00:13soreau: HdkR: had to call glUniform1i() to set both textures :P
00:21HdkR: soreau: Oh yes, that would be an important part.
00:23soreau: HdkR: not important enough for me to know better by now ;)
06:22dolphin: airlied: drm-intel-gt-next PR sent for v6.16, just fixes.
07:09mlankhorst: The real fun is always figuring out where to enable pushes to drm-misc-next-fixes again
08:50mlankhorst: airlied: perhaps instead of putting in memcg to be, we should put in a pointer to the current cgroup, and use it for dmem too to unify it
08:51mlankhorst: Also, we should add get_mem_cgroup_from_css() or something
11:10pq: hwentlan_, FYI, https://gitlab.freedesktop.org/drm/amd/-/issues/3358 seems to gaining a "me too" every few months.
13:01lumag: robclark, I've been trying to fix (some of) remaining issues with A702. One of the sources of Piglit failures is that Mesa doesn't have a problem to advertise high OpenGL compat levels even without ARB_geometry_shader4. I was able to find commit b272a01879d6 ("main: Allow for the possibility of GL 3.2 without ARB_geometry_shader4."). Do you remember / know a context by chance?
13:03lumag: (asking on #dri-devel in hope to get wider response)
13:28danylo: lumag I think the comment in that commit is clear enough - in order for GL 3.2 to not be advertised we should set glsl feature level to 140
13:29danylo: Because 150 adds geometry shaders
13:35lumag: danylo, hmm. And in case of GLX setting MESA_GLSL_VERSION_OVERRIDE=140 doesn't cause any difference, most likely because everything is computed on the server...
13:36lumag: danylo, I still think that the logic is a bit strange: it should be 'allow geometry if glsl > 1.50 || GL_ARB_geometry_shader4' in my opinion
13:36lumag: and require GS for GL 3.2
13:37danylo: But glsl >= 1.50 implies geometry shader
13:37danylo: Ah, I see what you mean, then maybe no one supports GL_ARB_geometry_shader4 but doesn't glsl 1.50 ?
13:38lumag: danylo, might be
13:40lumag: In a702 it's quite opposite, I think. We declare high GLSL compat, but extensions.OES_geometry_shader is false
13:42lumag: anyway, I'll try fixing the glsl_feature_level
14:00mattst88: FWIW, geometry shaders in GL 3.2 and GL_ARB_geometry_shader4 are not equivalent
14:08alyssa: mattst88: dare i ask?
14:26mattst88: alyssa: I don't remember the details off hand, but https://registry.khronos.org/OpenGL/extensions/ARB/ARB_geometry_shader4.txt has a "Dependencies on OpenGL 3.2
14:26mattst88: "
14:26mattst88: section that gives some information
14:27mattst88: I recall that we didn't think that implementing GL_ARB_geometry_shader4 was a good stepping stone towards GL3.2 geometry shaders
14:27alyssa: oh that's not bad
14:27alyssa: just gl3.2 geometry shaders with a slightly worse API
14:36karolherbst: does it allow running the gs multiple times 🙃
16:06alyssa: pro-tip: gpu's dont like it when you infinite loop them
16:06alyssa:reboots
19:12glehmann: zmike: was https://gitlab.freedesktop.org/mesa/mesa/-/tree/main/src/gallium/targets/opencl intentionally not deleted or did you make a rebase mistake with those two files?
19:50zmike: probably mistakenly not deleted