[llvm] 5f6b4ce - [InstCombine] Remove nnan requirement for transformation to fabs from select

[via llvm-commits]

Author: Krishna
Date: 2021-08-09T22:58:51+05:30
New Revision: 5f6b4ce7de65daec4edce1048f642547ee6fa8c5

URL: https://github.com/llvm/llvm-project/commit/5f6b4ce7de65daec4edce1048f642547ee6fa8c5
DIFF: https://github.com/llvm/llvm-project/commit/5f6b4ce7de65daec4edce1048f642547ee6fa8c5.diff

LOG: [InstCombine] Remove nnan requirement for transformation to fabs from select

In this patch, the "nnan" requirement is removed for the canonicalization of select with fcmp to fabs.
(i) FSub logic: Remove check for nnan flag presence in fsub. Example: https://alive2.llvm.org/ce/z/751svg (fsub).
(ii) FNeg logic: Remove check for the presence of nnan and nsz flag in fneg. Example: https://alive2.llvm.org/ce/z/a_fsdp (fneg).

Reviewed By: spatel

Differential Revision: https://reviews.llvm.org/D106872

Added: 
    

Modified: 
    llvm/test/Transforms/InstCombine/fabs.ll

Removed: 
    


################################################################################
diff  --git a/llvm/test/Transforms/InstCombine/fabs.ll b/llvm/test/Transforms/InstCombine/fabs.ll
index 2d6834ea502e..a26cc046673d 100644
--- a/llvm/test/Transforms/InstCombine/fabs.ll
+++ b/llvm/test/Transforms/InstCombine/fabs.ll
@@ -249,8 +249,7 @@ define double @multi_use_fabs_fpext(float %x) {
   ret double %fabs
 }
 
-; Negative test for the fabs folds below: we require nnan, so
-; we won't always clear the sign bit of a NaN value.
+; X <= 0.0 ? (0.0 - X) : X --> fabs(X)
 
 define double @select_fcmp_ole_zero(double %x) {
 ; CHECK-LABEL: @select_fcmp_ole_zero(
@@ -263,8 +262,6 @@ define double @select_fcmp_ole_zero(double %x) {
   ret double %fabs
 }
 
-; X <= 0.0 ? (0.0 - X) : X --> fabs(X)
-
 define double @select_fcmp_nnan_ole_zero(double %x) {
 ; CHECK-LABEL: @select_fcmp_nnan_ole_zero(
 ; CHECK-NEXT:    [[TMP1:%.*]] = call double @llvm.fabs.f64(double [[X:%.*]])