[PATCH] [DependenceAnalysis] Extend unifySubscriptType for handling coupled subscript groups.

Vaivaswatha Nagaraj vn at compilertree.com
Tue May 12 21:29:30 PDT 2015


>I may be missing something. Why not calling `unifySubscriptType(Pairs[SJ])`
here? Do we have to unify the subscript types across the entire group?
Thats right. All subscripts across the entire group need to have the same
type. This is because when testing coupled groups, the information obtained
by testing one subscript (constraints) may be propagated to another
subscript in the group. These constraints are used when testing the other
subscript, and at this point, different SCEVs may be created, all of which
need to have the same operand type.

So in short, its not just within a pair, but also across pairs in a coupled
group that the type must match. Hence the extension of unifySubscriptType().

  - Vaivaswatha

On Wed, May 13, 2015 at 5:34 AM, Jingyue Wu <jingyue at google.com> wrote:

> REPOSITORY
>   rL LLVM
>
> ================
> Comment at: lib/Analysis/DependenceAnalysis.cpp:3599
> @@ -3564,2 +3598,3 @@
>            Mivs.set(SJ);
> +        PairsInGroup.push_back(&Pair[SJ]);
>        }
> ----------------
> I may be missing something. Why not calling
> `unifySubscriptType(Pairs[SJ])` here? Do we have to unify the subscript
> types across the entire group?
>
> http://reviews.llvm.org/D9698
>
> EMAIL PREFERENCES
>   http://reviews.llvm.org/settings/panel/emailpreferences/
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150513/7e9d6023/attachment.html>


More information about the llvm-commits mailing list