[LLVMdev] Vectorizing global struct pointers

[Renato Golin]

On 5 February 2013 16:38, Arnold Schwaighofer <aschwaighofer at apple.com>wrote:

> If I understand you correctly, conceptually you want two different objects
> to be returned for Foo.bl and Foo.al?
>

Not necessarily. The vectorizer is implemented expecting that the objects
will be different, but that's a limitation on the vectorizer itself.

However, changing the vectorizer to recognize GEP offsets is probably not
the best course of action, but by the time we get the object back, we have
no information if it was a GEP, and if so, what was its offset.

Maybe I could work back (via uses) and identify the GEP and store the
offset with the object locally, so that the list would not think different
members are the same object. But I wanted to know first if the underlying
object concept is correct. From the docs you sent me, it seems it is.

TBAA seems a bit too much for this case, though.

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20130205/b46eb697/attachment.html>