ChanServ changed the topic of #lima to: Development channel for open source lima driver for ARM Mali4** GPUs - Kernel has landed in mainline, userspace driver is part of mesa - Logs at https://people.freedesktop.org/~cbrill/dri-log/index.php?channel=lima and https://freenode.irclog.whitequark.org/lima - Contact ARM for binary driver support!
buzzmarshall has quit [Quit: Leaving]
<mariogrip> anarsoul: Thanks! that did fix the issue :D
<mariogrip> (you'r patch on mesa that is, sorry wrong replied room)
<mariogrip> (your patch on mesa that is, sorry wrong replied room)
chewitt has quit [Read error: Connection reset by peer]
warpme_ has quit [Quit: Connection closed for inactivity]
yuq825 has joined #lima
buzzmarshall has joined #lima
Kwiboo has joined #lima
chewitt has joined #lima
chewitt has quit [Read error: Connection reset by peer]
megi has quit [Ping timeout: 268 seconds]
yuq825 has quit [Remote host closed the connection]
buzzmarshall has quit [Quit: Leaving]
dddddd has quit [Remote host closed the connection]
_whitelogger has joined #lima
_whitelogger has joined #lima
<MoeIcenowy> anarsoul: maybe the fix to sun4i-drm will fix real-world problems?
<anarsoul> MoeIcenowy: it's more likely that https://gitlab.freedesktop.org/mesa/mesa/merge_requests/3564 will fix your issues :)
<MoeIcenowy> okay, lemme try
<anarsoul> MoeIcenowy: also there's https://gitlab.freedesktop.org/mesa/mesa/issues/2425
<anarsoul> that's easy to fix if you have few hours to work on it
<MoeIcenowy> ah I don't know well about the pipes
<MoeIcenowy> although I may have enough time
<anarsoul> you don't need to
<anarsoul> the problem is that pipe_context can be freed earlier than pipe_surface
<anarsoul> currently you'll hit assert in lima_context_destroy() at assert(!_mesa_hash_table_num_entries(ctx->plb_pp_stream)); when that happens
<anarsoul> and if you have release build you'll get use-after-free and crash somewhere
<anarsoul> solution is to improve plb stream caching
<anarsoul> (and likely get rid of reference counting)
<MoeIcenowy> how to reproduce this?
<anarsoul> replaying trace from https://gitlab.freedesktop.org/mesa/mesa/issues/2382 on X11 reproduces it
<MoeIcenowy> anarsoul: so should we bind pp_plb_stream to surface rather than context?
<MoeIcenowy> (or maybe it should be bound to the whole library?
megi has joined #lima
warpme_ has joined #lima
yann has joined #lima
dddddd has joined #lima
deesix_ has joined #lima
dddddd_ has joined #lima
deesix has quit [Ping timeout: 272 seconds]
dddddd has quit [Ping timeout: 265 seconds]
dddddd_ is now known as dddddd
deesix_ is now known as deesix
<enunes> anarsoul: btw I noticed that the shader from that trace does use multiple writes to the same reg in different blocks
<enunes> I think we have to roll back to the previous algorithm for non-ssa registers
<enunes> which I think is to just find the first write and last read and just have a big live range across blocks?
<enunes> there must be another solution but more complicated and definitely not before 20.0 branch
<enunes> I think that's not too bad if the improved results from the new algorithm for ssa registers still improve regalloc enough so we can compile the glamor shaders
<enunes> looks like branch is in 2-3 days https://www.mesa3d.org/release-calendar.html
<MoeIcenowy> anarsoul: do we still have mipmap issue?
<MoeIcenowy> on SuperTuxKart some grass is misrendered
<MoeIcenowy> a simple analyze of the apitrace shows that the grass is mipmapped
<MoeIcenowy> oh seems not mipmap issue... some covered object failed to be drawn
<MoeIcenowy> maybe blend or depth issue
buzzmarshall has joined #lima
<MoeIcenowy> could anyone dump the blob command stream of https://github.com/Icenowy/gfx/tree/depth-discard/gbm-surface-depth-discard ?
<MoeIcenowy> this program currently misbehaves on Lima
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #lima
megi has quit [Ping timeout: 272 seconds]
<anarsoul> enunes: yeah, sonds good
<anarsoul> MoeIcenowy: I think I know what's wrong
<anarsoul> let me try something
<anarsoul> MoeIcenowy: yeah
<anarsoul> aux0 bit9 is early-z
<anarsoul> looks like we have to disable it if shader contains discard
<anarsoul> let me fix that...
<MoeIcenowy> oops, I hadn't checked limadriver.org for a long time...
<MoeIcenowy> since it's down
<anarsoul> :)
<anarsoul> almost done
<MoeIcenowy> anarsoul: it works in STK
<anarsoul> cool
<anarsoul> some screenshots? :)
<MoeIcenowy> ah in fact I am replaying my trace
yann has quit [Remote host closed the connection]
enunes has quit [Read error: Connection reset by peer]
enunes has joined #lima
egbert has quit [Disconnected by services]
egbert has joined #lima
megi has joined #lima