raster has quit [Remote host closed the connection]
vstehle has quit [Ping timeout: 240 seconds]
marcodiego has quit [Quit: Leaving]
davidlt has joined #panfrost
vstehle has joined #panfrost
stikonas has joined #panfrost
<tomeu>
narmstrong: what do you think about keeping a single build job per arch, but containing as many lava job descriptions as devices it can run on?
<tomeu>
it will be specially important when we start running armhf jobs on arm64-capable machines
<narmstrong>
tomeu: sure no problem
<tomeu>
cool
<tomeu>
narmstrong: are the patches for t820 going to be in 5.4?
<narmstrong>
tomeu: will daecon said so, but his -rc1 PR didn't have the patches
<tomeu>
narmstrong: didn't he say that it would be in his first fixes PR?
<tomeu>
btw, it's better to update test expectations along with the code changes that caused the update
<tomeu>
so they are more easily kept in sync
<bbrezillon>
tomeu: duh, there's been a problem
<bbrezillon>
the title has been dropped
<bbrezillon>
it was supposed to be
<bbrezillon>
"
<bbrezillon>
panfrost: More tests are passing
<bbrezillon>
<bbrezillon>
Remove the tests that are now passing.
<bbrezillon>
<bbrezillon>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
<bbrezillon>
"
<bbrezillon>
oops, no, it's my fault
<bbrezillon>
sorry about that
<bbrezillon>
and I'll squash things in the commit fixing the tests next time
<tomeu>
cool!
yann has quit [Ping timeout: 268 seconds]
megi has joined #panfrost
<narmstrong>
tomeu: seems I hit a drm-sched issue on v5.2
<narmstrong>
tomeu: it goes well on v5.2
<narmstrong>
*dsrm-sched issue on v5.3
davidlt has quit [Ping timeout: 245 seconds]
<alyssa>
So I've been working on getting the no-scheduler scheduler to zero regressions
<alyssa>
Since it's a big infra change and so doing a totally dumby scheduling algorithm atop the core infrastructure is probably the easiest way to validate the infra first, and then we can do the scheduling after with the new infra
<narmstrong>
tomeu: bisecting it, over lava :-) very handy !
<tomeu>
indeed!
* tomeu
wonders how many schedulers we have, and why they always seem to break everything