[llvm] 60bffe2 - [InstCombine] Handle commuted variant of sqrt transform

[Nikita Popov via llvm-commits]

Author: Nikita Popov
Date: 2024-08-16T16:47:13+02:00
New Revision: 60bffe221a1d615ffc7c6b632287d0fbd27ef863

URL: https://github.com/llvm/llvm-project/commit/60bffe221a1d615ffc7c6b632287d0fbd27ef863
DIFF: https://github.com/llvm/llvm-project/commit/60bffe221a1d615ffc7c6b632287d0fbd27ef863.diff

LOG: [InstCombine] Handle commuted variant of sqrt transform

Added: 
    

Modified: 
    llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
    llvm/test/Transforms/InstCombine/fast-math.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
index be4d4590cfd839..fb2efe581ac6bb 100644
--- a/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -2758,14 +2758,17 @@ Value *LibCallSimplifier::optimizeSqrt(CallInst *CI, IRBuilderBase &B) {
     // Note: We don't bother looking any deeper than this first level or for
     // variations of this pattern because instcombine's visitFMUL and/or the
     // reassociation pass should give us this form.
-    Value *OtherMul0, *OtherMul1;
-    if (match(Op0, m_FMul(m_Value(OtherMul0), m_Value(OtherMul1)))) {
-      // Pattern: sqrt((x * y) * z)
-      if (OtherMul0 == OtherMul1 && cast<Instruction>(Op0)->isFast()) {
-        // Matched: sqrt((x * x) * z)
-        RepeatOp = OtherMul0;
-        OtherOp = Op1;
-      }
+    Value *MulOp;
+    if (match(Op0, m_FMul(m_Value(MulOp), m_Deferred(MulOp))) &&
+        cast<Instruction>(Op0)->isFast()) {
+      // Pattern: sqrt((x * x) * z)
+      RepeatOp = MulOp;
+      OtherOp = Op1;
+    } else if (match(Op1, m_FMul(m_Value(MulOp), m_Deferred(MulOp))) &&
+               cast<Instruction>(Op1)->isFast()) {
+      // Pattern: sqrt(z * (x * x))
+      RepeatOp = MulOp;
+      OtherOp = Op0;
     }
   }
   if (!RepeatOp)

diff  --git a/llvm/test/Transforms/InstCombine/fast-math.ll b/llvm/test/Transforms/InstCombine/fast-math.ll
index 7f83328d360d66..d7c90e82ab520a 100644
--- a/llvm/test/Transforms/InstCombine/fast-math.ll
+++ b/llvm/test/Transforms/InstCombine/fast-math.ll
@@ -712,9 +712,9 @@ define double @sqrt_intrinsic_three_args5(double %x, double %y) {
 define double @sqrt_intrinsic_three_args6(double %x, ptr %yp) {
 ; CHECK-LABEL: @sqrt_intrinsic_three_args6(
 ; CHECK-NEXT:    [[Y:%.*]] = load double, ptr [[YP:%.*]], align 8
-; CHECK-NEXT:    [[MUL:%.*]] = fmul fast double [[X:%.*]], [[X]]
-; CHECK-NEXT:    [[MUL2:%.*]] = fmul fast double [[Y]], [[MUL]]
-; CHECK-NEXT:    [[SQRT:%.*]] = call fast double @llvm.sqrt.f64(double [[MUL2]])
+; CHECK-NEXT:    [[FABS:%.*]] = call fast double @llvm.fabs.f64(double [[X:%.*]])
+; CHECK-NEXT:    [[SQRT1:%.*]] = call fast double @llvm.sqrt.f64(double [[Y]])
+; CHECK-NEXT:    [[SQRT:%.*]] = fmul fast double [[FABS]], [[SQRT1]]
 ; CHECK-NEXT:    ret double [[SQRT]]
 ;
   %y = load double, ptr %yp ; thwart complexity-based canonicalization