06:31 <bshah> I wonder how best I can try to reduce the problem I am having with QML text elements so that it is reproducible for lima devs? I've dumped the shaders used by my test program. Although I don't know how or if I can try those shaders individually somehow and see if it is problem?

06:35 <bshah> none of shader seems to be using if/else and I don't get gpir/ppir warnings in my log when running test program.

06:44 <bshah> on 2nd look

06:45 <bshah> I found : https://gitlab.freedesktop.org/lima/mesa/issues/97

06:45 <bshah> and there indeed is user of that in my shaders,

06:45 <bshah> "uniform sampler2D qt_Texture;" etc

07:53 <yuq825> sampler2D is not rect texture, it's supported

07:54 <bshah> ah

07:54 <bshah> hm, back to 0 I guess :/

15:00 <anarsoul> bshah: you may be hitting max vertices per draw (65536)

15:09 <bshah> anarsoul: any way to verify this?

15:09 <anarsoul> check your code? :)

15:10 <enunes> anarsoul: we should probably include a check to abort the application if we hit that, like we do for 512 instructions in gpir

15:11 <enunes> not sure where :)

15:11 <anarsoul> enunes: or better ressurect https://gitlab.freedesktop.org/mesa/mesa/merge_requests/768

15:13 <enunes> yeah or that if it can be done in a timely manner, looking at it I don't know if it's close to a solution or not

15:15 <anarsoul> it's abandoned

15:16 <anarsoul> it should be easy to ressurect it but I'm in the middle of rebasing my pp cf patches

15:17 <enunes> maybe we can start pinging there and asking if the owner still has interest in it or not

15:19 <anarsoul> feel free to try :)

15:20 <enunes> ok I'll add a comment

15:20 <anarsoul> btw we also need to add support for heap bo into kernel driver, something similar to https://patchwork.kernel.org/patch/11048159/

15:20 <anarsoul> and use it for gp heap

15:21 <anarsoul> GP can signal when it's out of memory and resume when more memory is provided

15:21 <anarsoul> we should use that instead of statically allocating 2mb for each context

15:22 <bshah> although I am kinda starting to wonder if probably Qt is at blame. If I am understanding this bug correctly, I would see that bits and pieces are not consistently rendered, some are rendered properly but not all. But here what I see is all keys are rendered perfectly just not all of text.

15:22 <bshah> (ignore my ramblings if it doesn't make sense :P

15:22 <bshah> )

15:28 <enunes> anarsoul: seems good, that falls into "optimixations we can do after we have a functional driver"?

15:30 <anarsoul> yeah, but assuming it takes 6 months to get your change into mainline linux it would be nice to start early :)

15:31 <anarsoul> also it'll break ABI

15:45 <wens> pp cf!

16:40 <anarsoul> don't get too excited