmul_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fmul.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fmul float %a, %b ret float %1 } @@ -50,6 +76,14 @@ ; RV32IF-NEXT: fdiv.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fdiv_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fdiv.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fdiv float %a, %b ret float %1 } @@ -63,6 +97,13 @@ ; RV32IF-NEXT: fsqrt.s ft0, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fsqrt_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fsqrt.s ft0, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = call float @llvm.sqrt.f32(float %a) ret float %1 } @@ -77,6 +118,14 @@ ; RV32IF-NEXT: fsgnj.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fsgnj_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fsgnj.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = call float @llvm.copysign.f32(float %a, float %b) ret float %1 } @@ -91,6 +140,14 @@ ; RV32IF-NEXT: fneg.s ft1, ft0 ; RV32IF-NEXT: feq.s a0, ft0, ft1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fneg_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fadd.s ft0, ft0, ft0 +; RV64IF-NEXT: fneg.s ft1, ft0 +; RV64IF-NEXT: feq.s a0, ft0, ft1 +; RV64IF-NEXT: ret %1 = fadd float %a, %a %2 = fneg float %1 %3 = fcmp oeq float %1, %2 @@ -109,6 +166,15 @@ ; RV32IF-NEXT: fsgnjn.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fsgnjn_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fsgnjn.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fadd float %a, %b %2 = fneg float %1 %3 = call float @llvm.copysign.f32(float %a, float %2) @@ -129,6 +195,16 @@ ; RV32IF-NEXT: fadd.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fabs_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fabs.s ft1, ft0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fadd float %a, %b %2 = call float @llvm.fabs.f32(float %1) %3 = fadd float %2, %1 @@ -145,6 +221,14 @@ ; RV32IF-NEXT: fmin.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmin_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fmin.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = call float @llvm.minnum.f32(float %a, float %b) ret float %1 } @@ -159,6 +243,14 @@ ; RV32IF-NEXT: fmax.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmax_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fmax.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = call float @llvm.maxnum.f32(float %a, float %b) ret float %1 } @@ -170,6 +262,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a0 ; RV32IF-NEXT: feq.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: feq_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp oeq float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -182,6 +281,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a0 ; RV32IF-NEXT: flt.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: flt_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp olt float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -194,6 +300,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a0 ; RV32IF-NEXT: fle.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fle_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp ole float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -210,6 +323,15 @@ ; RV32IF-NEXT: fmadd.s ft0, ft2, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmadd_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a2 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft2, a0 +; RV64IF-NEXT: fmadd.s ft0, ft2, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = call float @llvm.fma.f32(float %a, float %b, float %c) ret float %1 } @@ -227,6 +349,19 @@ ; RV32IF-NEXT: fmsub.s ft0, ft2, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmsub_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a2 +; RV64IF-NEXT: lui a2, %hi(.LCPI15_0) +; RV64IF-NEXT: addi a2, a2, %lo(.LCPI15_0) +; RV64IF-NEXT: flw ft1, 0(a2) +; RV64IF-NEXT: fadd.s ft0, ft0, ft1 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft2, a0 +; RV64IF-NEXT: fmsub.s ft0, ft2, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %c_ = fadd float 0.0, %c ; avoid negation using xor %negc = fsub float -0.0, %c_ %1 = call float @llvm.fma.f32(float %a, float %b, float %negc) @@ -247,6 +382,20 @@ ; RV32IF-NEXT: fnmadd.s ft0, ft1, ft2, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fnmadd_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a2 +; RV64IF-NEXT: lui a2, %hi(.LCPI16_0) +; RV64IF-NEXT: addi a2, a2, %lo(.LCPI16_0) +; RV64IF-NEXT: flw ft1, 0(a2) +; RV64IF-NEXT: fadd.s ft0, ft0, ft1 +; RV64IF-NEXT: fmv.w.x ft2, a0 +; RV64IF-NEXT: fadd.s ft1, ft2, ft1 +; RV64IF-NEXT: fmv.w.x ft2, a1 +; RV64IF-NEXT: fnmadd.s ft0, ft1, ft2, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %a_ = fadd float 0.0, %a %c_ = fadd float 0.0, %c %nega = fsub float -0.0, %a_ @@ -268,6 +417,19 @@ ; RV32IF-NEXT: fnmsub.s ft0, ft0, ft2, ft1 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fnmsub_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: lui a0, %hi(.LCPI17_0) +; RV64IF-NEXT: addi a0, a0, %lo(.LCPI17_0) +; RV64IF-NEXT: flw ft1, 0(a0) +; RV64IF-NEXT: fadd.s ft0, ft0, ft1 +; RV64IF-NEXT: fmv.w.x ft1, a2 +; RV64IF-NEXT: fmv.w.x ft2, a1 +; RV64IF-NEXT: fnmsub.s ft0, ft0, ft2, ft1 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %a_ = fadd float 0.0, %a %nega = fsub float -0.0, %a_ %1 = call float @llvm.fma.f32(float %nega, float %b, float %c) Index: llvm/trunk/test/CodeGen/RISCV/float-bitmanip-dagcombines.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-bitmanip-dagcombines.ll +++ llvm/trunk/test/CodeGen/RISCV/float-bitmanip-dagcombines.ll @@ -5,6 +5,8 @@ ; RUN: | FileCheck -check-prefix=RV32IF %s ; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV64I %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s ; This file tests cases where simple floating point operations can be ; profitably handled though bit manipulation if a soft-float ABI is being used @@ -30,6 +32,12 @@ ; RV64I-NEXT: lui a1, 524288 ; RV64I-NEXT: xor a0, a0, a1 ; RV64I-NEXT: ret +; +; RV64IF-LABEL: fneg: +; RV64IF: # %bb.0: +; RV64IF-NEXT: lui a1, 524288 +; RV64IF-NEXT: xor a0, a0, a1 +; RV64IF-NEXT: ret %1 = fneg float %a ret float %1 } @@ -57,6 +65,13 @@ ; RV64I-NEXT: addiw a1, a1, -1 ; RV64I-NEXT: and a0, a0, a1 ; RV64I-NEXT: ret +; +; RV64IF-LABEL: fabs: +; RV64IF: # %bb.0: +; RV64IF-NEXT: lui a1, 524288 +; RV64IF-NEXT: addiw a1, a1, -1 +; RV64IF-NEXT: and a0, a0, a1 +; RV64IF-NEXT: ret %1 = call float @llvm.fabs.f32(float %a) ret float %1 } @@ -97,6 +112,14 @@ ; RV64I-NEXT: and a0, a0, a2 ; RV64I-NEXT: or a0, a0, a1 ; RV64I-NEXT: ret +; +; RV64IF-LABEL: fcopysign_fneg: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fsgnjn.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fneg float %b %2 = call float @llvm.copysign.f32(float %a, float %1) ret float %2 Index: llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll +++ llvm/trunk/test/CodeGen/RISCV/float-br-fcmp.ll @@ -1,6 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s declare void @abort() declare void @exit(i32) @@ -19,6 +21,19 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB0_2: # %if.else ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_false: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: addi a0, zero, 1 +; RV64IF-NEXT: bnez a0, .LBB0_2 +; RV64IF-NEXT: # %bb.1: # %if.then +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB0_2: # %if.else +; RV64IF-NEXT: call abort %1 = fcmp false float %a, %b br i1 %1, label %if.then, label %if.else if.then: @@ -43,6 +58,21 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB1_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_oeq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB1_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB1_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp oeq float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -71,6 +101,22 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB2_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_oeq_alt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: beqz a0, .LBB2_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB2_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp oeq float %a, %b br i1 %1, label %if.then, label %if.else if.then: @@ -95,6 +141,21 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB3_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ogt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB3_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB3_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ogt float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -119,6 +180,21 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB4_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_oge: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB4_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB4_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp oge float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -143,6 +219,21 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB5_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_olt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB5_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB5_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp olt float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -167,6 +258,21 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB6_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ole: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB6_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB6_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ole float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -199,6 +305,28 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB7_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_one: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: feq.s a0, ft1, ft1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: feq.s a1, ft0, ft1 +; RV64IF-NEXT: not a1, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: bnez a0, .LBB7_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB7_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp one float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -227,6 +355,25 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB8_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ord: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: feq.s a0, ft0, ft0 +; RV64IF-NEXT: and a0, a0, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB8_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB8_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ord float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -256,6 +403,26 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB9_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ueq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: feq.s a2, ft1, ft1 +; RV64IF-NEXT: and a1, a2, a1 +; RV64IF-NEXT: seqz a1, a1 +; RV64IF-NEXT: or a0, a0, a1 +; RV64IF-NEXT: bnez a0, .LBB9_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB9_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ueq float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -281,6 +448,22 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB10_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ugt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB10_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB10_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ugt float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -306,6 +489,22 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB11_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_uge: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB11_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB11_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp uge float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -331,6 +530,22 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB12_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ult: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB12_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB12_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ult float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -356,6 +571,22 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB13_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_ule: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB13_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB13_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp ule float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -381,6 +612,22 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB14_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_une: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB14_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB14_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp une float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -409,6 +656,24 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB15_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_uno: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: feq.s a0, ft0, ft0 +; RV64IF-NEXT: and a0, a0, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: bnez a0, .LBB15_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB15_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp uno float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -431,6 +696,19 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB16_2: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_true: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: addi a0, zero, 1 +; RV64IF-NEXT: bnez a0, .LBB16_2 +; RV64IF-NEXT: # %bb.1: # %if.else +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB16_2: # %if.then +; RV64IF-NEXT: call abort %1 = fcmp true float %a, %b br i1 %1, label %if.then, label %if.else if.else: @@ -471,6 +749,38 @@ ; RV32IF-NEXT: ret ; RV32IF-NEXT: .LBB17_3: # %if.then ; RV32IF-NEXT: call abort +; +; RV64IF-LABEL: br_fcmp_store_load_stack_slot: +; RV64IF: # %bb.0: # %entry +; RV64IF-NEXT: addi sp, sp, -32 +; RV64IF-NEXT: sd ra, 24(sp) +; RV64IF-NEXT: sd s1, 16(sp) +; RV64IF-NEXT: lui a0, %hi(.LCPI17_0) +; RV64IF-NEXT: addi a0, a0, %lo(.LCPI17_0) +; RV64IF-NEXT: flw ft0, 0(a0) +; RV64IF-NEXT: fsw ft0, 12(sp) +; RV64IF-NEXT: fmv.x.w s1, ft0 +; RV64IF-NEXT: mv a0, s1 +; RV64IF-NEXT: call dummy +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: flw ft1, 12(sp) +; RV64IF-NEXT: feq.s a0, ft0, ft1 +; RV64IF-NEXT: beqz a0, .LBB17_3 +; RV64IF-NEXT: # %bb.1: # %if.end +; RV64IF-NEXT: mv a0, s1 +; RV64IF-NEXT: call dummy +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: flw ft1, 12(sp) +; RV64IF-NEXT: feq.s a0, ft0, ft1 +; RV64IF-NEXT: beqz a0, .LBB17_3 +; RV64IF-NEXT: # %bb.2: # %if.end4 +; RV64IF-NEXT: mv a0, zero +; RV64IF-NEXT: ld s1, 16(sp) +; RV64IF-NEXT: ld ra, 24(sp) +; RV64IF-NEXT: addi sp, sp, 32 +; RV64IF-NEXT: ret +; RV64IF-NEXT: .LBB17_3: # %if.then +; RV64IF-NEXT: call abort entry: %call = call float @dummy(float 0.000000e+00) %cmp = fcmp une float %call, 0.000000e+00 Index: llvm/trunk/test/CodeGen/RISCV/float-convert.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-convert.ll +++ llvm/trunk/test/CodeGen/RISCV/float-convert.ll @@ -1,23 +1,41 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s +; For RV64F, fcvt.l.s is semantically equivalent to fcvt.w.s in this case +; because fptosi will produce poison if the result doesn't fit into an i32. define i32 @fcvt_w_s(float %a) nounwind { ; RV32IF-LABEL: fcvt_w_s: ; RV32IF: # %bb.0: ; RV32IF-NEXT: fmv.w.x ft0, a0 ; RV32IF-NEXT: fcvt.w.s a0, ft0, rtz ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_w_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.l.s a0, ft0, rtz +; RV64IF-NEXT: ret %1 = fptosi float %a to i32 ret i32 %1 } +; For RV64F, fcvt.lu.s is semantically equivalent to fcvt.wu.s in this case +; because fptoui will produce poison if the result doesn't fit into an i32. define i32 @fcvt_wu_s(float %a) nounwind { ; RV32IF-LABEL: fcvt_wu_s: ; RV32IF: # %bb.0: ; RV32IF-NEXT: fmv.w.x ft0, a0 ; RV32IF-NEXT: fcvt.wu.s a0, ft0, rtz ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_wu_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.lu.s a0, ft0, rtz +; RV64IF-NEXT: ret %1 = fptoui float %a to i32 ret i32 %1 } @@ -30,6 +48,14 @@ ; RV32IF-NEXT: fadd.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmv_x_w: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret ; Ensure fmv.x.w is generated even for a soft float calling convention %1 = fadd float %a, %b %2 = bitcast float %1 to i32 @@ -42,6 +68,12 @@ ; RV32IF-NEXT: fcvt.s.w ft0, a0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_s_w: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.w ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = sitofp i32 %a to float ret float %1 } @@ -52,6 +84,12 @@ ; RV32IF-NEXT: fcvt.s.wu ft0, a0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_s_wu: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.wu ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = uitofp i32 %a to float ret float %1 } @@ -64,9 +102,93 @@ ; RV32IF-NEXT: fadd.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fmv_w_x: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret ; Ensure fmv.w.x is generated even for a soft float calling convention %1 = bitcast i32 %a to float %2 = bitcast i32 %b to float %3 = fadd float %1, %2 ret float %3 } + +define i64 @fcvt_l_s(float %a) nounwind { +; RV32IF-LABEL: fcvt_l_s: +; RV32IF: # %bb.0: +; RV32IF-NEXT: addi sp, sp, -16 +; RV32IF-NEXT: sw ra, 12(sp) +; RV32IF-NEXT: call __fixsfdi +; RV32IF-NEXT: lw ra, 12(sp) +; RV32IF-NEXT: addi sp, sp, 16 +; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_l_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.l.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptosi float %a to i64 + ret i64 %1 +} + +define i64 @fcvt_lu_s(float %a) nounwind { +; RV32IF-LABEL: fcvt_lu_s: +; RV32IF: # %bb.0: +; RV32IF-NEXT: addi sp, sp, -16 +; RV32IF-NEXT: sw ra, 12(sp) +; RV32IF-NEXT: call __fixunssfdi +; RV32IF-NEXT: lw ra, 12(sp) +; RV32IF-NEXT: addi sp, sp, 16 +; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_lu_s: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.lu.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptoui float %a to i64 + ret i64 %1 +} + +define float @fcvt_s_l(i64 %a) nounwind { +; RV32IF-LABEL: fcvt_s_l: +; RV32IF: # %bb.0: +; RV32IF-NEXT: addi sp, sp, -16 +; RV32IF-NEXT: sw ra, 12(sp) +; RV32IF-NEXT: call __floatdisf +; RV32IF-NEXT: lw ra, 12(sp) +; RV32IF-NEXT: addi sp, sp, 16 +; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_s_l: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.l ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = sitofp i64 %a to float + ret float %1 +} + +define float @fcvt_s_lu(i64 %a) nounwind { +; RV32IF-LABEL: fcvt_s_lu: +; RV32IF: # %bb.0: +; RV32IF-NEXT: addi sp, sp, -16 +; RV32IF-NEXT: sw ra, 12(sp) +; RV32IF-NEXT: call __floatundisf +; RV32IF-NEXT: lw ra, 12(sp) +; RV32IF-NEXT: addi sp, sp, 16 +; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcvt_s_lu: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.lu ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = uitofp i64 %a to float + ret float %1 +} Index: llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll +++ llvm/trunk/test/CodeGen/RISCV/float-fcmp.ll @@ -1,12 +1,19 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s define i32 @fcmp_false(float %a, float %b) nounwind { ; RV32IF-LABEL: fcmp_false: ; RV32IF: # %bb.0: ; RV32IF-NEXT: mv a0, zero ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_false: +; RV64IF: # %bb.0: +; RV64IF-NEXT: mv a0, zero +; RV64IF-NEXT: ret %1 = fcmp false float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -19,6 +26,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a0 ; RV32IF-NEXT: feq.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_oeq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp oeq float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -31,6 +45,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a1 ; RV32IF-NEXT: flt.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ogt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp ogt float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -43,6 +64,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a1 ; RV32IF-NEXT: fle.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_oge: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp oge float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -55,6 +83,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a0 ; RV32IF-NEXT: flt.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_olt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp olt float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -67,6 +102,13 @@ ; RV32IF-NEXT: fmv.w.x ft1, a0 ; RV32IF-NEXT: fle.s a0, ft1, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ole: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: ret %1 = fcmp ole float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -86,6 +128,20 @@ ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: and a0, a1, a0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_one: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: feq.s a0, ft1, ft1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: feq.s a1, ft0, ft1 +; RV64IF-NEXT: not a1, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: ret %1 = fcmp one float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -102,6 +158,17 @@ ; RV32IF-NEXT: seqz a0, a0 ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ord: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: feq.s a0, ft0, ft0 +; RV64IF-NEXT: and a0, a0, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: ret %1 = fcmp ord float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -119,6 +186,18 @@ ; RV32IF-NEXT: seqz a1, a1 ; RV32IF-NEXT: or a0, a0, a1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ueq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: feq.s a2, ft1, ft1 +; RV64IF-NEXT: and a1, a2, a1 +; RV64IF-NEXT: seqz a1, a1 +; RV64IF-NEXT: or a0, a0, a1 +; RV64IF-NEXT: ret %1 = fcmp ueq float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -132,6 +211,14 @@ ; RV32IF-NEXT: fle.s a0, ft1, ft0 ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ugt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: ret %1 = fcmp ugt float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -145,6 +232,14 @@ ; RV32IF-NEXT: flt.s a0, ft1, ft0 ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_uge: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: ret %1 = fcmp uge float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -158,6 +253,14 @@ ; RV32IF-NEXT: fle.s a0, ft1, ft0 ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ult: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: ret %1 = fcmp ult float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -171,6 +274,14 @@ ; RV32IF-NEXT: flt.s a0, ft1, ft0 ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_ule: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: ret %1 = fcmp ule float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -184,6 +295,14 @@ ; RV32IF-NEXT: feq.s a0, ft1, ft0 ; RV32IF-NEXT: xori a0, a0, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_une: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: ret %1 = fcmp une float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -199,6 +318,16 @@ ; RV32IF-NEXT: and a0, a0, a1 ; RV32IF-NEXT: seqz a0, a0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_uno: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: feq.s a0, ft0, ft0 +; RV64IF-NEXT: and a0, a0, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: ret %1 = fcmp uno float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 @@ -209,6 +338,11 @@ ; RV32IF: # %bb.0: ; RV32IF-NEXT: addi a0, zero, 1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fcmp_true: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi a0, zero, 1 +; RV64IF-NEXT: ret %1 = fcmp true float %a, %b %2 = zext i1 %1 to i32 ret i32 %2 Index: llvm/trunk/test/CodeGen/RISCV/float-imm.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-imm.ll +++ llvm/trunk/test/CodeGen/RISCV/float-imm.ll @@ -1,13 +1,24 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s +; TODO: constant pool shouldn't be necessary for RV64IF. define float @float_imm() nounwind { ; RV32IF-LABEL: float_imm: ; RV32IF: # %bb.0: ; RV32IF-NEXT: lui a0, 263313 ; RV32IF-NEXT: addi a0, a0, -37 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: float_imm: +; RV64IF: # %bb.0: +; RV64IF-NEXT: lui a0, %hi(.LCPI0_0) +; RV64IF-NEXT: addi a0, a0, %lo(.LCPI0_0) +; RV64IF-NEXT: flw ft0, 0(a0) +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret ret float 3.14159274101257324218750 } @@ -22,6 +33,16 @@ ; RV32IF-NEXT: fadd.s ft0, ft0, ft1 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: float_imm_op: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: lui a0, %hi(.LCPI1_0) +; RV64IF-NEXT: addi a0, a0, %lo(.LCPI1_0) +; RV64IF-NEXT: flw ft1, 0(a0) +; RV64IF-NEXT: fadd.s ft0, ft0, ft1 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fadd float %a, 1.0 ret float %1 } Index: llvm/trunk/test/CodeGen/RISCV/float-mem.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-mem.ll +++ llvm/trunk/test/CodeGen/RISCV/float-mem.ll @@ -1,6 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s define float @flw(float *%a) nounwind { ; RV32IF-LABEL: flw: @@ -10,6 +12,14 @@ ; RV32IF-NEXT: fadd.s ft0, ft1, ft0 ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: flw: +; RV64IF: # %bb.0: +; RV64IF-NEXT: flw ft0, 12(a0) +; RV64IF-NEXT: flw ft1, 0(a0) +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = load float, float* %a %2 = getelementptr float, float* %a, i32 3 %3 = load float, float* %2 @@ -30,6 +40,15 @@ ; RV32IF-NEXT: fsw ft0, 32(a0) ; RV32IF-NEXT: fsw ft0, 0(a0) ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fsw: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a2 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fsw ft0, 32(a0) +; RV64IF-NEXT: fsw ft0, 0(a0) +; RV64IF-NEXT: ret %1 = fadd float %b, %c store float %1, float* %a %2 = getelementptr float, float* %a, i32 8 @@ -56,6 +75,20 @@ ; RV32IF-NEXT: fsw ft0, 36(a0) ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: flw_fsw_global: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: lui a0, %hi(G) +; RV64IF-NEXT: flw ft1, %lo(G)(a0) +; RV64IF-NEXT: fsw ft0, %lo(G)(a0) +; RV64IF-NEXT: addi a0, a0, %lo(G) +; RV64IF-NEXT: flw ft1, 36(a0) +; RV64IF-NEXT: fsw ft0, 36(a0) +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fadd float %a, %b %2 = load volatile float, float* @G store float %1, float* @G @@ -76,6 +109,18 @@ ; RV32IF-NEXT: fsw ft0, -273(a0) ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: flw_fsw_constant: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: lui a0, 56 +; RV64IF-NEXT: addiw a0, a0, -1353 +; RV64IF-NEXT: slli a0, a0, 14 +; RV64IF-NEXT: flw ft1, -273(a0) +; RV64IF-NEXT: fadd.s ft0, ft0, ft1 +; RV64IF-NEXT: fsw ft0, -273(a0) +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = inttoptr i32 3735928559 to float* %2 = load volatile float, float* %1 %3 = fadd float %a, %2 @@ -102,6 +147,23 @@ ; RV32IF-NEXT: lw ra, 12(sp) ; RV32IF-NEXT: addi sp, sp, 16 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: flw_stack: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -32 +; RV64IF-NEXT: sd ra, 24(sp) +; RV64IF-NEXT: sd s1, 16(sp) +; RV64IF-NEXT: mv s1, a0 +; RV64IF-NEXT: addi a0, sp, 12 +; RV64IF-NEXT: call notdead +; RV64IF-NEXT: fmv.w.x ft0, s1 +; RV64IF-NEXT: flw ft1, 12(sp) +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ld s1, 16(sp) +; RV64IF-NEXT: ld ra, 24(sp) +; RV64IF-NEXT: addi sp, sp, 32 +; RV64IF-NEXT: ret %1 = alloca float, align 4 %2 = bitcast float* %1 to i8* call void @notdead(i8* %2) @@ -124,6 +186,20 @@ ; RV32IF-NEXT: lw ra, 12(sp) ; RV32IF-NEXT: addi sp, sp, 16 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: fsw_stack: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fsw ft0, 4(sp) +; RV64IF-NEXT: addi a0, sp, 4 +; RV64IF-NEXT: call notdead +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret %1 = fadd float %a, %b ; force store from FPR32 %2 = alloca float, align 4 store float %1, float* %2 Index: llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll +++ llvm/trunk/test/CodeGen/RISCV/float-select-fcmp.ll @@ -1,12 +1,19 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc -mtriple=riscv32 -mattr=+f -verify-machineinstrs < %s \ ; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s define float @select_fcmp_false(float %a, float %b) nounwind { ; RV32IF-LABEL: select_fcmp_false: ; RV32IF: # %bb.0: ; RV32IF-NEXT: mv a0, a1 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_false: +; RV64IF: # %bb.0: +; RV64IF-NEXT: mv a0, a1 +; RV64IF-NEXT: ret %1 = fcmp false float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -24,6 +31,18 @@ ; RV32IF-NEXT: .LBB1_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_oeq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: feq.s a0, ft0, ft1 +; RV64IF-NEXT: bnez a0, .LBB1_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB1_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp oeq float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -41,6 +60,18 @@ ; RV32IF-NEXT: .LBB2_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ogt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB2_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB2_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ogt float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -58,6 +89,18 @@ ; RV32IF-NEXT: .LBB3_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_oge: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB3_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB3_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp oge float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -75,6 +118,18 @@ ; RV32IF-NEXT: .LBB4_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_olt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: flt.s a0, ft0, ft1 +; RV64IF-NEXT: bnez a0, .LBB4_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB4_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp olt float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -92,6 +147,18 @@ ; RV32IF-NEXT: .LBB5_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ole: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fle.s a0, ft0, ft1 +; RV64IF-NEXT: bnez a0, .LBB5_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB5_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ole float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -117,6 +184,25 @@ ; RV32IF-NEXT: .LBB6_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_one: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: feq.s a0, ft1, ft1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: feq.s a1, ft0, ft1 +; RV64IF-NEXT: not a1, a1 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: bnez a0, .LBB6_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB6_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp one float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -138,6 +224,22 @@ ; RV32IF-NEXT: .LBB7_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ord: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: feq.s a0, ft1, ft1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB7_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB7_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ord float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -160,6 +262,23 @@ ; RV32IF-NEXT: .LBB8_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ueq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: feq.s a0, ft1, ft1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: feq.s a1, ft0, ft1 +; RV64IF-NEXT: or a0, a1, a0 +; RV64IF-NEXT: bnez a0, .LBB8_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB8_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ueq float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -178,6 +297,19 @@ ; RV32IF-NEXT: .LBB9_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ugt: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fle.s a0, ft0, ft1 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB9_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB9_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ugt float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -196,6 +328,19 @@ ; RV32IF-NEXT: .LBB10_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_uge: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: flt.s a0, ft0, ft1 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB10_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB10_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp uge float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -214,6 +359,19 @@ ; RV32IF-NEXT: .LBB11_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ult: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fle.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB11_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB11_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ult float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -232,6 +390,19 @@ ; RV32IF-NEXT: .LBB12_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_ule: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: flt.s a0, ft1, ft0 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB12_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB12_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp ule float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -250,6 +421,19 @@ ; RV32IF-NEXT: .LBB13_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_une: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: feq.s a0, ft0, ft1 +; RV64IF-NEXT: xori a0, a0, 1 +; RV64IF-NEXT: bnez a0, .LBB13_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB13_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp une float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -271,6 +455,21 @@ ; RV32IF-NEXT: .LBB14_2: ; RV32IF-NEXT: fmv.x.w a0, ft0 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_uno: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fmv.w.x ft1, a1 +; RV64IF-NEXT: feq.s a0, ft1, ft1 +; RV64IF-NEXT: feq.s a1, ft0, ft0 +; RV64IF-NEXT: and a0, a1, a0 +; RV64IF-NEXT: seqz a0, a0 +; RV64IF-NEXT: bnez a0, .LBB14_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: fmv.s ft0, ft1 +; RV64IF-NEXT: .LBB14_2: +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret %1 = fcmp uno float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -280,6 +479,10 @@ ; RV32IF-LABEL: select_fcmp_true: ; RV32IF: # %bb.0: ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: select_fcmp_true: +; RV64IF: # %bb.0: +; RV64IF-NEXT: ret %1 = fcmp true float %a, %b %2 = select i1 %1, float %a, float %b ret float %2 @@ -298,6 +501,18 @@ ; RV32IF-NEXT: .LBB16_2: ; RV32IF-NEXT: mv a0, a2 ; RV32IF-NEXT: ret +; +; RV64IF-LABEL: i32_select_fcmp_oeq: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: feq.s a0, ft1, ft0 +; RV64IF-NEXT: bnez a0, .LBB16_2 +; RV64IF-NEXT: # %bb.1: +; RV64IF-NEXT: mv a2, a3 +; RV64IF-NEXT: .LBB16_2: +; RV64IF-NEXT: mv a0, a2 +; RV64IF-NEXT: ret %1 = fcmp oeq float %a, %b %2 = select i1 %1, i32 %c, i32 %d ret i32 %2 Index: llvm/trunk/test/CodeGen/RISCV/rv32i-rv64i-float-double.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/rv32i-rv64i-float-double.ll +++ llvm/trunk/test/CodeGen/RISCV/rv32i-rv64i-float-double.ll @@ -0,0 +1,86 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV32IF %s +; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \ +; RUN: | FileCheck -check-prefix=RV64IF %s + +; This file provides a simple sanity check of float and double operations for +; RV32I and RV64I. This is primarily intended to ensure that custom +; legalisation or DAG combines aren't incorrectly triggered when the F +; extension isn't enabled. + +; TODO: f32 parameters on RV64 with a soft-float ABI are anyext. + +define float @float_test(float %a, float %b) nounwind { +; RV32IF-LABEL: float_test: +; RV32IF: # %bb.0: +; RV32IF-NEXT: addi sp, sp, -16 +; RV32IF-NEXT: sw ra, 12(sp) +; RV32IF-NEXT: sw s1, 8(sp) +; RV32IF-NEXT: mv s1, a1 +; RV32IF-NEXT: call __addsf3 +; RV32IF-NEXT: mv a1, s1 +; RV32IF-NEXT: call __divsf3 +; RV32IF-NEXT: lw s1, 8(sp) +; RV32IF-NEXT: lw ra, 12(sp) +; RV32IF-NEXT: addi sp, sp, 16 +; RV32IF-NEXT: ret +; +; RV64IF-LABEL: float_test: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: sd s1, 0(sp) +; RV64IF-NEXT: slli a0, a0, 32 +; RV64IF-NEXT: srli a0, a0, 32 +; RV64IF-NEXT: slli a1, a1, 32 +; RV64IF-NEXT: srli s1, a1, 32 +; RV64IF-NEXT: mv a1, s1 +; RV64IF-NEXT: call __addsf3 +; RV64IF-NEXT: mv a1, s1 +; RV64IF-NEXT: call __divsf3 +; RV64IF-NEXT: ld s1, 0(sp) +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret + %1 = fadd float %a, %b + %2 = fdiv float %1, %b + ret float %2 +} + +define double @double_test(double %a, double %b) nounwind { +; RV32IF-LABEL: double_test: +; RV32IF: # %bb.0: +; RV32IF-NEXT: addi sp, sp, -16 +; RV32IF-NEXT: sw ra, 12(sp) +; RV32IF-NEXT: sw s1, 8(sp) +; RV32IF-NEXT: sw s2, 4(sp) +; RV32IF-NEXT: mv s2, a3 +; RV32IF-NEXT: mv s1, a2 +; RV32IF-NEXT: call __adddf3 +; RV32IF-NEXT: mv a2, s1 +; RV32IF-NEXT: mv a3, s2 +; RV32IF-NEXT: call __divdf3 +; RV32IF-NEXT: lw s2, 4(sp) +; RV32IF-NEXT: lw s1, 8(sp) +; RV32IF-NEXT: lw ra, 12(sp) +; RV32IF-NEXT: addi sp, sp, 16 +; RV32IF-NEXT: ret +; +; RV64IF-LABEL: double_test: +; RV64IF: # %bb.0: +; RV64IF-NEXT: addi sp, sp, -16 +; RV64IF-NEXT: sd ra, 8(sp) +; RV64IF-NEXT: sd s1, 0(sp) +; RV64IF-NEXT: mv s1, a1 +; RV64IF-NEXT: call __adddf3 +; RV64IF-NEXT: mv a1, s1 +; RV64IF-NEXT: call __divdf3 +; RV64IF-NEXT: ld s1, 0(sp) +; RV64IF-NEXT: ld ra, 8(sp) +; RV64IF-NEXT: addi sp, sp, 16 +; RV64IF-NEXT: ret + %1 = fadd double %a, %b + %2 = fdiv double %1, %b + ret double %2 +} Index: llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll =================================================================== --- llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll +++ llvm/trunk/test/CodeGen/RISCV/rv64f-float-convert.ll @@ -0,0 +1,187 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \ +; RUN: | FileCheck %s -check-prefix=RV64IF + +; This file exhaustively checks float<->i32 conversions. In general, +; fcvt.l[u].s can be selected instead of fcvt.w[u].s because poison is +; generated for an fpto[s|u]i conversion if the result doesn't fit in the +; target type. + +define i32 @aext_fptosi(float %a) nounwind { +; RV64IF-LABEL: aext_fptosi: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.l.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptosi float %a to i32 + ret i32 %1 +} + +define signext i32 @sext_fptosi(float %a) nounwind { +; RV64IF-LABEL: sext_fptosi: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.l.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptosi float %a to i32 + ret i32 %1 +} + +define zeroext i32 @zext_fptosi(float %a) nounwind { +; RV64IF-LABEL: zext_fptosi: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.l.s a0, ft0, rtz +; RV64IF-NEXT: slli a0, a0, 32 +; RV64IF-NEXT: srli a0, a0, 32 +; RV64IF-NEXT: ret + %1 = fptosi float %a to i32 + ret i32 %1 +} + +define i32 @aext_fptoui(float %a) nounwind { +; RV64IF-LABEL: aext_fptoui: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.lu.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptoui float %a to i32 + ret i32 %1 +} + +define signext i32 @sext_fptoui(float %a) nounwind { +; RV64IF-LABEL: sext_fptoui: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.wu.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptoui float %a to i32 + ret i32 %1 +} + +define zeroext i32 @zext_fptoui(float %a) nounwind { +; RV64IF-LABEL: zext_fptoui: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a0 +; RV64IF-NEXT: fcvt.lu.s a0, ft0, rtz +; RV64IF-NEXT: ret + %1 = fptoui float %a to i32 + ret i32 %1 +} + +define i32 @bcvt_f32_to_aext_i32(float %a, float %b) nounwind { +; RV64IF-LABEL: bcvt_f32_to_aext_i32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = fadd float %a, %b + %2 = bitcast float %1 to i32 + ret i32 %2 +} + +define signext i32 @bcvt_f32_to_sext_i32(float %a, float %b) nounwind { +; RV64IF-LABEL: bcvt_f32_to_sext_i32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = fadd float %a, %b + %2 = bitcast float %1 to i32 + ret i32 %2 +} + +define zeroext i32 @bcvt_f32_to_zext_i32(float %a, float %b) nounwind { +; RV64IF-LABEL: bcvt_f32_to_zext_i32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: slli a0, a0, 32 +; RV64IF-NEXT: srli a0, a0, 32 +; RV64IF-NEXT: ret + %1 = fadd float %a, %b + %2 = bitcast float %1 to i32 + ret i32 %2 +} + +define float @bcvt_i64_to_f32_via_i32(i64 %a, i64 %b) nounwind { +; RV64IF-LABEL: bcvt_i64_to_f32_via_i32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fmv.w.x ft0, a1 +; RV64IF-NEXT: fmv.w.x ft1, a0 +; RV64IF-NEXT: fadd.s ft0, ft1, ft0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = trunc i64 %a to i32 + %2 = trunc i64 %b to i32 + %3 = bitcast i32 %1 to float + %4 = bitcast i32 %2 to float + %5 = fadd float %3, %4 + ret float %5 +} + +define float @uitofp_aext_i32_to_f32(i32 %a) nounwind { +; RV64IF-LABEL: uitofp_aext_i32_to_f32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.wu ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = uitofp i32 %a to float + ret float %1 +} + +define float @uitofp_sext_i32_to_f32(i32 signext %a) nounwind { +; RV64IF-LABEL: uitofp_sext_i32_to_f32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.wu ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = uitofp i32 %a to float + ret float %1 +} + +define float @uitofp_zext_i32_to_f32(i32 zeroext %a) nounwind { +; RV64IF-LABEL: uitofp_zext_i32_to_f32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.wu ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = uitofp i32 %a to float + ret float %1 +} + +define float @sitofp_aext_i32_to_f32(i32 %a) nounwind { +; RV64IF-LABEL: sitofp_aext_i32_to_f32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.w ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = sitofp i32 %a to float + ret float %1 +} + +define float @sitofp_sext_i32_to_f32(i32 signext %a) nounwind { +; RV64IF-LABEL: sitofp_sext_i32_to_f32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.l ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = sitofp i32 %a to float + ret float %1 +} + +define float @sitofp_zext_i32_to_f32(i32 zeroext %a) nounwind { +; RV64IF-LABEL: sitofp_zext_i32_to_f32: +; RV64IF: # %bb.0: +; RV64IF-NEXT: fcvt.s.w ft0, a0 +; RV64IF-NEXT: fmv.x.w a0, ft0 +; RV64IF-NEXT: ret + %1 = sitofp i32 %a to float + ret float %1 +}