[NARY] Don't optimize min/max if there are side uses

[ebrevnov]

Summary

Say we have
%1=min(%a,%b)
%2=min(%b,%c)
%3=min(%2,%a)

The optimization will try to reassociate the later one so that we can rewrite it to %3=min(%1, %c) and remove %2.
But if %2 has another uses outside of %3 then we can't remove %2 and end up with:

%1=min(%a,%b)
%2=min(%b,%c)
%3=min(%1, %c)

This doesn't harm by itself except it is not profitable and changes IR for no good reason.
What is bad it triggers next iteration which finds out that optimization is applicable to %2 and %3 and generates:

%1=min(%a,%b)
%2=min(%b,%c)
%3=min(%1,%c)
%4=min(%2,%a)

and so on...

The solution is to prevent optimization in the first place if intermediate result (%2) has side uses and
known to be not removed.

[mkazantsev]

LGTM

llvm/test/Transforms/NaryReassociate/nary-req.ll

6

Please precommit the test with current output and put your patch on top of it so that it clearly shows the difference.

[vdmitrie]

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

595

Just a hint:
/ This is a linear time operation. Use hasOneUse, hasNUses, or
/ hasNUsesOrMore to check for specific values.

[ebrevnov]

Thanks for noting. Fixed.

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

 
 template <typename MaxMinT>
 Value *NaryReassociatePass::tryReassociateMinOrMax(Instruction *I,
                                                    MaxMinT MaxMinMatch,
                                                    Value *LHS, Value *RHS) {
   Value *A = nullptr, *B = nullptr;
   MaxMinT m_MaxMin(m_Value(A), m_Value(B));
   for (unsigned int i = 0; i < 2; ++i) {
-    if (match(LHS, m_MaxMin)) {
+    if (LHS->getNumUses() <= 2 && match(LHS, m_MaxMin)) {
       const SCEV *AExpr = SE->getSCEV(A), *BExpr = SE->getSCEV(B);
       const SCEV *RHSExpr = SE->getSCEV(RHS);
       for (unsigned int j = 0; j < 2; ++j) {
         if (j == 0) {
           if (BExpr == RHSExpr)
             continue;
           // Transform 'I = (A op B) op RHS' to 'I = (A op RHS) op B' on the
           // first iteration.
           std::swap(BExpr, RHSExpr);
         } else {
           if (AExpr == RHSExpr)
             continue;
           // Transform 'I = (A op RHS) op B' 'I = (B op RHS) op A' on the second
           // iteration.
           std::swap(AExpr, RHSExpr);
         }
 
+        // The optimization is profitable only if LHS can be removed in the end.
+        // In other words LHS should be used (directly or indirectly) by I only.
+        for (User *U : LHS->users())
+          if (U != I || !(U->hasOneUser() && *U->users().begin() == I))
+            continue;
+
         SCEVExpander Expander(*SE, *DL, "nary-reassociate");
         SmallVector<const SCEV *, 2> Ops1{ BExpr, AExpr };
         const SCEVTypes SCEVType = convertToSCEVype(m_MaxMin);
         const SCEV *R1Expr = SE->getMinMaxExpr(SCEVType, Ops1);
 
         Instruction *R1MinMax = findClosestMatchingDominator(R1Expr, I);
 
         if (!R1MinMax)

llvm/test/Transforms/NaryReassociate/nary-req.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -nary-reassociate -S | FileCheck %s
+; RUN: opt < %s -passes='nary-reassociate' -S | FileCheck %s
+
+declare i32 @llvm.smax.i32(i32 %a, i32 %b)
+
+; This is a negative test. We should not optimize if intermediate result
+; has a use outside of optimizaple pattern. In other words %smax2 has one
+; use from %smax3 and side use from %res2.
+define i32 @smax_test1(i32 %a, i32 %b, i32 %c) {
+; CHECK-LABEL: @smax_test1(
+; CHECK-NEXT:    [[C1:%.*]] = icmp sgt i32 [[A:%.*]], [[B:%.*]]
+; CHECK-NEXT:    [[SMAX1:%.*]] = select i1 [[C1]], i32 [[A]], i32 [[B]]
+; CHECK-NEXT:    [[C2:%.*]] = icmp sgt i32 [[B]], [[C:%.*]]
+; CHECK-NEXT:    [[SMAX2:%.*]] = select i1 [[C2]], i32 [[B]], i32 [[C]]
+; CHECK-NEXT:    [[C3:%.*]] = icmp sgt i32 [[SMAX2]], [[A]]
+; CHECK-NEXT:    [[SMAX3:%.*]] = select i1 [[C3]], i32 [[SMAX2]], i32 [[A]]
+; CHECK-NEXT:    [[RES:%.*]] = add i32 [[SMAX1]], [[SMAX3]]
+; CHECK-NEXT:    [[RES2:%.*]] = add i32 [[RES]], [[SMAX2]]
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %c1 = icmp sgt i32 %a, %b
+  %smax1 = select i1 %c1, i32 %a, i32 %b
+  %c2 = icmp sgt i32 %b, %c
+  %smax2 = select i1 %c2, i32 %b, i32 %c
+  %c3 = icmp sgt i32 %smax2, %a
+  %smax3 = select i1 %c3, i32 %smax2, i32 %a
+  %res = add i32 %smax1, %smax3
+  %res2 = add i32 %res, %smax2
+  ret i32 %res
+}
+