[NARY] Don't optimize min/max if there are side uses (part2)

[ebrevnov]

Summary

Previous attempt to fix infinite recursion in min/max reassociation was not fully successful (D100170). Newly discovered failing case is due to not properly handled when there is a single use. It should be processed separately from 2 uses case.

[ebrevnov]

Renaming

[mkazantsev]

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

617

How is this even possible? I could not find explaining comment here, but intuitively LHS and RHS are expected to be arguments of I. If LHS has the only user then it must be I. no?

Could you please add a method comment about what I, LHS and RHS actually are?

[ebrevnov]

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

617

Yes, your understanding is correct. LHS & RHS simply left and right operands of I (which is min/max or select). I will updated the code shortly

[ebrevnov]

Added comment + removed useless check + replaced explicit for with llvm::any_of

[mkazantsev]

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

615–626

Please limit capture scope to &I (it doesn'tseem you need anythong else).

616

Is U != I even happens to be false? An instruction can only use itself it it's a Phi, and I believe I is not a Phi here.

[mkazantsev]

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

616

Ah it's a user of LHS and not I. Makes sense then.

[mkazantsev]

Looks good now, thanks.

llvm/lib/Transforms/Scalar/NaryReassociate.cpp

if (!LHS->hasNUsesOrMore(3) && match(LHS, m_MaxMin)) {
             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.
+        bool HasSideUse = false;
+        if (LHS->hasOneUser()) {
+          if (LHS->user_back() != I)
+            HasSideUse = true;
+        } else {
           for (User *U : LHS->users())
-          if (U != I || !(U->hasOneUser() && *U->users().begin() == I))
-            continue;
+            if (U != I && !(U->hasOneUser() && *U->users().begin() == I))
+              HasSideUse = true;
+        }
+
+        if (HasSideUse)
 
         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);
 

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)
+declare i64 @llvm.umin.i64(i64, i64) 
 
 ; 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
+; has a use outside of optimizable 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
 }
 
+; This is a negative test. It similar to the previous one
+; but a bit more complex. In particular after first iteration
+; e10 is replaced with %e10.nary = call i64 @llvm.umin.i64(i64 %e5, i64 %e).
+; No more reassociation should be applied to %e10.nary since
+; %e5 has side use in %e6.
+define void @test2(i64 %arg) {
+; CHECK-LABEL: @test2(
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[E:%.*]] = sub i64 undef, 0
+; CHECK-NEXT:    [[E1:%.*]] = sub i64 [[ARG:%.*]], 0
+; CHECK-NEXT:    [[E2:%.*]] = call i64 @llvm.umin.i64(i64 [[E]], i64 [[E1]])
+; CHECK-NEXT:    [[E3:%.*]] = call i64 @llvm.umin.i64(i64 [[E2]], i64 16384)
+; CHECK-NEXT:    [[E4:%.*]] = sub i64 [[ARG]], 0
+; CHECK-NEXT:    [[E5:%.*]] = call i64 @llvm.umin.i64(i64 [[E4]], i64 16384)
+; CHECK-NEXT:    [[E6:%.*]] = icmp ugt i64 [[E5]], 0
+; CHECK-NEXT:    [[E10_NARY:%.*]] = call i64 @llvm.umin.i64(i64 [[E5]], i64 [[E]])
+; CHECK-NEXT:    unreachable
+;
+bb:
+  %e = sub i64 undef, 0
+  %e1 = sub i64 %arg, 0
+  %e2 = call i64 @llvm.umin.i64(i64 %e, i64 %e1)
+  %e3 = call i64 @llvm.umin.i64(i64 %e2, i64 16384)
+  %e4 = sub i64 %arg, 0
+  %e5 = call i64 @llvm.umin.i64(i64 %e4, i64 16384)
+  %e6 = icmp ugt i64 %e5, 0
+  %e7 = sub i64 undef, 0
+  %e8 = sub i64 %arg, 0
+  %e9 = call i64 @llvm.umin.i64(i64 %e7, i64 %e8)
+  %e10 = call i64 @llvm.umin.i64(i64 %e9, i64 16384)
+  unreachable
+}
+