[Mesa-stable] Release-candidate branch for upcoming 10.2.4

[Kenneth Graunke]

On Monday, July 14, 2014 09:49:48 PM Carl Worth wrote:
> Hi folks,
> 
> I've pushed out an update to the 10.2 branch and I need some specific
> testing in the next three days.
> 
> I've tested the branch on Intel (Haswell) as well as both swrast and
> Gallium softpipe and found no piglit regressions compared to the 10.2.3
> release.
> 
> The branch includes a few patches to nouveau and radeonsi which I have
> not been able to test. If someone will test one of these drivers with
> piglit and let me know that all looks good, I'll be happy to include the
> patches in the release. Otherwise, I'll drop any untested patches before
> making the final release on Friday.
> 
> Also, there's still time in the next three days for someone to nominate
> further driver-specific changes. I'll just need positive piglit test
> results for any such patches, (on top of the branch as it stands now),
> before I'll accept them.
> 
> Thanks, all.
> 
> -Carl

Hi Carl,

Here's a last minute pull request targeted at improving Broadwell.  It's pretty big, and a couple of the patches just landed, so this might be too much to ask.

Branch:
http://cgit.freedesktop.org/~kwg/mesa/log/?h=10.2.4-broadwell-pull

Piglit result files and summary:
http://people.freedesktop.org/~kwg/10.2.4-broadwell-pull/

Patches:

i965/fs: Refactor check for potential copy propagated instructions.
i965/fs: skip copy-propate for logical instructions with negated src entries
i965/vec4: skip copy-propate for logical instructions with negated src entries
i965: Don't copy propagate abs into Broadwell logic instructions.

These fix 541 Piglit tests.  All but the first should only affect Gen8+.

i965/vec4: Don't return void from a void function.
i965/vec4: Don't fix_math_operand() on Gen >= 8.
i965/fs: Don't fix_math_operand() on Gen >= 8.
i965/fs: Make try_constant_propagate() static.
i965/fs: Constant propagate into 2-src math instructions on Gen8.
i965/vec4: Constant propagate into 2-src math instructions on Gen8.

These are small optimizations that only affect Gen8+.

i965: Set execution size to 8 for instructions with force_sechalf set.

Broadwell only.  Fixes bad assembly in MSAA tests.

i965/fs: Set force_uncompressed and force_sechalf on samplepos setup.
i965/fs: Don't use brw_imm_* unnecessarily.
   (only necessary for the next patch to pick cleanly)
i965/fs: Use WE_all for gl_SampleID header register munging.
i965: Generalize the pixel_x/y workaround for all UW types.

These fix MSAA issues on Broadwell, and possibly on other platforms.
I can't point to specific tests fixed, but it fixes some issues you can
see by inspecting the assembly.

Without them, I saw regressions from the later patches.  With them,
everything kept working great.  I've tested them quite a bit, so I feel
pretty confident in them.

Plus, multisampling isn't 100% solid on Broadwell at the moment, so as long
as it doesn't hurt other platforms, I'm not too concerned.

i965/fs: Set correct number of regs_written for MCS fetches.

Fixes a bug.  Could help all platforms.

i965: Add auxiliary surface field #defines for Broadwell.
i965: Add plumbing for Broadwell's auxiliary surface support.
i965: Drop SINT workaround for CMS layout on Broadwell.
i965: Hook up the MCS buffers in SURFACE_STATE on Broadwell.
i965: Add 2x MSAA support to the MCS allocation function.
i965: Enable compressed multisample support (CMS) on Broadwell.

These enable an optimization that's important for multisampled programs,
such as anything using WebGL.  With the earlier fixes, they don't appear
to break anything, and have been on master for a while now.

Thanks!  Let me know what you think, and if I can answer any questions.

--Ken
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/mesa-stable/attachments/20140715/aa964eb9/attachment.sig>