[PATCH] Global Structure Vectorization

[Renato Golin]

On 23 February 2013 17:55, Arnold Schwaighofer <aschwaighofer at apple.com>wrote:

> What is missing of course is a qualitative analysis: do any of those
> matter?
>

I had a look at those that changed in standard test-suite and the changes
seemed randomly distributed up/down/stable.

http://llvm.org/perf/db_default/v4/nts/8772

That still doesn't mean much, though, as those tests are only worth over
time, and even so, some run for so little time that they end up with a huge
deviation, so sometimes longer runs are actually faster (if you take out OS
interventions).

In any case, I wouldn't expect that big difference over current benchmarks,
since the use of global variables is limited to old C code.

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130223/2b3f3aa9/attachment.html>