[Mesa-dev] [PATCH 2/2] i965: Add an implementation of intel_miptree_map using streaming loads.

[Chad Versace]

On 11/07/2013 10:55 AM, Matt Turner wrote:
> Improves performance of RoboHornet's 2D Canvas toDataURL benchmark
> [http://www.robohornet.org/#e=canvastodataurl] by approximately 5x
> on Baytrail on ChromiumOS.
> ---
> v2: Added benchmark results
>      Remove unnecessary _mesa_align_free in error case
>      Rename stride to width_bytes and make const
>      Remove mt->region->cpp from malloc size calculation
>      Remove misalignment from malloc size calculation
>
> Chad: I think you're right about not needing misalignment in the malloc
> size calculation, but if that's incorrect it could mean that the src/dst
> buffers aren't coaligned and we'll get no performance improvement. Can
> you double check that this patch still gives the expected improvement?

Elapsed time drops by -81.4861% +/- 1.22619% (n=3 s=14.9105, confidence=95%).

This patch has my r-b.