alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
fysa has joined #panfrost
fysa has quit [Ping timeout: 265 seconds]
raster has quit [Quit: Gettin' stinky!]
megi has quit [Ping timeout: 240 seconds]
fysa has joined #panfrost
fysa has quit [Ping timeout: 240 seconds]
fysa has joined #panfrost
BenG83 has quit [Ping timeout: 250 seconds]
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #panfrost
davidlt has joined #panfrost
fysa has quit [Remote host closed the connection]
<tomeu> alyssa: amazing about the blend tests!
<tomeu> I only see commits in the compiler, that was it?
<tomeu> narmstrong: that skips file is a good start
fysa has joined #panfrost
fysa has quit [Ping timeout: 265 seconds]
Elpaulo has joined #panfrost
stikonas has joined #panfrost
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #panfrost
stikonas has quit [Ping timeout: 276 seconds]
warpme_ has joined #panfrost
BenG83 has joined #panfrost
megi has joined #panfrost
fysa has joined #panfrost
fysa has quit [Ping timeout: 240 seconds]
raster has joined #panfrost
<alyssa> tomeu: Yeah, not sure what the problem is
<alyssa> If you were *that* curous you could bisect across the branch I merged but shrug
fysa has joined #panfrost
fysa has quit [Ping timeout: 246 seconds]
<bbrezillon> alyssa: I reverted "panfrost: Disable frame throttling" to try and reproduce the FD leak issue, but glmark doesn't seem to crash here
<robmur01> yay, pagetable fixes finally made rc4!
<robmur01> bbrezillon: I think the underlying leak was fixed in the meantime
<bbrezillon> robmur01: ok, great! so it was something in the core
<TheCycoONE> is mesa git working on kevin again? 4c9b9e from this weekend core dumped launching sddm in 'ra_get_node_class'
<TheCycoONE> (ra_get_node_class, mir_spill_register, midgard_compile_shader_nir, panfrost_shader_compile, panfrost_bind_shader_state, st_update_fp, st_validate_state, prepare_draw, _mesa_validated_drawrangeelements, _mesa_DrawElements. (no line numbers, sorry)
<alyssa> bbrezillon: Yeah, it was a core bug. It's all good now :)
<alyssa> robmur01: Eyy
<alyssa> TheCycoONE: Umm
<alyssa> I'm rewriting the RA right now so that function won't exist anymore and you'll have a different set of SEGFAULTs, so wontfix..? ;p
<TheCycoONE> :D ok
<TheCycoONE> worth bisecting?
<alyssa> TheCycoONE: No, literally all the code in midgard_ra.c is getting rewritten :)
<daniels> bbrezillon: btw are you waiting for more review on !2290, or ... ?
<bbrezillon> daniels: nope
<bbrezillon> I rebased and pushed a new version
<bbrezillon> was waiting for CI
<daniels> bbrezillon: cool :) you can just hit 'merge when pipeline succeeds' and it'll do that for you
<shadeslayer> bbrezillon: http://paste.ubuntu.com/p/kKNd9ympD9/ thoughts?
<shadeslayer> since Thomas asked for a userspace application using it
<bbrezillon> shadeslayer: SEGFAULT :P
<shadeslayer> /o\
<bbrezillon> struct drm_handle_set_label_object args = {};
<shadeslayer> bbrezillon: of course :P
<bbrezillon> shadeslayer: maybe add another patch adding labels to all BOs allocated by panfrost
<shadeslayer> bbrezillon: yeah, that's gonna be next, though ... what do we label them as :)
<bbrezillon> shadeslayer: 'git grep panfrost_bo_create' should give you a good idea of who is allocating BOs and what they do with those buffers
<bbrezillon> shadeslayer: 'git grep panfrost_batch_create_bo' too
<shadeslayer> bbrezillon: do you reckon that we should label only on debug builds?
<bbrezillon> probably
fysa has joined #panfrost
<alyssa> shadeslayer: yes
<shadeslayer> alyssa: bbrezillon http://paste.ubuntu.com/p/BrrZFncfWd/
<shadeslayer> actually, this is pretty fugly, we can just ifdef DEBUG everything inside set_label
<bbrezillon> shadeslayer: I wonder if we should patch the panfrost_bo_create() prototype to pass a label
<shadeslayer> bbrezillon: I considered that, I reasoned against it as label'ing is optional
<bbrezillon> new users of panfrost_create_bo() are likely to forget the set_label() call if we don't do that
<bbrezillon> it's optional, but we would like to have all BOs allocated by mesa to be labeled, right?
<bbrezillon> oh, and you should reset the label when returning a BO to the cache
<shadeslayer> ah yes, I knew I was forgetting something :P
<bbrezillon> shadeslayer: where is bo->label defined?
<bbrezillon> we could cache the value, but I wonder if it's really useful
<bbrezillon> and if we don't cache the value, we should have an ioctl to retrieve a label
<bbrezillon> (we'll need that for imported BOs anyway)
<shadeslayer> bbrezillon: yeah, I was thinking the same
<shadeslayer> bbrezillon: http://paste.ubuntu.com/p/jrkvs8njCj/
<shadeslayer> I need to cleanup the kernel uapi to match this and sync the header, but this is starting to look pretty decent imo
<shadeslayer> enough for me to open a MR
fysa has quit [Ping timeout: 268 seconds]
<shadeslayer> bbrezillon: I still don't have a answer to adjusting panfrost_bo_create to label all BO's
<shadeslayer> but maybe we can discuss that on MR
BenG83 has quit [Quit: Leaving]
BenG83 has joined #panfrost
<bbrezillon> shadeslayer: I'm pretty sure it's not compile tested :P
<bbrezillon> elsif does not exists in C
<bbrezillon> BTW, the if (is_texture) ... else if (is_2d) ... else if (is_streaming) ... is wrong
<bbrezillon> those 3 things are orthogonal
<shadeslayer> bbrezillon: I'm pretty sure of that as well, but I fixed it ;)
<shadeslayer> PS: too much ruby
<shadeslayer> bbrezillon: so you can have all 3 at the same time?
<shadeslayer> a streaming 2D texture?
<bbrezillon> well, a 2D texture, definitely
<bbrezillon> streaming is just a hint about the type of access and its frequency
<shadeslayer> bbrezillon: hmmm ... should we just string those together then
<bbrezillon> we could
fysa has joined #panfrost
<bbrezillon> not sure if it's important to know the kind of texture or the resource usage
fysa has quit [Remote host closed the connection]
<bbrezillon> alyssa: ^
<bbrezillon> shadeslayer: you should at least print the bind attribute https://elixir.bootlin.com/mesa/latest/source/src/gallium/docs/source/screen.rst#L711
pH5 has joined #panfrost
<shadeslayer> bbrezillon: mmmhhh ... I suppose we can print it in int .. and a userspace tool would have to convert it back into the flags
enunes has quit [Ping timeout: 250 seconds]
davidlt has quit [Remote host closed the connection]
davidlt has joined #panfrost
enunes has joined #panfrost
stikonas has joined #panfrost
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #panfrost
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #panfrost
<alyssa> bbrezillon: kind is more important, but the usage can help for debugging perf issues
davidlt has quit [Ping timeout: 240 seconds]
stikonas has quit [Read error: Connection reset by peer]
stikonas has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
stikonas has quit [Ping timeout: 246 seconds]
stikonas has joined #panfrost
BenG83 has quit [Quit: Leaving]
BenG83 has joined #panfrost
pH5 has quit [Remote host closed the connection]
enunes has quit [Ping timeout: 265 seconds]
enunes has joined #panfrost
HdkR has joined #panfrost
stikonas has quit [Remote host closed the connection]
<alyssa> alyssa: Sort constraints by alignment reqs
<alyssa> How so?Erm
<alyssa> Erm
<alyssa> Differing alignment causes messiness but to be addressed I guess