llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT--> @llvm/pr-subscribers-llvm-transforms Author: Matt Arsenault (arsenm) <details> <summary>Changes</summary> --- Patch is 22.68 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/180703.diff 2 Files Affected: - (modified) llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp (+37) - (modified) llvm/test/Transforms/InstCombine/simplify-demanded-fpclass-ldexp.ll (+31-48) ``````````diff diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp index c097e74456ff7..7b3c50b4dda59 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp @@ -3129,6 +3129,43 @@ Value *InstCombinerImpl::SimplifyDemandedUseFPClass(Instruction *I, return simplifyDemandedFPClassResult(CI, FMF, DemandedMask, Known, {KnownSrc}); } + case Intrinsic::ldexp: { + FPClassTest SrcDemandedMask = DemandedMask & fcInf; + if (DemandedMask & fcNan) + SrcDemandedMask |= fcNan; + + if (DemandedMask & fcPosInf) + SrcDemandedMask |= fcPosNormal | fcPosSubnormal; + if (DemandedMask & fcNegInf) + SrcDemandedMask |= fcNegNormal | fcNegSubnormal; + + if (DemandedMask & (fcPosNormal | fcPosSubnormal)) + SrcDemandedMask |= fcPosNormal | fcPosSubnormal; + if (DemandedMask & (fcNegNormal | fcNegSubnormal)) + SrcDemandedMask |= fcNegNormal | fcNegSubnormal; + + if (DemandedMask & fcPosZero) + SrcDemandedMask |= fcPosFinite; + if (DemandedMask & fcNegZero) + SrcDemandedMask |= fcNegFinite; + + KnownFPClass KnownSrc; + if (SimplifyDemandedFPClass(CI, 0, SrcDemandedMask, KnownSrc, Depth + 1)) + return CI; + + Type *EltTy = VTy->getScalarType(); + const fltSemantics &FltSem = EltTy->getFltSemantics(); + DenormalMode Mode = F.getDenormalMode(FltSem); + + KnownBits KnownExpBits = + ::computeKnownBits(CI->getArgOperand(1), SQ, Depth + 1); + + Known = KnownFPClass::ldexp(KnownSrc, KnownExpBits, FltSem, Mode); + Known.knownNot(~DemandedMask); + + return simplifyDemandedFPClassResult(CI, FMF, DemandedMask, Known, + {KnownSrc}); + } case Intrinsic::trunc: case Intrinsic::floor: case Intrinsic::ceil: diff --git a/llvm/test/Transforms/InstCombine/simplify-demanded-fpclass-ldexp.ll b/llvm/test/Transforms/InstCombine/simplify-demanded-fpclass-ldexp.ll index 72d687375e946..b110f58f64c23 100644 --- a/llvm/test/Transforms/InstCombine/simplify-demanded-fpclass-ldexp.ll +++ b/llvm/test/Transforms/InstCombine/simplify-demanded-fpclass-ldexp.ll @@ -24,7 +24,7 @@ declare nofpclass(pinf pnorm psub pzero) float @returns_negative_or_nan() define nofpclass(qnan inf norm sub zero) float @ret_only_snan__ldexp(float %x, i32 %exp) { ; CHECK-LABEL: define nofpclass(qnan inf zero sub norm) float @ret_only_snan__ldexp( ; CHECK-SAME: float [[X:%.*]], i32 [[EXP:%.*]]) { -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %exp) @@ -34,8 +34,7 @@ define nofpclass(qnan inf norm sub zero) float @ret_only_snan__ldexp(float %x, i define nofpclass(snan inf norm sub zero) float @ret_only_qnan__ldexp(float %x, i32 %exp) { ; CHECK-LABEL: define nofpclass(snan inf zero sub norm) float @ret_only_qnan__ldexp( ; CHECK-SAME: float [[X:%.*]], i32 [[EXP:%.*]]) { -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) -; CHECK-NEXT: ret float [[LDEXP]] +; CHECK-NEXT: ret float 0x7FF8000000000000 ; %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %exp) ret float %ldexp @@ -44,8 +43,7 @@ define nofpclass(snan inf norm sub zero) float @ret_only_qnan__ldexp(float %x, i define nofpclass(inf norm sub zero) float @ret_only_nan__ldexp(float %x, i32 %exp) { ; CHECK-LABEL: define nofpclass(inf zero sub norm) float @ret_only_nan__ldexp( ; CHECK-SAME: float [[X:%.*]], i32 [[EXP:%.*]]) { -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) -; CHECK-NEXT: ret float [[LDEXP]] +; CHECK-NEXT: ret float 0x7FF8000000000000 ; %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %exp) ret float %ldexp @@ -54,7 +52,7 @@ define nofpclass(inf norm sub zero) float @ret_only_nan__ldexp(float %x, i32 %ex define nofpclass(nan norm sub zero) float @ret_only_inf__ldexp(float %x, i32 %exp) { ; CHECK-LABEL: define nofpclass(nan zero sub norm) float @ret_only_inf__ldexp( ; CHECK-SAME: float [[X:%.*]], i32 [[EXP:%.*]]) { -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %exp) @@ -82,7 +80,7 @@ define nofpclass(nan pinf norm sub zero) float @ret_only_ninf__ldexp(float %x, i define nofpclass(inf nan norm sub) float @ret_only_zero__ldexp(float %x, i32 %exp) { ; CHECK-LABEL: define nofpclass(nan inf sub norm) float @ret_only_zero__ldexp( ; CHECK-SAME: float [[X:%.*]], i32 [[EXP:%.*]]) { -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan ninf float @llvm.ldexp.f32.i32(float [[X]], i32 [[EXP]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %ldexp = call float @llvm.ldexp.f32.i32(float %x, i32 %exp) @@ -131,8 +129,7 @@ define nofpclass(nan) float @ret_nofpclass_nan__ldexp_select_nan_or_unknown__unk ; CHECK-LABEL: define nofpclass(nan) float @ret_nofpclass_nan__ldexp_select_nan_or_unknown__unknown( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[EXP:%.*]]) { ; CHECK-NEXT: [[NAN:%.*]] = call float @returns_nan() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NAN]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[EXP]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[EXP]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %nan = call float @returns_nan() @@ -145,8 +142,7 @@ define nofpclass(nan) <2 x float> @ret_nofpclass_nan__ldexp_select_nan_or_unknow ; CHECK-LABEL: define nofpclass(nan) <2 x float> @ret_nofpclass_nan__ldexp_select_nan_or_unknown__unknown_vec( ; CHECK-SAME: i1 [[COND:%.*]], <2 x float> [[UNKNOWN:%.*]], <2 x i32> [[EXP:%.*]]) { ; CHECK-NEXT: [[NAN:%.*]] = call <2 x float> @returns_nan_vec() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], <2 x float> [[NAN]], <2 x float> [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> [[SELECT]], <2 x i32> [[EXP]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan <2 x float> @llvm.ldexp.v2f32.v2i32(<2 x float> [[UNKNOWN]], <2 x i32> [[EXP]]) ; CHECK-NEXT: ret <2 x float> [[LDEXP]] ; %nan = call <2 x float> @returns_nan_vec() @@ -159,8 +155,7 @@ define nofpclass(inf) float @ret_nofpclass_inf__ldexp_select_inf_or_unknown__unk ; CHECK-LABEL: define nofpclass(inf) float @ret_nofpclass_inf__ldexp_select_inf_or_unknown__unknown( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[EXP:%.*]]) { ; CHECK-NEXT: [[INF:%.*]] = call float @returns_inf() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[INF]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[EXP]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[EXP]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %inf = call float @returns_inf() @@ -174,7 +169,7 @@ define nofpclass(nan inf norm sub) float @zero_result_demands_sub_source(i1 %con ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[SUB:%.*]] = call float @returns_sub() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[SUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %sub = call float @returns_sub() @@ -188,7 +183,7 @@ define nofpclass(inf norm psub nzero) float @pzero_result_demands_psub_source(i1 ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[PSUB:%.*]] = call float @returns_psub() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[PSUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %psub = call float @returns_psub() @@ -202,7 +197,7 @@ define nofpclass(inf norm nsub pzero) float @nzero_result_demands_nsub_source(i1 ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NSUB:%.*]] = call float @returns_nsub() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NSUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %nsub = call float @returns_nsub() @@ -216,7 +211,7 @@ define nofpclass(inf norm psub nzero) float @pzero_result_demands_pnorm_source(i ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[PNORM:%.*]] = call float @returns_pnorm() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[PNORM]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %pnorm = call float @returns_pnorm() @@ -230,7 +225,7 @@ define nofpclass(inf norm psub pzero) float @nzero_result_demands_nnorm_source(i ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NNORM:%.*]] = call float @returns_nnorm() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NNORM]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %nnorm = call float @returns_nnorm() @@ -243,8 +238,7 @@ define nofpclass(inf norm nsub nzero) float @pzero_result_no_demand_nnorm_source ; CHECK-LABEL: define nofpclass(inf nzero nsub norm) float @pzero_result_no_demand_nnorm_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NNORM:%.*]] = call float @returns_nnorm() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NNORM]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %nnorm = call float @returns_nnorm() @@ -257,8 +251,7 @@ define nofpclass(inf norm psub pzero) float @nzero_result_no_demand_pnorm_source ; CHECK-LABEL: define nofpclass(inf pzero psub norm) float @nzero_result_no_demand_pnorm_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[PNORM:%.*]] = call float @returns_pnorm() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[PNORM]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %pnorm = call float @returns_pnorm() @@ -271,8 +264,7 @@ define nofpclass(inf norm sub nzero) float @pzero_result_no_demand_nsub_source(i ; CHECK-LABEL: define nofpclass(inf nzero sub norm) float @pzero_result_no_demand_nsub_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NSUB:%.*]] = call float @returns_nsub() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NSUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %nsub = call float @returns_nsub() @@ -285,8 +277,7 @@ define nofpclass(inf norm sub pzero) float @nzero_result_no_demand_psub_source(i ; CHECK-LABEL: define nofpclass(inf pzero sub norm) float @nzero_result_no_demand_psub_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[PSUB:%.*]] = call float @returns_psub() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[PSUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call ninf float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %psub = call float @returns_psub() @@ -300,7 +291,7 @@ define nofpclass(nan inf norm sub) float @zero_result_demands_norm_source_lhs(i1 ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NORM:%.*]] = call float @returns_norm() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NORM]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %norm = call float @returns_norm() @@ -314,7 +305,7 @@ define nofpclass(nan inf sub zero) float @norm_result_demands_sub_source(i1 %con ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[SUB:%.*]] = call float @returns_sub() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[SUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan ninf float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %sub = call float @returns_sub() @@ -328,7 +319,7 @@ define nofpclass(nan norm sub) float @inf_result_demands_sub_source(i1 %cond, fl ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[SUB:%.*]] = call float @returns_sub() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[SUB]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %sub = call float @returns_sub() @@ -342,7 +333,7 @@ define nofpclass(nan norm sub) float @inf_result_demands_norm_source(i1 %cond, f ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NORM:%.*]] = call float @returns_norm() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NORM]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %norm = call float @returns_norm() @@ -356,7 +347,7 @@ define nofpclass(nan norm sub) float @inf_result_demands_inf_source(i1 %cond, fl ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[INF:%.*]] = call float @returns_inf() ; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[INF]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %inf = call float @returns_inf() @@ -369,8 +360,7 @@ define nofpclass(nan inf sub zero) float @zero_result_no_zero_source(i1 %cond, f ; CHECK-LABEL: define nofpclass(nan inf zero sub) float @zero_result_no_zero_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[ZERO:%.*]] = call float @returns_zero() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[ZERO]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan ninf float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %zero = call float @returns_zero() @@ -383,8 +373,7 @@ define nofpclass(pinf pnorm psub pzero) float @negative_result_implies_no_positi ; CHECK-LABEL: define nofpclass(pinf pzero psub pnorm) float @negative_result_implies_no_positive_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[POS:%.*]] = call float @returns_positive() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[POS]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %pos = call float @returns_positive() @@ -397,8 +386,7 @@ define nofpclass(ninf nnorm nsub nzero) float @positive_result_implies_no_negati ; CHECK-LABEL: define nofpclass(ninf nzero nsub nnorm) float @positive_result_implies_no_negative_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NEG:%.*]] = call float @returns_negative() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NEG]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %neg = call float @returns_negative() @@ -411,8 +399,7 @@ define nofpclass(nan pinf pnorm psub pzero) float @negative_nonan_result_implies ; CHECK-LABEL: define nofpclass(nan pinf pzero psub pnorm) float @negative_nonan_result_implies_no_positive_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[POS_OR_NAN:%.*]] = call float @returns_positive_or_nan() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[POS_OR_NAN]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %pos.or.nan = call float @returns_positive_or_nan() @@ -425,8 +412,7 @@ define nofpclass(nan ninf nnorm nsub nzero) float @positive_nonan_result_implies ; CHECK-LABEL: define nofpclass(nan ninf nzero nsub nnorm) float @positive_nonan_result_implies_no_negative_source( ; CHECK-SAME: i1 [[COND:%.*]], float [[UNKNOWN:%.*]], i32 [[UNKNOWN_INT:%.*]]) { ; CHECK-NEXT: [[NEG_OR_NAN:%.*]] = call float @returns_negative_or_nan() -; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], float [[NEG_OR_NAN]], float [[UNKNOWN]] -; CHECK-NEXT: [[LDEXP:%.*]] = call float @llvm.ldexp.f32.i32(float [[SELECT]], i32 [[UNKNOWN_INT]]) +; CHECK-NEXT: [[LDEXP:%.*]] = call nnan float @llvm.ldexp.f32.i32(float [[UNKNOWN]], i32 [[UNKNOWN_INT]]) ; CHECK-NEXT: ret float [[LDEXP]] ; %neg.or.nan = call float @returns_negative_or_nan() @@ -439,8 +425,7 @@ define no... [truncated] `````````` </details> https://github.com/llvm/llvm-project/pull/180703 _______________________________________________ llvm-branch-commits mailing list [email protected] https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
