ChanServ changed the topic of #zink to: official development channel for the mesa3d zink driver ||
sinanmohd has joined #zink
sinanmoh- has quit [Ping timeout: 480 seconds]
sinanmohd has quit [Ping timeout: 480 seconds]
sinanmohd has joined #zink
hch12907 has quit [Quit: Bridge terminating on SIGTERM]
MTCoster has quit [Remote host closed the connection]
MTCoster has joined #zink
hch12907 has joined #zink
kusma has joined #zink
neobrain[m] has joined #zink
pac85[m] has joined #zink
Soroush has joined #zink
<zmike> Soroush: is IMG able to implement robustBufferAccess2 ?
<zmike> anholt: would be nice to get in before bp, though I'll probably backport it anyway if it doesn't
<zmike> also how is vvl on anv going?
<zmike> ah neat
<anholt> I think gl46 cts is actually inefficient enough that we'd probably be better off testing gles + some subset of piglit pre-merge and kicking gl cts and full piglit to post.
<anholt> (in terms of catching useful bugs)
<zmike> hmm not sure I'm okay with that; there's a lot of cases where 46 has caught bugs
<zmike> GL WG is looking to spend some resources on cts in the near future, so maybe we should try to pinpoint the long tests
<anholt> yeah. just fixing the tests would be great.
<zmike> another factor for cts runtimes is actually just the initial cts binary execution, which is being tracked
<anholt> initial cts execution seems quite fast here
<zmike> err wrong link
<anholt> (glcts, that is. vk is atrocious)
<zmike> really? it's incredibly slow for me on glcts too
<zmike> this ticket is mainly about vk, but I assume the same change would apply to gl
<zmike> for piglit I know there's one blend test that takes a very long time on sw drivers
<anholt> 0-test caselist takes .1s here.
<zmike> huh
<zmike> I've been meaning to try splitting up the blend test into multiple cases
<zmike> since it's 60+s each time
<zmike> are you looking to merge the anv vvl thing asap or do I have some time to analyze?
<anholt> I would love to merge as is and fix xfails from there
<zmike> k
<anholt> because otherwise it'll be extra work days of rebasing and rebuilding containers
<anholt> piglit gl-1.0-blend needs to do many draws then many probes, rather than draw probe draw probe draw probe.
<zmike> yeah that's the one
<zmike> ahh you're bumping validation version
<zmike> hm
<zmike> the one thing I'll say about that is this is a very bad time to bump
<zmike> there's known regressions they're working on which will make VVL explode any time GPL is used with zink
<zmike> which is going to be problematic in the near future
<anholt> current release? or tot?
<zmike> unsure, I didn't bisect to see how far back the breakage went
<zmike> (I only tested on tot)
<zmike> anholt: also re: did you mean to mark this as active? it's WIP now so nobody will see it
<anholt> thanks for noticing. I actually have no idea how khronos process works
<anholt> (other than "poorly"
<zmike> cc Sachiel
<zmike> I think I usually just click mark as active and then add the opengl reviewers
<zmike> can't remember if I need something else
<zmike> (click "add reviewer" on the left, type in "openglcts-review" and send)
<anholt> thanks!
<Sachiel> yeah, you need to mark them active at least. The reviewers can get added by the cts chair if you don't feel like it/forget to do it. For extra points, add the 'opengl-review' tag
hch12907 has quit []
neobrain[m] has quit []
jhli_ has quit []
jhli has joined #zink
kchibisov has joined #zink
<kchibisov> Is the stuff like eglSwapBuffersWithDamage in scope for zink, given that it's basically a wsi thing?
<zmike> probably