diff options
| -rw-r--r-- | llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp | 8 | ||||
| -rw-r--r-- | llvm/test/Transforms/InstCombine/x86-sse.ll | 7 | ||||
| -rw-r--r-- | llvm/test/Transforms/InstCombine/x86-sse2.ll | 7 |
3 files changed, 11 insertions, 11 deletions
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp index da2617c57cf..856eb0c39c7 100644 --- a/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp +++ b/llvm/lib/Transforms/InstCombine/InstCombineSimplifyDemanded.cpp @@ -1229,11 +1229,12 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts, case Intrinsic::x86_sse_add_ss: case Intrinsic::x86_sse_sub_ss: case Intrinsic::x86_sse_mul_ss: + case Intrinsic::x86_sse_div_ss: case Intrinsic::x86_sse2_add_sd: case Intrinsic::x86_sse2_sub_sd: case Intrinsic::x86_sse2_mul_sd: + case Intrinsic::x86_sse2_div_sd: // TODO: Lower MIN/MAX/etc. - // TODO: Lower DIV (with rounding/exceptions checks). Value *LHS = II->getArgOperand(0); Value *RHS = II->getArgOperand(1); // Extract the element as scalars. @@ -1259,6 +1260,11 @@ Value *InstCombiner::SimplifyDemandedVectorElts(Value *V, APInt DemandedElts, TmpV = InsertNewInstWith(BinaryOperator::CreateFMul(LHS, RHS, II->getName()), *II); break; + case Intrinsic::x86_sse_div_ss: + case Intrinsic::x86_sse2_div_sd: + TmpV = InsertNewInstWith(BinaryOperator::CreateFDiv(LHS, RHS, + II->getName()), *II); + break; } Instruction *New = diff --git a/llvm/test/Transforms/InstCombine/x86-sse.ll b/llvm/test/Transforms/InstCombine/x86-sse.ll index 831c55e340c..c465c53e401 100644 --- a/llvm/test/Transforms/InstCombine/x86-sse.ll +++ b/llvm/test/Transforms/InstCombine/x86-sse.ll @@ -235,11 +235,8 @@ define <4 x float> @test_div_ss(<4 x float> %a, <4 x float> %b) { define float @test_div_ss_0(float %a, float %b) { ; CHECK-LABEL: @test_div_ss_0( -; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x float> undef, float %a, i32 0 -; CHECK-NEXT: [[TMP2:%.*]] = insertelement <4 x float> undef, float %b, i32 0 -; CHECK-NEXT: [[TMP3:%.*]] = tail call <4 x float> @llvm.x86.sse.div.ss(<4 x float> [[TMP1]], <4 x float> [[TMP2]]) -; CHECK-NEXT: [[R:%.*]] = extractelement <4 x float> [[TMP3]], i32 0 -; CHECK-NEXT: ret float [[R]] +; CHECK-NEXT: [[TMP1:%.*]] = fdiv float %a, %b +; CHECK-NEXT: ret float [[TMP1]] ; %1 = insertelement <4 x float> undef, float %a, i32 0 %2 = insertelement <4 x float> %1, float 1.000000e+00, i32 1 diff --git a/llvm/test/Transforms/InstCombine/x86-sse2.ll b/llvm/test/Transforms/InstCombine/x86-sse2.ll index f0c0947ac18..560930bea6e 100644 --- a/llvm/test/Transforms/InstCombine/x86-sse2.ll +++ b/llvm/test/Transforms/InstCombine/x86-sse2.ll @@ -150,11 +150,8 @@ define <2 x double> @test_div_sd(<2 x double> %a, <2 x double> %b) { define double @test_div_sd_0(double %a, double %b) { ; CHECK-LABEL: @test_div_sd_0( -; CHECK-NEXT: [[TMP1:%.*]] = insertelement <2 x double> undef, double %a, i32 0 -; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x double> undef, double %b, i32 0 -; CHECK-NEXT: [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.sse2.div.sd(<2 x double> [[TMP1]], <2 x double> [[TMP2]]) -; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 0 -; CHECK-NEXT: ret double [[TMP4]] +; CHECK-NEXT: [[TMP1:%.*]] = fdiv double %a, %b +; CHECK-NEXT: ret double [[TMP1]] ; %1 = insertelement <2 x double> undef, double %a, i32 0 %2 = insertelement <2 x double> %1, double 1.000000e+00, i32 1 |
