[Fast-math] Disable "(C1/X)*C2 => (C1*C2)/X" if C1/X has multiple uses.

  If "C1/X" were having multiple uses, the only benefit of this
transformation is to potentially shorten critical path. But it is at the
cost of instroducing additional div.

  The additional div may or may not incur cost depending on how div is
implemented. If it is implemented using Newton–Raphson iteration, it dosen't
seem to incur any cost (FIXME). However, if the div blocks the entire
pipeline, that sounds to be pretty expensive. Let CodeGen to take care 
this transformation.

  This patch sees 6% on a benchmark.

rdar://15032743


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@191037 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 12 insertions, 0 deletions

diff --git a/test/Transforms/InstCombine/fast-math.ll b/test/Transforms/InstCombine/fast-math.ll
index a9a7015..d8ba2a5 100644
--- a/test/Transforms/InstCombine/fast-math.ll
+++ b/test/Transforms/InstCombine/fast-math.ll
@@ -202,6 +202,18 @@ define float @fmul2(float %f1) {
 ; CHECK: fdiv fast float 1.200000e+07, %f1
 }
 
+; X/C1 * C2 => X * (C2/C1) is disabled if X/C1 has multiple uses
+@fmul2_external = external global float
+define float @fmul2_disable(float %f1) {
+  %div = fdiv fast float 1.000000e+00, %f1 
+  store float %div, float* @fmul2_external
+  %mul = fmul fast float %div, 2.000000e+00
+  ret float %mul
+; CHECK-LABEL: @fmul2_disable
+; CHECK: store
+; CHECK: fmul fast
+}
+
 ; X/C1 * C2 => X * (C2/C1) (if C2/C1 is normal Fp)
 define float @fmul3(float %f1, float %f2) {
   %t1 = fdiv float %f1, 2.0e+3