Optimize load from aggregate stores

David Blaikie dblaikie at gmail.com
Sun Mar 30 10:23:11 PDT 2014

On Sun, Mar 30, 2014 at 1:18 AM, deadal nix <deadalnix at gmail.com> wrote:
> Hi again,
> What should I do to get this even considered ? I got very clowny codegen in
> many cases without this patch.

General advice for patch submission is ping the  thread every week or so.

My specific advice would be that if this is a general optimization
issue (not one that only shows up with a custom (non-Clang) frontend,
etc), especially if it manifests with optimizations enabled (-O2/-O3),
that providing some specific examples to highlight the value of this
patch. Show some reasonable piece of source code (if you can use a
real world, open source, important/hot code (compression algorithms,
etc), even better) and show the IR before and after your change to
show the value/importance of the patch, that might increase
visibility/interest in reviewing/committing your patch.

> I allow myself to forward my previous message. I would also advice for a
> better system to submit patches (use github, mount a phabricator somewhere,

We have one of those: http://llvm.org/docs/Phabricator.html

> option aren't limited).
> ---------- Forwarded message ----------
> From: deadal nix <deadalnix at gmail.com>
> Date: 2014-03-22 22:40 GMT-07:00
> Subject: Optimize load from aggregate stores
> To: llvm-commits at cs.uiuc.edu
> I can't find the thread where this was discussed. I made a diff some time
> ago. Here is the same updated and with improved test coverage.
> Currently, LLVM really misoptimize aggregate load/store. This diff is a big
> improvement on that side.
> Amaury SECHET
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits

More information about the llvm-commits mailing list