=== chihchun is now known as chihchun_afk | ||
=== chihchun_afk is now known as chihchun | ||
=== chihchun is now known as chihchun_afk | ||
duflu | greyback: I think I love you right now. INTEL_DEBUG=perf was a great suggestion | 09:13 |
---|---|---|
duflu | (love you more than usual) | 09:13 |
greyback | :) | 09:13 |
greyback | duflu: I bestow http://mesa3d.org/envvars.html upon you, use it with care! | 09:14 |
duflu | greyback: I know. Still never used that variable before | 09:14 |
greyback | I had to read through the i915 mesa code to figure out what env vars actually returned some debugging info | 09:14 |
duflu | Arguably perf issues are errors to us... but if you ask only for errors you'd never see it | 09:14 |
greyback | what did you learn? | 09:15 |
=== chihchun_afk is now known as chihchun | ||
duflu | greyback: Old and new(!) Intel GPUs fall back to software rendering regularly | 09:17 |
duflu | But Atoms moreso | 09:17 |
duflu | Possibly caused by ES/non-ES confusion | 09:18 |
duflu | The Atom issue seems to be that our shaders are 18% longer than the hardware limit | 09:18 |
greyback | that's interesting | 09:21 |
duflu | It's a big deal | 09:21 |
duflu | And exciting to imagine after we fix it | 09:21 |
greyback | indeed | 09:21 |
duflu | greyback: The offending function seems to only be in qtubuntu and qtquick but maybe I missed more | 09:23 |
greyback | qtubuntu is just a glue between mirclient/platform-api and Qt - aside from setting up GL context, it has no other impact on GL | 09:24 |
greyback | qtquick most likely | 09:24 |
duflu | I logged bugs. You will find them... | 09:24 |
greyback | nice | 09:24 |
greyback | duflu: fyi, here are the shaders QtQuick has: http://code.qt.io/cgit/qt/qtdeclarative.git/tree/src/quick/scenegraph/shaders | 09:26 |
greyback | that is trunk (5.7 pre-release), we're using 5.5, but I don't see the shaders changing in some time | 09:26 |
duflu | greyback: One thing at a time. The first issue seems to be bad format args to glCopyTexSubImage | 09:27 |
greyback | ah, that was what you were referring to | 09:27 |
greyback | yes | 09:27 |
greyback | I had hoped a proper OpenGL context would have resolved that | 09:27 |
duflu | Which I found one instance of that switches behaviour for ES | 09:27 |
duflu | Yep | 09:27 |
greyback | http://pastebin.ubuntu.com/16518134/ <- qt only uses it for the font distance field | 09:36 |
greyback | uitk doesn't | 09:36 |
=== hikiko is now known as hikiko|ln | ||
=== chihchun is now known as chihchun_afk | ||
=== hikiko|ln is now known as hikiko | ||
=== dandrader is now known as dandrader|afk | ||
=== dandrader|afk is now known as dandrader | ||
=== dandrader is now known as dandrader|afk | ||
alan_g | kdub: you're probably not looking at it yet. But you had an MP that got the titlebars wrong. I've just seen a similar behavior in code I'm working on and it looks like a deeper issue that may also have caused the problem on your branch. Will update if I figure it out. | 15:42 |
kdub | alan_g, ack, yeah got stuck on that one because i couldnt get the terminal to start | 16:11 |
alan_g | kdub: I bottomed my problem - unfortunately, it was new code so it fix yours. | 16:38 |
kdub | *it doesnt fix? | 16:38 |
alan_g | yes doesn't | 16:41 |
alan_g | /sigh | 16:41 |
alan_g | Oh well, 20min of week left and code is in a stable intermediate form. Might as well MP it and start looking forward to xp2016 | 16:43 |
=== alan_g is now known as alan_g|EOW | ||
=== dandrader|afk is now known as dandrader | ||
=== dandrader is now known as dandrader|afk | ||
=== dandrader|afk is now known as dandrader |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!