[llvm] r371139 - [AliasSetTracker] Correct AAInfo check.

[Finkel, Hal J. via llvm-commits]

On 9/5/19 6:00 PM, Alina Sbirlea via llvm-commits wrote:
> Author: asbirlea
> Date: Thu Sep  5 16:00:36 2019
> New Revision: 371139
>
> URL: http://llvm.org/viewvc/llvm-project?rev=371139&view=rev
> Log:
> [AliasSetTracker] Correct AAInfo check.
>
> Properly check if NewAAInfo conflicts with AAInfo.
> Update local variable and alias set that a change occured when a conflict is found.
> Resolves PR42969.
>
> Added:
>      llvm/trunk/test/Transforms/LICM/pr42969.ll
> Modified:
>      llvm/trunk/include/llvm/Analysis/AliasSetTracker.h
>
> Modified: llvm/trunk/include/llvm/Analysis/AliasSetTracker.h
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/AliasSetTracker.h?rev=371139&r1=371138&r2=371139&view=diff
> ==============================================================================
> --- llvm/trunk/include/llvm/Analysis/AliasSetTracker.h (original)
> +++ llvm/trunk/include/llvm/Analysis/AliasSetTracker.h Thu Sep  5 16:00:36 2019
> @@ -87,10 +87,11 @@ class AliasSet : public ilist_node<Alias
>           AAInfo = NewAAInfo;
>         else {
>           AAMDNodes Intersection(AAInfo.intersect(NewAAInfo));
> -        if (!Intersection) {
> +        if (!Intersection.TBAA || !Intersection.Scope ||
> +            !Intersection.NoAlias) {


Thanks for tracking this down!

Is now too strong? What if both AAInfo and NewAAInfo started out with 
one of these as nullptr?

  -Hal


>             // NewAAInfo conflicts with AAInfo.
>             AAInfo = DenseMapInfo<AAMDNodes>::getTombstoneKey();
> -          return SizeChanged;
> +          SizeChanged = true;
>           }
>           AAInfo = Intersection;
>         }
>
> Added: llvm/trunk/test/Transforms/LICM/pr42969.ll
> URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LICM/pr42969.ll?rev=371139&view=auto
> ==============================================================================
> --- llvm/trunk/test/Transforms/LICM/pr42969.ll (added)
> +++ llvm/trunk/test/Transforms/LICM/pr42969.ll Thu Sep  5 16:00:36 2019
> @@ -0,0 +1,71 @@
> +; RUN: opt %s -S -scoped-noalias -enable-mssa-loop-dependency=true -licm -debug-only=licm | FileCheck %s
> +
> +define i16 @main(i1 %a_b_mayalias, i16* %a, i16* %b) {
> +; CHECK-NOT: LICM: Promoting value stored to in loop
> +; CHECK:       scalar.body:
> +; CHECK-NEXT:    [[J:%.*]] = phi i64
> +; CHECK-NEXT:    [[TMP3:%.*]] = load i16
> +; CHECK-NEXT:    [[RESULT:%.*]] = add i16 [[TMP3]], 1
> +; CHECK-NEXT:    store i16 [[RESULT]]
> +
> +entry:
> +  br label %outer
> +
> +outer:                                            ; preds = %scalar.cleanup, %entry
> +; 4 = MemoryPhi({entry,liveOnEntry},{scalar.cleanup,2})
> +  %i = phi i16 [ 0, %entry ], [ %i.next, %scalar.cleanup ]
> +  br i1 %a_b_mayalias, label %scalar.ph, label %vector.ph
> +
> +vector.ph:                                        ; preds = %outer
> +; MemoryUse(4) MayAlias
> +  %tmp1 = load i16, i16* %a, align 1, !alias.scope !0, !tbaa !7
> +  %tmp2 = add i16 %tmp1, 1
> +; 1 = MemoryDef(4)
> +  store i16 %tmp2, i16* %b, align 1, !alias.scope !3, !noalias !0, !tbaa !7
> +  br label %vector.body
> +
> +vector.body:                                      ; preds = %vector.body, %vector.ph
> +  %index = phi i64 [ %index.next, %vector.body ], [ 0, %vector.ph ]
> +  %index.next = add i64 %index, 1
> +  %cmp1 = icmp eq i64 %index.next, 16
> +  br i1 %cmp1, label %middle.block, label %vector.body
> +
> +middle.block:                                     ; preds = %vector.body
> +  br label %scalar.ph
> +
> +scalar.ph:                                        ; preds = %middle.block, %outer
> +; 5 = MemoryPhi({outer,4},{middle.block,1})
> +  %j.start = phi i64 [ 0, %outer ], [ 16, %middle.block ]
> +  br label %scalar.body
> +
> +scalar.body:                                      ; preds = %scalar.body, %scalar.ph
> +; 3 = MemoryPhi({scalar.ph,5},{scalar.body,2})
> +  %j = phi i64 [ %j.next, %scalar.body ], [ %j.start, %scalar.ph ]
> +; MemoryUse(3) MayAlias
> +  %tmp3 = load i16, i16* %a, align 1, !tbaa !7
> +  %result = add i16 %tmp3, 1
> +; 2 = MemoryDef(3)
> +  store i16 %result, i16* %b, align 1, !tbaa !7
> +  %j.next = add nuw nsw i64 %j, 1
> +  %cmp2 = icmp ult i64 %j.next, 20
> +  br i1 %cmp2, label %scalar.body, label %scalar.cleanup
> +
> +scalar.cleanup:                                   ; preds = %scalar.body
> +  %result.lcssa = phi i16 [ %result, %scalar.body ]
> +  %i.next = add nuw nsw i16 %i, 1
> +  %exitcond = icmp eq i16 %i.next, 10
> +  br i1 %exitcond, label %exit.block, label %outer
> +
> +exit.block:                                       ; preds = %scalar.cleanup
> +  %result.lcssa.lcssa = phi i16 [ %result.lcssa, %scalar.cleanup ]
> +  ret i16 %result.lcssa.lcssa
> +}
> +
> +!0 = !{!1}
> +!1 = distinct !{!1, !2}
> +!2 = distinct !{!2, !"LVerDomain"}
> +!3 = !{!4}
> +!4 = distinct !{!4, !2}
> +!5 = !{!"Simple C++ TBAA"}
> +!6 = !{!"omnipotent char", !5, i64 0}
> +!7 = !{!6, !6, i64 0}
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at lists.llvm.org
> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits

-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory