[PATCH] D119746: [FPEnv][InstSimplify] Fold fsub X, -0 ==> X, when we know X is not -0

[Kevin P. Neal via Phabricator via llvm-commits]

kpn created this revision.
kpn added reviewers: sepavloff, spatel, nlopes, lebedev.ri.
Herald added a subscriber: hiraditya.
kpn requested review of this revision.
Herald added a project: LLVM.
Herald added a subscriber: llvm-commits.

Currently the fsub optimizations in InstSimplify don't know how to fold X - -0.0 to X when we know X is not zero and the constrained intrinsics are used. This adds the support.

This review is split out from D107285 <https://reviews.llvm.org/D107285>.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D119746

Files:
  llvm/lib/Analysis/InstructionSimplify.cpp
  llvm/test/Transforms/InstSimplify/strictfp-fsub.ll


Index: llvm/test/Transforms/InstSimplify/strictfp-fsub.ll
===================================================================
--- llvm/test/Transforms/InstSimplify/strictfp-fsub.ll
+++ llvm/test/Transforms/InstSimplify/strictfp-fsub.ll
@@ -127,11 +127,9 @@
   ret float %sub
 }
 
-; TODO: This will fold if we allow non-default floating point environments.
 define float @fold_fsub_nnan_nsz_x_n0_ebmaytrap(float %a) #0 {
 ; CHECK-LABEL: @fold_fsub_nnan_nsz_x_n0_ebmaytrap(
-; CHECK-NEXT:    [[SUB:%.*]] = call nnan nsz float @llvm.experimental.constrained.fsub.f32(float [[A:%.*]], float -0.000000e+00, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
-; CHECK-NEXT:    ret float [[SUB]]
+; CHECK-NEXT:    ret float [[A:%.*]]
 ;
   %sub = call nnan nsz float @llvm.experimental.constrained.fsub.f32(float %a, float -0.0, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
   ret float %sub
@@ -147,12 +145,11 @@
   ret float %sub
 }
 
-; TODO: This will fold if we allow non-default floating point environments.
 ; TODO: The instruction is expected to remain, but the result isn't used.
 define float @fold_fsub_nsz_nnan_x_n0_ebstrict(float %a) #0 {
 ; CHECK-LABEL: @fold_fsub_nsz_nnan_x_n0_ebstrict(
 ; CHECK-NEXT:    [[SUB:%.*]] = call nnan nsz float @llvm.experimental.constrained.fsub.f32(float [[A:%.*]], float -0.000000e+00, metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
-; CHECK-NEXT:    ret float [[SUB]]
+; CHECK-NEXT:    ret float [[A]]
 ;
   %sub = call nsz nnan float @llvm.experimental.constrained.fsub.f32(float %a, float -0.0, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
   ret float %sub
@@ -185,12 +182,10 @@
   ret float %sub
 }
 
-; TODO: This will fold if we allow non-default floating point environments.
 define float @fold_fsub_fabs_nnan_x_n0_ebmaytrap(float %a) #0 {
 ; CHECK-LABEL: @fold_fsub_fabs_nnan_x_n0_ebmaytrap(
 ; CHECK-NEXT:    [[ABSA:%.*]] = call float @llvm.fabs.f32(float [[A:%.*]])
-; CHECK-NEXT:    [[SUB:%.*]] = call nnan float @llvm.experimental.constrained.fsub.f32(float [[ABSA]], float -0.000000e+00, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #[[ATTR0]]
-; CHECK-NEXT:    ret float [[SUB]]
+; CHECK-NEXT:    ret float [[ABSA]]
 ;
   %absa = call float @llvm.fabs.f32(float %a)
   %sub = call nnan float @llvm.experimental.constrained.fsub.f32(float %absa, float -0.0, metadata !"round.tonearest", metadata !"fpexcept.maytrap") #0
@@ -209,13 +204,12 @@
   ret float %sub
 }
 
-; TODO: This will fold if we allow non-default floating point environments.
 ; TODO: The instruction is expected to remain, but the result isn't used.
 define float @fold_fsub_fabs_nnan_x_n0_ebstrict(float %a) #0 {
 ; CHECK-LABEL: @fold_fsub_fabs_nnan_x_n0_ebstrict(
 ; CHECK-NEXT:    [[ABSA:%.*]] = call float @llvm.fabs.f32(float [[A:%.*]])
 ; CHECK-NEXT:    [[SUB:%.*]] = call nnan float @llvm.experimental.constrained.fsub.f32(float [[ABSA]], float -0.000000e+00, metadata !"round.tonearest", metadata !"fpexcept.strict") #[[ATTR0]]
-; CHECK-NEXT:    ret float [[SUB]]
+; CHECK-NEXT:    ret float [[ABSA]]
 ;
   %absa = call float @llvm.fabs.f32(float %a)
   %sub = call nnan float @llvm.experimental.constrained.fsub.f32(float %absa, float -0.0, metadata !"round.tonearest", metadata !"fpexcept.strict") #0
Index: llvm/lib/Analysis/InstructionSimplify.cpp
===================================================================
--- llvm/lib/Analysis/InstructionSimplify.cpp
+++ llvm/lib/Analysis/InstructionSimplify.cpp
@@ -5143,14 +5143,15 @@
     if (match(Op1, m_PosZeroFP()))
       return Op0;
 
+  // fsub X, -0 ==> X, when we know X is not -0
+  if (canIgnoreSNaN(ExBehavior, FMF))
+    if (match(Op1, m_NegZeroFP()) &&
+        (FMF.noSignedZeros() || CannotBeNegativeZero(Op0, Q.TLI)))
+      return Op0;
+
   if (!isDefaultFPEnvironment(ExBehavior, Rounding))
     return nullptr;
 
-  // fsub X, -0 ==> X, when we know X is not -0
-  if (match(Op1, m_NegZeroFP()) &&
-      (FMF.noSignedZeros() || CannotBeNegativeZero(Op0, Q.TLI)))
-    return Op0;
-
   // fsub -0.0, (fsub -0.0, X) ==> X
   // fsub -0.0, (fneg X) ==> X
   Value *X;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D119746.408511.patch
Type: text/x-patch
Size: 4156 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220214/ec4236eb/attachment.bin>