IRC Logs of #dri-devel on for 2014-04-18

Choose date

02:14 #dri-devel: < ari-_-e> the libOpenVG that you get when you build from master right now does not export any of its symbols
06:50 #dri-devel: < cand> is there a minimum required mmio size?
06:51 #dri-devel: < cand> ie, does some spec require it to be at least X bytes?
06:52 #dri-devel: < daniels> ari-_-e: do you think anyone will notice?
07:18 #dri-devel: < imirkin> ari-_-e: i'm guessing it's due to some of xexaxo^'s build changes...
07:40 #dri-devel: < xexaxo> ari-_-e: iirc there was a bugreport on the topic, not sure if it's yours or not. Either way a git bisect would be great
07:56 #dri-devel: < imirkin> is piglit building broken for anyone else with commit 6b71f21705 ("util/egl: Add piglit_egl_get_default_display()") due to unknown EGL_PLATFORM_X11_EXT/etc?
07:57 #dri-devel: < dcbaker> imirkin: It builds on i965
07:58 #dri-devel: < imirkin> dcbaker: i think it's a headers issue...
08:01 #dri-devel: < imirkin> like it expecting that i have the very-latest headers installed or something. which i don't think should have to be the case for piglit... but i guess i'm not sure where those defines are supposed to come from in the first place
08:02 #dri-devel: < dcbaker> I am building against mesa master
08:03 #dri-devel: < imirkin> i'm building against a released mesa (10.0.4 iirc)
08:04 #dri-devel: < imirkin> it does look like those are defined in eglext.h (in mesa git)
08:04 #dri-devel: < imirkin> but not on my system one
08:05 #dri-devel: < imirkin> i guess i'll send chad an email and see what the right fix is, or if everyone should be forced to have the latest headers to build piglit
08:05 #dri-devel: < dcbaker> I dont think the latter is an option
08:05 #dri-devel: < imirkin> it might be, as long as you don't have to have the latest libraries when running... dunno
08:07 #dri-devel: < dcbaker> That would make stable branches interesting
08:07 #dri-devel: < imirkin> well, i almost always build and run piglit against different versions :)
08:07 #dri-devel: < imirkin> i build against system
08:07 #dri-devel: < imirkin> and run against my hacked-up copies
08:18 #dri-devel: < daniels> imirkin: those defines are from a new-ish egl extension which allows you to actually correctly type the arguments to eglGetDisplay etc
08:19 #dri-devel: < daniels> it would be nice if that weren't a hard requirement, given that a lot of proprietary stacks don't support that yet
08:19 #dri-devel: < daniels> and piglit is a great tool to use to shame them into fixing stuff :)
08:19 #dri-devel: < imirkin> daniels: yeah, like how 50% of tests fail on the blob (for totally silly reasons)? that's definitely made them improve things.... not =/
08:20 #dri-devel: < imirkin> er, make that "nvidia proprietary driver"
08:20 #dri-devel: < daniels> imirkin: heh, for better or worse (though i'm pretty sure i know which), i deal with a whole different set of proprietary stacks
08:22 #dri-devel: < imirkin> yeah, there are a lot of blobs out there, i tend to forget that
09:26 #dri-devel: < mareko> cand: regarding the perf. regression on cayman, what resolution did you test with?
09:28 #dri-devel: < cand> mareko, 1600x900
09:28 #dri-devel: < mareko> cand: on aruba?
09:28 #dri-devel: < cand> yes
09:28 #dri-devel: < cand> Richland A8-6500
09:30 #dri-devel: < cand> it doesn't affect all ioq3 games: padman, warsow, urban terror were unaffected
10:16 #dri-devel: < EdB> hello tstellar, any hint for someone who want to start developing on clover ?
10:34 #dri-devel: < mannerov> I get almost 90% of the performance of catalyst with my radeonsi card on unigine heaven. I'm curious to know what's the remaining 10%
10:46 #dri-devel: < agd5f> mannerov, you enable hyperz?
10:47 #dri-devel: < mannerov> is it possible to enable it for radeonsi?
10:47 #dri-devel: < agd5f> mannerov, yes. it's off by default however
10:48 #dri-devel: < mannerov> what should I do
10:49 #dri-devel: < agd5f> RADEON_DEBUG=hyperz
10:55 #dri-devel: < mannerov> I have a 2% performance improvement with unigine heaven
10:58 #dri-devel: < mannerov> if I'm correct only a subset of hyperz features is supported for the moment
11:00 #dri-devel: < agd5f> mannerov, I don't think it's the only factor

Written by Christoph Brill © 2007-2014

Valid XHTML 1.0 Transitional

Available in Android Market