From: Florian Hahn <flo@fhahn.com> Date: Wed, 11 Sep 2019 16:17:03 +0000 (+0000) Subject: Revert [InstCombine] Use SimplifyFMulInst to simplify multiply in fma. X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=c60c66e7ae06be18dd14d03c340ea752c299ba97;p=llvm Revert [InstCombine] Use SimplifyFMulInst to simplify multiply in fma. This introduces additional rounding error in some cases. See D67434. This reverts r371518 (git commit 18a1f0818b659cee13865b4fad2648d85984a4ed) git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@371634 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/lib/Transforms/InstCombine/InstCombineCalls.cpp b/lib/Transforms/InstCombine/InstCombineCalls.cpp index 1cef8d956b7..e96e45e802a 100644 --- a/lib/Transforms/InstCombine/InstCombineCalls.cpp +++ b/lib/Transforms/InstCombine/InstCombineCalls.cpp @@ -2258,11 +2258,9 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) { return II; } - // Try to simplify the underlying FMul. - if (Value *V = SimplifyFMulInst(II->getArgOperand(0), II->getArgOperand(1), - II->getFastMathFlags(), - SQ.getWithInstruction(II))) { - auto *FAdd = BinaryOperator::CreateFAdd(V, II->getArgOperand(2)); + // fma x, 1, z -> fadd x, z + if (match(Src1, m_FPOne())) { + auto *FAdd = BinaryOperator::CreateFAdd(Src0, II->getArgOperand(2)); FAdd->copyFastMathFlags(II); return FAdd; } diff --git a/test/Transforms/InstCombine/fma.ll b/test/Transforms/InstCombine/fma.ll index 42ed587bb3d..89fdc6bc9a5 100644 --- a/test/Transforms/InstCombine/fma.ll +++ b/test/Transforms/InstCombine/fma.ll @@ -372,7 +372,8 @@ define float @fmuladd_x_1_z_fast(float %x, float %z) { define <2 x double> @fmuladd_a_0_b(<2 x double> %a, <2 x double> %b) { ; CHECK-LABEL: @fmuladd_a_0_b( ; CHECK-NEXT: entry: -; CHECK-NEXT: ret <2 x double> [[B:%.*]] +; CHECK-NEXT: [[RES:%.*]] = call nnan nsz <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[A:%.*]], <2 x double> zeroinitializer, <2 x double> [[B:%.*]]) +; CHECK-NEXT: ret <2 x double> [[RES]] ; entry: %res = call nnan nsz <2 x double> @llvm.fmuladd.v2f64(<2 x double> %a, <2 x double> zeroinitializer, <2 x double> %b) @@ -382,7 +383,8 @@ entry: define <2 x double> @fmuladd_0_a_b(<2 x double> %a, <2 x double> %b) { ; CHECK-LABEL: @fmuladd_0_a_b( ; CHECK-NEXT: entry: -; CHECK-NEXT: ret <2 x double> [[B:%.*]] +; CHECK-NEXT: [[RES:%.*]] = call nnan nsz <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[A:%.*]], <2 x double> zeroinitializer, <2 x double> [[B:%.*]]) +; CHECK-NEXT: ret <2 x double> [[RES]] ; entry: %res = call nnan nsz <2 x double> @llvm.fmuladd.v2f64(<2 x double> zeroinitializer, <2 x double> %a, <2 x double> %b) @@ -405,7 +407,8 @@ declare <2 x double> @llvm.fmuladd.v2f64(<2 x double>, <2 x double>, <2 x double define <2 x double> @fma_a_0_b(<2 x double> %a, <2 x double> %b) { ; CHECK-LABEL: @fma_a_0_b( ; CHECK-NEXT: entry: -; CHECK-NEXT: ret <2 x double> [[B:%.*]] +; CHECK-NEXT: [[RES:%.*]] = call nnan nsz <2 x double> @llvm.fma.v2f64(<2 x double> [[A:%.*]], <2 x double> zeroinitializer, <2 x double> [[B:%.*]]) +; CHECK-NEXT: ret <2 x double> [[RES]] ; entry: %res = call nnan nsz <2 x double> @llvm.fma.v2f64(<2 x double> %a, <2 x double> zeroinitializer, <2 x double> %b) @@ -415,7 +418,8 @@ entry: define <2 x double> @fma_0_a_b(<2 x double> %a, <2 x double> %b) { ; CHECK-LABEL: @fma_0_a_b( ; CHECK-NEXT: entry: -; CHECK-NEXT: ret <2 x double> [[B:%.*]] +; CHECK-NEXT: [[RES:%.*]] = call nnan nsz <2 x double> @llvm.fma.v2f64(<2 x double> [[A:%.*]], <2 x double> zeroinitializer, <2 x double> [[B:%.*]]) +; CHECK-NEXT: ret <2 x double> [[RES]] ; entry: %res = call nnan nsz <2 x double> @llvm.fma.v2f64(<2 x double> zeroinitializer, <2 x double> %a, <2 x double> %b) @@ -436,7 +440,8 @@ entry: define <2 x double> @fma_sqrt(<2 x double> %a, <2 x double> %b) { ; CHECK-LABEL: @fma_sqrt( ; CHECK-NEXT: entry: -; CHECK-NEXT: [[RES:%.*]] = fadd fast <2 x double> [[A:%.*]], [[B:%.*]] +; CHECK-NEXT: [[SQRT:%.*]] = call fast <2 x double> @llvm.sqrt.v2f64(<2 x double> [[A:%.*]]) +; CHECK-NEXT: [[RES:%.*]] = call fast <2 x double> @llvm.fma.v2f64(<2 x double> [[SQRT]], <2 x double> [[SQRT]], <2 x double> [[B:%.*]]) ; CHECK-NEXT: ret <2 x double> [[RES]] ; entry: