[llvm] r251747 - [SimplifyLibCalls] Add test to ensure transform is not executed if fast-math

Davide Italiano via llvm-commits llvm-commits at lists.llvm.org
Sat Oct 31 13:59:33 PDT 2015


Author: davide
Date: Sat Oct 31 15:59:32 2015
New Revision: 251747

URL: http://llvm.org/viewvc/llvm-project?rev=251747&view=rev
Log:
[SimplifyLibCalls] Add test to ensure transform is not executed if fast-math
attribute is not present.

During my refactor in r251595 I changed the behavior of optimizeSqrt(),
skipping the transformation if the function wasn't marked with unsafe-fp-math
attribute. This fixed a bug, as confirmed by Sanjay (before the optimization
was silently executed anyway), although it wasn't my primary aim.
This commit adds a test to ensure the code doesn't break again.

Reported by: Marcello Maggioni
Discussed with: Sanjay Patel

Added:
    llvm/trunk/test/Transforms/InstCombine/sqrt-nofast.ll

Added: llvm/trunk/test/Transforms/InstCombine/sqrt-nofast.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/sqrt-nofast.ll?rev=251747&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/sqrt-nofast.ll (added)
+++ llvm/trunk/test/Transforms/InstCombine/sqrt-nofast.ll Sat Oct 31 15:59:32 2015
@@ -0,0 +1,25 @@
+; Check that we skip transformations if the attribute unsafe-fp-math
+; is not set.
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+define float @mysqrt(float %x, float %y) #0 {
+entry:
+  %x.addr = alloca float, align 4
+  %y.addr = alloca float, align 4
+  store float %x, float* %x.addr, align 4
+  store float %y, float* %y.addr, align 4
+  %0 = load float, float* %x.addr, align 4
+  %1 = load float, float* %x.addr, align 4
+  %mul = fmul fast float %0, %1
+  %2 = call float @llvm.sqrt.f32(float %mul)
+  ret float %2
+}
+
+declare float @llvm.sqrt.f32(float) #1
+
+; CHECK: define float @mysqrt(float %x, float %y) {
+; CHECK: entry:
+; CHECK:   %mul = fmul fast float %x, %x
+; CHECK:   %0 = call float @llvm.sqrt.f32(float %mul)
+; CHECK:   ret float %0
+; CHECK: }




More information about the llvm-commits mailing list