[InstCombine] Enable FAdd simplifications when user can ignore sign bit (#157757)

When FAdd result is used by fabs, we can safely ignore the sign bit of
fp zero. This patch enables an instruction simplification optimization
that folds fadd x, 0 ==> x, which would otherwise not work as the
compiler cannot prove that the zero isn't -0. But if the result of the
fadd is used by fabs we can simply ignore this and still do the
optimization.

Fixes #154238

Author: VedantParanjape

llvm/lib/Transforms/InstCombine/InstCombineAddSub.cpp

@@ -2002,6 +2002,16 @@ Instruction *InstCombinerImpl::visitFAdd(BinaryOperator &I) {
   if (Instruction *FoldedFAdd = foldBinOpIntoSelectOrPhi(I))
     return FoldedFAdd;
 
+  // B = fadd A, 0.0
+  // Z = Op B
+  // can be transformed into
+  // Z = Op A
+  // Where Op is such that we can ignore sign of 0 in fadd
+  Value *A;
+  if (match(&I, m_OneUse(m_FAdd(m_Value(A), m_AnyZeroFP()))) &&
+      canIgnoreSignBitOfZero(*I.use_begin()))
+    return replaceInstUsesWith(I, A);
+
   // (-X) + Y --> Y - X
   Value *X, *Y;
   if (match(&I, m_c_FAdd(m_FNeg(m_Value(X)), m_Value(Y))))
@@ -3145,6 +3155,16 @@ Instruction *InstCombinerImpl::visitFSub(BinaryOperator &I) {
   Value *X, *Y;
   Constant *C;
 
+  // B = fsub A, 0.0
+  // Z = Op B
+  // can be transformed into
+  // Z = Op A
+  // Where Op is such that we can ignore sign of 0 in fsub
+  Value *A;
+  if (match(&I, m_OneUse(m_FSub(m_Value(A), m_AnyZeroFP()))) &&
+      canIgnoreSignBitOfZero(*I.use_begin()))
+    return replaceInstUsesWith(I, A);
+
   Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
   // If Op0 is not -0.0 or we can ignore -0.0: Z - (X - Y) --> Z + (Y - X)
   // Canonicalize to fadd to make analysis easier.

llvm/test/Transforms/InstCombine/fold-fadd-with-zero-gh154238.ll

@@ -0,0 +1,37 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt < %s -passes=instcombine -S | FileCheck %s
+define float @src(float %arg1) {
+; CHECK-LABEL: define float @src(
+; CHECK-SAME: float [[ARG1:%.*]]) {
+; CHECK-NEXT:    [[V3:%.*]] = call float @llvm.fabs.f32(float [[ARG1]])
+; CHECK-NEXT:    ret float [[V3]]
+;
+  %v2 = fadd float %arg1, 0.000000e+00
+  %v3 = call float @llvm.fabs.f32(float %v2)
+  ret float %v3
+}
+
+define float @src2(float %arg1) {
+; CHECK-LABEL: define float @src2(
+; CHECK-SAME: float [[ARG1:%.*]]) {
+; CHECK-NEXT:    [[V2:%.*]] = fadd float [[ARG1]], 0.000000e+00
+; CHECK-NEXT:    [[V3:%.*]] = call float @llvm.fabs.f32(float [[V2]])
+; CHECK-NEXT:    [[V4:%.*]] = fsub float [[V2]], [[V3]]
+; CHECK-NEXT:    ret float [[V4]]
+;
+  %v2 = fadd float %arg1, 0.000000e+00
+  %v3 = call float @llvm.fabs.f32(float %v2)
+  %v4 = fsub float %v2, %v3
+  ret float %v4
+}
+
+define float @src_sub(float %arg1) {
+; CHECK-LABEL: define float @src_sub(
+; CHECK-SAME: float [[ARG1:%.*]]) {
+; CHECK-NEXT:    [[V3:%.*]] = call float @llvm.fabs.f32(float [[ARG1]])
+; CHECK-NEXT:    ret float [[V3]]
+;
+  %v2 = fsub float %arg1, 0.000000e+00
+  %v3 = call float @llvm.fabs.f32(float %v2)
+  ret float %v3
+}