[llvm] Zjaffal/fix fold fdiv sqrt (PR #81970)

[via llvm-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-llvm-transforms

Author: Zain Jaffal (zjaffal)

<details>
<summary>Changes</summary>

This patch fixes the issues introduced in https://github.com/llvm/llvm-project/commit/bb5c3899d1936ebdf7ebf5ca4347ee2e057bee7f. 

I moved the check for the instruction to be div before I check for the fast math flags which resolves the crash in 

```
float a, b;
double sqrt();
void c() { b = a / sqrt(a); }
```

---
Full diff: https://github.com/llvm/llvm-project/pull/81970.diff


2 Files Affected:

- (modified) llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp (+31) 
- (modified) llvm/test/Transforms/InstCombine/fdiv-sqrt.ll (+26-9) 


``````````diff
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
index 0bd4b6d1a835af..912d9ac404052a 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineMulDivRem.cpp
@@ -1706,6 +1706,34 @@ static Instruction *foldFDivPowDivisor(BinaryOperator &I,
   return BinaryOperator::CreateFMulFMF(Op0, Pow, &I);
 }
 
+/// Convert div to mul if we have an sqrt divisor iff sqrt's operand is a fdiv
+/// instruction.
+static Instruction *foldFDivSqrtDivisor(BinaryOperator &I,
+                                        InstCombiner::BuilderTy &Builder) {
+  // X / sqrt(Y / Z) -->  X * sqrt(Z / Y)
+  if (!I.hasAllowReassoc() || !I.hasAllowReciprocal())
+    return nullptr;
+  Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
+  auto *II = dyn_cast<IntrinsicInst>(Op1);
+  if (!II || II->getIntrinsicID() != Intrinsic::sqrt || !II->hasOneUse() ||
+      !II->hasAllowReassoc() || !II->hasAllowReciprocal())
+    return nullptr;
+
+  Value *Y, *Z;
+  auto *DivOp = dyn_cast<Instruction>(II->getOperand(0));
+  if (!DivOp)
+    return nullptr;
+  if (!match(DivOp, m_FDiv(m_Value(Y), m_Value(Z))))
+    return nullptr;
+  if (!DivOp->hasAllowReassoc() || !I.hasAllowReciprocal() ||
+      !DivOp->hasOneUse())
+    return nullptr;
+  Value *SwapDiv = Builder.CreateFDivFMF(Z, Y, DivOp);
+  Value *NewSqrt =
+      Builder.CreateUnaryIntrinsic(II->getIntrinsicID(), SwapDiv, II);
+  return BinaryOperator::CreateFMulFMF(Op0, NewSqrt, &I);
+}
+
 Instruction *InstCombinerImpl::visitFDiv(BinaryOperator &I) {
   Module *M = I.getModule();
 
@@ -1813,6 +1841,9 @@ Instruction *InstCombinerImpl::visitFDiv(BinaryOperator &I) {
   if (Instruction *Mul = foldFDivPowDivisor(I, Builder))
     return Mul;
 
+  if (Instruction *Mul = foldFDivSqrtDivisor(I, Builder))
+    return Mul;
+
   // pow(X, Y) / X --> pow(X, Y-1)
   if (I.hasAllowReassoc() &&
       match(Op0, m_OneUse(m_Intrinsic<Intrinsic::pow>(m_Specific(Op1),
diff --git a/llvm/test/Transforms/InstCombine/fdiv-sqrt.ll b/llvm/test/Transforms/InstCombine/fdiv-sqrt.ll
index 346271be7da761..58cc7c297e90a0 100644
--- a/llvm/test/Transforms/InstCombine/fdiv-sqrt.ll
+++ b/llvm/test/Transforms/InstCombine/fdiv-sqrt.ll
@@ -6,9 +6,9 @@ declare double @llvm.sqrt.f64(double)
 define double @sqrt_div_fast(double %x, double %y, double %z) {
 ; CHECK-LABEL: @sqrt_div_fast(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[DIV:%.*]] = fdiv fast double [[Y:%.*]], [[Z:%.*]]
-; CHECK-NEXT:    [[SQRT:%.*]] = call fast double @llvm.sqrt.f64(double [[DIV]])
-; CHECK-NEXT:    [[DIV1:%.*]] = fdiv fast double [[X:%.*]], [[SQRT]]
+; CHECK-NEXT:    [[TMP0:%.*]] = fdiv fast double [[Z:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call fast double @llvm.sqrt.f64(double [[TMP0]])
+; CHECK-NEXT:    [[DIV1:%.*]] = fmul fast double [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    ret double [[DIV1]]
 ;
 entry:
@@ -36,9 +36,9 @@ entry:
 define double @sqrt_div_reassoc_arcp(double %x, double %y, double %z) {
 ; CHECK-LABEL: @sqrt_div_reassoc_arcp(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[DIV:%.*]] = fdiv reassoc arcp double [[Y:%.*]], [[Z:%.*]]
-; CHECK-NEXT:    [[SQRT:%.*]] = call reassoc arcp double @llvm.sqrt.f64(double [[DIV]])
-; CHECK-NEXT:    [[DIV1:%.*]] = fdiv reassoc arcp double [[X:%.*]], [[SQRT]]
+; CHECK-NEXT:    [[TMP0:%.*]] = fdiv reassoc arcp double [[Z:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call reassoc arcp double @llvm.sqrt.f64(double [[TMP0]])
+; CHECK-NEXT:    [[DIV1:%.*]] = fmul reassoc arcp double [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    ret double [[DIV1]]
 ;
 entry:
@@ -96,9 +96,9 @@ entry:
 define double @sqrt_div_arcp_missing(double %x, double %y, double %z) {
 ; CHECK-LABEL: @sqrt_div_arcp_missing(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[DIV:%.*]] = fdiv reassoc double [[Y:%.*]], [[Z:%.*]]
-; CHECK-NEXT:    [[SQRT:%.*]] = call reassoc arcp double @llvm.sqrt.f64(double [[DIV]])
-; CHECK-NEXT:    [[DIV1:%.*]] = fdiv reassoc arcp double [[X:%.*]], [[SQRT]]
+; CHECK-NEXT:    [[TMP0:%.*]] = fdiv reassoc double [[Z:%.*]], [[Y:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call reassoc arcp double @llvm.sqrt.f64(double [[TMP0]])
+; CHECK-NEXT:    [[DIV1:%.*]] = fmul reassoc arcp double [[TMP1]], [[X:%.*]]
 ; CHECK-NEXT:    ret double [[DIV1]]
 ;
 entry:
@@ -173,3 +173,20 @@ entry:
   ret double %div1
 }
 
+; Function Attrs: nounwind ssp uwtable(sync)
+define float @sqrt_non_div_operator(float %a) {
+; CHECK-LABEL: @sqrt_non_div_operator(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CONV:%.*]] = fpext float [[A:%.*]] to double
+; CHECK-NEXT:    [[SQRT:%.*]] = call fast double @llvm.sqrt.f64(double [[CONV]])
+; CHECK-NEXT:    [[DIV:%.*]] = fdiv fast double [[CONV]], [[SQRT]]
+; CHECK-NEXT:    [[CONV2:%.*]] = fptrunc double [[DIV]] to float
+; CHECK-NEXT:    ret float [[CONV2]]
+;
+entry:
+  %conv = fpext float %a to double
+  %sqrt = call fast double @llvm.sqrt.f64(double %conv)
+  %div = fdiv fast double %conv, %sqrt
+  %conv2 = fptrunc double %div to float
+  ret float %conv2
+}

``````````

</details>


https://github.com/llvm/llvm-project/pull/81970