[Mesa-dev] [PATCH 2/2] i965: Avoid generating MOVs for most ir_assignment handling.

[Eric Anholt]

On Fri, 21 Oct 2011 11:05:53 -0700, Stéphane Marchesin <stephane.marchesin at gmail.com> wrote:
> 2011/10/19 Stéphane Marchesin <stephane.marchesin at gmail.com>:
> > On Sat, Aug 27, 2011 at 03:23, Kenneth Graunke <kenneth at whitecape.org> wrote:
> >> This is a port of vec4_visitor::try_rewrite_rhs_to_dst to fs_visitor.
> >>
> >> Not only is this technique less invasive and more robust, it also
> >> generates better code.  Over and above the previous technique, this
> >> reduced instruction count in shader-db:
> >>
> >> With all FS optimizations (register coalescing and such) disabled, this
> >> reduced instruction count by 6.23% on average and 36% in the best case.
> >>
> >> With optimizations left enabled, it still reduced instruction count by
> >> 0.28% on average and 1.4% in the best case.
> >>
> >> In no case did this technique result in more code than the prior method.
> >>
> >
> > Hi,
> >
> > This regresses yuv tests in glbench here. The shaders regressing are
> > the .glslf files there, and glbench can be found in the same repo:
> >
> > http://git.chromium.org/gitweb/?p=chromiumos/third_party/autotest.git;a=tree;f=client/deps/glbench/src;h=8c533f2d21e87c143f512f693947d7a660c2b4bc;hb=HEAD
> >
> 
> Should I push a revert while you guys investigate?

Cold you post the shader code from MESA_GLSL=dump?  Hopefully the shader
is trivial and we can just push a fix out.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20111021/7eb247a7/attachment.pgp>