Move three folds for FADD, FSUB and FMUL in the DAG combiner away from Unsafe to...

Summary: Unsafe does not map well alone for each of these three cases as it is missing NoNan context when accessed directly with clang.  I have migrated the fold guards to reflect the expectations of handing nan and zero contexts directly (NoNan, NSZ) and some tests with it.  Unsafe does include NSZ, however there is already precedent for using the target option directly to reflect that context.

Reviewers: spatel, wristow, hfinkel, craig.topper, arsenm

Reviewed By: arsenm

Subscribers: michele.scandale, wdng, javed.absar

Differential Revision: https://reviews.llvm.org/D64450

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@365679 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 232bfe46dc3cc268fa2077e431b39eae42b2ff12..320db83566b681d090c1d6d889d6c2f4199f643b 100644 (file)
--- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -12030,8 +12030,8 @@ SDValue DAGCombiner::visitFADD(SDNode *N) {
   // Selection pass has a hard time dealing with FP constants.
   bool AllowNewConst = (Level < AfterLegalizeDAG);
 
-  // If 'unsafe math' or nnan is enabled, fold lots of things.
-  if ((Options.UnsafeFPMath || Flags.hasNoNaNs()) && AllowNewConst) {
+  // If nnan is enabled, fold lots of things.
+  if ((Options.NoNaNsFPMath || Flags.hasNoNaNs()) && AllowNewConst) {
     // If allowed, fold (fadd (fneg x), x) -> 0.0
     if (N0.getOpcode() == ISD::FNEG && N0.getOperand(0) == N1)
       return DAG.getConstantFP(0.0, DL, VT);
@@ -12171,7 +12171,7 @@ SDValue DAGCombiner::visitFSUB(SDNode *N) {
 
   if (N0 == N1) {
     // (fsub x, x) -> 0.0
-    if (Options.UnsafeFPMath || Flags.hasNoNaNs())
+    if (Options.NoNaNsFPMath || Flags.hasNoNaNs())
       return DAG.getConstantFP(0.0f, DL, VT);
   }
 
@@ -12249,7 +12249,7 @@ SDValue DAGCombiner::visitFMUL(SDNode *N) {
   if (SDValue NewSel = foldBinOpIntoSelect(N))
     return NewSel;
 
-  if (Options.UnsafeFPMath ||
+  if ((Options.NoNaNsFPMath && Options.NoSignedZerosFPMath) ||
       (Flags.hasNoNaNs() && Flags.hasNoSignedZeros())) {
     // fold (fmul A, 0) -> 0
     if (N1CFP && N1CFP->isZero())

diff --git a/test/CodeGen/ARM/unsafe-fsub.ll b/test/CodeGen/ARM/nnan-fsub.ll
similarity index 76%
rename from test/CodeGen/ARM/unsafe-fsub.ll
rename to test/CodeGen/ARM/nnan-fsub.ll
index 0c5702aa5faba3029960f4598affaf70d2408f49..01839083547be354f8eaf864a75fb832894346d3 100644 (file)
--- a/test/CodeGen/ARM/unsafe-fsub.ll
+++ b/test/CodeGen/ARM/nnan-fsub.ll
@@ -1,5 +1,5 @@
 ; RUN: llc -mcpu=cortex-a9 < %s | FileCheck -check-prefix=SAFE %s
-; RUN: llc -mcpu=cortex-a9 -enable-unsafe-fp-math < %s | FileCheck -check-prefix=FAST %s
+; RUN: llc -mcpu=cortex-a9 --enable-no-nans-fp-math < %s | FileCheck -check-prefix=FAST %s
 
 target triple = "armv7-apple-ios"
 

diff --git a/test/CodeGen/X86/fmul-combines.ll b/test/CodeGen/X86/fmul-combines.ll
index f9843dced1b4216aadb9f6c1149249a4df00cb01..3a2b7bc4a828c8ed77e775b5c64d7a3457b90ec5 100644 (file)
--- a/test/CodeGen/X86/fmul-combines.ll
+++ b/test/CodeGen/X86/fmul-combines.ll
@@ -76,12 +76,12 @@ define <4 x float> @constant_fold_fmul_v4f32_undef(<4 x float> %x) {
   ret <4 x float> %y
 }
 
-define <4 x float> @fmul0_v4f32(<4 x float> %x) #0 {
-; CHECK-LABEL: fmul0_v4f32:
+define <4 x float> @fmul0_v4f32_nsz_nnan(<4 x float> %x) #0 {
+; CHECK-LABEL: fmul0_v4f32_nsz_nnan:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    xorps %xmm0, %xmm0
 ; CHECK-NEXT:    retq
-  %y = fmul <4 x float> %x, <float 0.0, float 0.0, float 0.0, float 0.0>
+  %y = fmul nnan nsz <4 x float> %x, <float 0.0, float 0.0, float 0.0, float 0.0>
   ret <4 x float> %y
 }
 
@@ -90,7 +90,7 @@ define <4 x float> @fmul0_v4f32_undef(<4 x float> %x) #0 {
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    xorps %xmm0, %xmm0
 ; CHECK-NEXT:    retq
-  %y = fmul <4 x float> %x, <float undef, float 0.0, float undef, float 0.0>
+  %y = fmul nnan nsz <4 x float> %x, <float undef, float 0.0, float undef, float 0.0>
   ret <4 x float> %y
 }
 

diff --git a/test/CodeGen/X86/fp-fast.ll b/test/CodeGen/X86/fp-fast.ll
index 7abedfd2d171b30e5d0735932c761fd849ee7af6..81dd983d94c3f9a9cca27610f39707e493df2057 100644 (file)
--- a/test/CodeGen/X86/fp-fast.ll
+++ b/test/CodeGen/X86/fp-fast.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=avx -enable-unsafe-fp-math < %s | FileCheck %s
+; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=avx -enable-unsafe-fp-math --enable-no-nans-fp-math < %s | FileCheck %s
 
 define float @test1(float %a) {
 ; CHECK-LABEL: test1: