12:37 <bbrezillon> stepri01: I've been reviewing tomeu's vkEvent MR, and it appeared to me that supporting 'synchronization' jobs (jobs with job->head = 0 whose only purpose is to add a synchronization point) would be a good replacement for the 'NULL' job emitted here

12:37 <bbrezillon> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/11709/diffs?commit_id=d7cef0c41c868d7dc1045f5c500fd119c7fee080#f3c061ff55836e648459396301a7e693316150ca_779_789

12:37 <bbrezillon> stepri01: IOW, would you be okay with this change => https://gitlab.freedesktop.org/bbrezillon/linux/-/commit/30132ab93890d1693c1fc6fdc391e412cda11720 ?

12:40 <bbrezillon> we'd keep the 'if (submit->jc != 0) return -EINVAL' in the old submit ioctl of course

12:40 <stepri01> bbrezillon: I'm not sure whether special casing head==0 is best, or introducing a new JD_REQ

12:40 <stepri01> kbase has the concept of "dependency only atoms" which is implemented with a special 'requirement'

12:40 <bbrezillon> sure, a JD_REQ flag would work too

12:40 <stepri01> that would be my preference

12:41 <bbrezillon> perfect, I'll go for this solution then

12:42 <stepri01> 👍

15:03 <alyssa> robmur01: ping

16:30 <alyssa> alyssa: util_blitter_save_so_targets call is probably wrong

16:32 <alyssa> so is save_vertex_shader

16:43 <alyssa> do we need to preserve min_samples?

17:15 <robmur01> alyssa: 'sup

17:17 <alyssa> robmur01: who do i need to hassle to get sven's DART patches pushed 😇

17:22 <robmur01> I'll try to give them a second look over soon, although where we are in the cycle it's probably at least a couple of weeks still until new stuff starts getting queued again

17:31 <robmur01> I can't remember off-hand if Joerg has a particular preference, but it's not uncommon for maintainers to focus just on fixes early on and not start their for-next queue until around rc3

17:32 <robmur01> certainly don't use drm-misc as a benchmark for the rest of the kernel ;)