diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp --- a/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp +++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeFloatTypes.cpp @@ -913,6 +913,24 @@ 0); } +// Even if the result type is legal, no libcall may exactly match. (e.g. We +// don't have FP-i8 conversions) This helper method looks for an appropriate +// promoted libcall. +static RTLIB::Libcall findFPToIntLibcall(EVT SrcVT, EVT RetVT, EVT &Promoted, + bool Signed) { + RTLIB::Libcall LC = RTLIB::UNKNOWN_LIBCALL; + for (unsigned IntVT = MVT::FIRST_INTEGER_VALUETYPE; + IntVT <= MVT::LAST_INTEGER_VALUETYPE && LC == RTLIB::UNKNOWN_LIBCALL; + ++IntVT) { + Promoted = (MVT::SimpleValueType)IntVT; + // The type needs to big enough to hold the result. + if (Promoted.bitsGE(RetVT)) + LC = Signed ? RTLIB::getFPTOSINT(SrcVT, Promoted) + : RTLIB::getFPTOUINT(SrcVT, Promoted); + } + return LC; +} + SDValue DAGTypeLegalizer::SoftenFloatOp_FP_TO_XINT(SDNode *N) { bool IsStrict = N->isStrictFPOpcode(); bool Signed = N->getOpcode() == ISD::FP_TO_SINT || @@ -928,16 +946,9 @@ // a larger type, eg: fp -> i32. Even if it is legal, no libcall may exactly // match, eg. we don't have fp -> i8 conversions. // Look for an appropriate libcall. - RTLIB::Libcall LC = RTLIB::UNKNOWN_LIBCALL; - for (unsigned IntVT = MVT::FIRST_INTEGER_VALUETYPE; - IntVT <= MVT::LAST_INTEGER_VALUETYPE && LC == RTLIB::UNKNOWN_LIBCALL; - ++IntVT) { - NVT = (MVT::SimpleValueType)IntVT; - // The type needs to big enough to hold the result. - if (NVT.bitsGE(RVT)) - LC = Signed ? RTLIB::getFPTOSINT(SVT, NVT) : RTLIB::getFPTOUINT(SVT, NVT); - } - assert(LC != RTLIB::UNKNOWN_LIBCALL && "Unsupported FP_TO_XINT!"); + RTLIB::Libcall LC = findFPToIntLibcall(SVT, RVT, NVT, Signed); + assert(LC != RTLIB::UNKNOWN_LIBCALL && NVT.isSimple() && + "Unsupported FP_TO_XINT!"); Op = GetSoftenedFloat(Op); SDValue Chain = IsStrict ? N->getOperand(0) : SDValue(); @@ -1895,12 +1906,14 @@ N->getOpcode() == ISD::STRICT_FP_TO_SINT; SDValue Op = N->getOperand(IsStrict ? 1 : 0); SDValue Chain = IsStrict ? N->getOperand(0) : SDValue(); - RTLIB::Libcall LC = Signed ? RTLIB::getFPTOSINT(Op.getValueType(), RVT) - : RTLIB::getFPTOUINT(Op.getValueType(), RVT); - assert(LC != RTLIB::UNKNOWN_LIBCALL && "Unsupported FP_TO_XINT!"); + + EVT NVT; + RTLIB::Libcall LC = findFPToIntLibcall(Op.getValueType(), RVT, NVT, Signed); + assert(LC != RTLIB::UNKNOWN_LIBCALL && NVT.isSimple() && + "Unsupported FP_TO_XINT!"); TargetLowering::MakeLibCallOptions CallOptions; std::pair Tmp = - TLI.makeLibCall(DAG, LC, RVT, Op, CallOptions, dl, Chain); + TLI.makeLibCall(DAG, LC, NVT, Op, CallOptions, dl, Chain); if (!IsStrict) return Tmp.first; diff --git a/llvm/test/CodeGen/PowerPC/ppcf128-constrained-fp-intrinsics.ll b/llvm/test/CodeGen/PowerPC/ppcf128-constrained-fp-intrinsics.ll --- a/llvm/test/CodeGen/PowerPC/ppcf128-constrained-fp-intrinsics.ll +++ b/llvm/test/CodeGen/PowerPC/ppcf128-constrained-fp-intrinsics.ll @@ -1975,6 +1975,100 @@ ret ppc_fp128 %conv } +define i1 @ppcq_to_s1(ppc_fp128 %a) { +; PC64LE-LABEL: ppcq_to_s1: +; PC64LE: # %bb.0: # %entry +; PC64LE-NEXT: mflr 0 +; PC64LE-NEXT: std 0, 16(1) +; PC64LE-NEXT: stdu 1, -32(1) +; PC64LE-NEXT: .cfi_def_cfa_offset 32 +; PC64LE-NEXT: .cfi_offset lr, 16 +; PC64LE-NEXT: bl __gcc_qtou +; PC64LE-NEXT: nop +; PC64LE-NEXT: addi 1, 1, 32 +; PC64LE-NEXT: ld 0, 16(1) +; PC64LE-NEXT: mtlr 0 +; PC64LE-NEXT: blr +; +; PC64LE9-LABEL: ppcq_to_s1: +; PC64LE9: # %bb.0: # %entry +; PC64LE9-NEXT: mflr 0 +; PC64LE9-NEXT: std 0, 16(1) +; PC64LE9-NEXT: stdu 1, -32(1) +; PC64LE9-NEXT: .cfi_def_cfa_offset 32 +; PC64LE9-NEXT: .cfi_offset lr, 16 +; PC64LE9-NEXT: bl __gcc_qtou +; PC64LE9-NEXT: nop +; PC64LE9-NEXT: addi 1, 1, 32 +; PC64LE9-NEXT: ld 0, 16(1) +; PC64LE9-NEXT: mtlr 0 +; PC64LE9-NEXT: blr +; +; PC64-LABEL: ppcq_to_s1: +; PC64: # %bb.0: # %entry +; PC64-NEXT: mflr 0 +; PC64-NEXT: std 0, 16(1) +; PC64-NEXT: stdu 1, -112(1) +; PC64-NEXT: .cfi_def_cfa_offset 112 +; PC64-NEXT: .cfi_offset lr, 16 +; PC64-NEXT: bl __gcc_qtou +; PC64-NEXT: nop +; PC64-NEXT: addi 1, 1, 112 +; PC64-NEXT: ld 0, 16(1) +; PC64-NEXT: mtlr 0 +; PC64-NEXT: blr +entry: + %conv = tail call i1 @llvm.experimental.constrained.fptosi.ppcf128.i1(ppc_fp128 %a, metadata !"fpexcept.strict") #1 + ret i1 %conv +} + +define i1 @ppcq_to_u1(ppc_fp128 %a) { +; PC64LE-LABEL: ppcq_to_u1: +; PC64LE: # %bb.0: # %entry +; PC64LE-NEXT: mflr 0 +; PC64LE-NEXT: std 0, 16(1) +; PC64LE-NEXT: stdu 1, -32(1) +; PC64LE-NEXT: .cfi_def_cfa_offset 32 +; PC64LE-NEXT: .cfi_offset lr, 16 +; PC64LE-NEXT: bl __fixunstfsi +; PC64LE-NEXT: nop +; PC64LE-NEXT: addi 1, 1, 32 +; PC64LE-NEXT: ld 0, 16(1) +; PC64LE-NEXT: mtlr 0 +; PC64LE-NEXT: blr +; +; PC64LE9-LABEL: ppcq_to_u1: +; PC64LE9: # %bb.0: # %entry +; PC64LE9-NEXT: mflr 0 +; PC64LE9-NEXT: std 0, 16(1) +; PC64LE9-NEXT: stdu 1, -32(1) +; PC64LE9-NEXT: .cfi_def_cfa_offset 32 +; PC64LE9-NEXT: .cfi_offset lr, 16 +; PC64LE9-NEXT: bl __fixunstfsi +; PC64LE9-NEXT: nop +; PC64LE9-NEXT: addi 1, 1, 32 +; PC64LE9-NEXT: ld 0, 16(1) +; PC64LE9-NEXT: mtlr 0 +; PC64LE9-NEXT: blr +; +; PC64-LABEL: ppcq_to_u1: +; PC64: # %bb.0: # %entry +; PC64-NEXT: mflr 0 +; PC64-NEXT: std 0, 16(1) +; PC64-NEXT: stdu 1, -112(1) +; PC64-NEXT: .cfi_def_cfa_offset 112 +; PC64-NEXT: .cfi_offset lr, 16 +; PC64-NEXT: bl __fixunstfsi +; PC64-NEXT: nop +; PC64-NEXT: addi 1, 1, 112 +; PC64-NEXT: ld 0, 16(1) +; PC64-NEXT: mtlr 0 +; PC64-NEXT: blr +entry: + %conv = tail call i1 @llvm.experimental.constrained.fptoui.ppcf128.i1(ppc_fp128 %a, metadata !"fpexcept.strict") #1 + ret i1 %conv +} + attributes #0 = { nounwind strictfp } attributes #1 = { strictfp } @@ -2008,8 +2102,10 @@ declare ppc_fp128 @llvm.experimental.constrained.trunc.ppcf128(ppc_fp128, metadata) declare i64 @llvm.experimental.constrained.fptosi.i64.ppcf128(ppc_fp128, metadata) declare i32 @llvm.experimental.constrained.fptosi.i32.ppcf128(ppc_fp128, metadata) +declare i1 @llvm.experimental.constrained.fptosi.ppcf128.i1(ppc_fp128, metadata) declare i64 @llvm.experimental.constrained.fptoui.i64.ppcf128(ppc_fp128, metadata) declare i32 @llvm.experimental.constrained.fptoui.i32.ppcf128(ppc_fp128, metadata) +declare i1 @llvm.experimental.constrained.fptoui.ppcf128.i1(ppc_fp128, metadata) declare ppc_fp128 @llvm.experimental.constrained.sitofp.ppcf128.i32(i32, metadata, metadata) declare ppc_fp128 @llvm.experimental.constrained.uitofp.ppcf128.i32(i32, metadata, metadata) declare ppc_fp128 @llvm.experimental.constrained.sitofp.ppcf128.i64(i64, metadata, metadata)