ChanServ changed the topic of #lima to: Development channel for open source lima driver for ARM Mali4** GPUs - Kernel driver has landed in mainline, userspace driver is part of mesa - Logs at https://oftc.irclog.whitequark.org/lima/
<anarsoul>
and looks like no one touched draw_stencil_pixels() since ~2011 except code refactoring
<anarsoul>
heh, it passes with LIMA_DEBUG=notiling (and one extra fix to always align resources to tile boundary)
<anarsoul>
:\
<anarsoul>
well, that's definitely something
camus has joined #lima
<anarsoul>
ugh
<anarsoul>
I think I understand what's the issue
<anarsoul>
we need to implement transfer_flush_region() for lima
<anarsoul>
and move panfrost_store_tiled_image there
<anarsoul>
basically blitter job races with panfrost_store_tiled_image in transfer_unmap_inner()
<anarsoul>
because flush_region() does blit of the resource that is still mapped (and it's tiled, so uses staging)
<anarsoul>
anyway, I think u_transfer_helper is broken
<anarsoul>
it does flush_region() before flushing its staging
Daanct12 has joined #lima
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #lima
tanty has quit []
tanty has joined #lima
<enunes>
anarsoul: interesting, so does your new MR together with the blit one solve it?
<enunes>
should we wait for that before merging so there is no intermediate state with a regression?
Daanct12 has quit [Remote host closed the connection]
camus1 has joined #lima
camus has quit [Remote host closed the connection]
<anarsoul>
enunes: yeah, we should wait for my new MR to merge before proceeding with blit and msaa
<anarsoul>
enunes: guess we need to figure out what test causes OOM?
<enunes>
anarsoul: I ran a 1-thread run to ensure its not due to too many parallel tests, I think its a specific 3d texture test that reproduces it
<anarsoul>
max size? :)
<enunes>
looks like it
camus has joined #lima
<anarsoul>
8192 * 8192 * 8192 is not going to fit into memory for sure :)
<anarsoul>
but I'm not sure if it tries to allocate it
<anarsoul>
spec@!opengl 1.2@tex3d-maxsize is in skips list of most of the drivers
despoil has joined #lima
<enunes>
and now it will likely be on ours too
despoil has quit [Remote host closed the connection]
camus has quit [Ping timeout: 480 seconds]
drod has joined #lima
<anarsoul>
enunes: looks like we need to allocate tile-aligned buffers even for linear, it looks like they can reach the blitter even with no PIPE_BIND_RENDER_TARGET set
<anarsoul>
and honestly I don't see a good reason to keep this mess with aligned/not-aligned buffers