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!
yuq825 has joined #lima
megi has quit [Ping timeout: 246 seconds]
Barada has joined #lima
dddddd has quit [Remote host closed the connection]
dllud has quit [Quit: ZNC 1.7.4 - https://znc.in]
smaeul has quit [Quit: cya]
dllud has joined #lima
smaeul has joined #lima
smaeul has quit [Quit: cya]
smaeul has joined #lima
Elpaulo has quit [Quit: Elpaulo]
<enunes> anarsoul: still on travel, back home this weekend
yuq825 has quit [Remote host closed the connection]
mripard_ has quit [*.net *.split]
yann has quit [Ping timeout: 250 seconds]
mripard_ has joined #lima
<rellla> 07:06 <anarsoul> rellla: so resetting ctx->index_offset fixes gp mmu fault, but tests are still broken
<rellla> do you mean resetting it to ctx->index_offset = 0?
<rellla> i'm trying to debug, why DRAW_ELEMENTS fails in the meantime
<rellla> we can ignore what i said before ("working with up to 125 indices per draw, not working with 126 and up"), because this is sth mesa does.
<rellla> s/does/wants to optimize/
yann has joined #lima
megi has joined #lima
kaspter has quit [Ping timeout: 250 seconds]
kaspter has joined #lima
cwabbott has quit [Remote host closed the connection]
cwabbott has joined #lima
monstr has joined #lima
Marex has quit [*.net *.split]
Marex has joined #lima
Barada has quit [Quit: Barada]
Marex has quit [*.net *.split]
Marex has joined #lima
<anarsoul> rellla: yes, you have to set it to 0
<rellla> thanks. btw, i tested your lima-enable-tiling branch: http://imkreisrum.de/deqp/master_5585b8eaddd_lima-enable-tiling/
<rellla> i get 9 new fails but i don't remember the exact reference fails.txt ... i think, we can ignore that. i don't want to recheck that tests.
<rellla> and i didn't run into kernel errors with this skip list.
<rellla> so i think i can add my rb and tb :)
yann has quit [Ping timeout: 268 seconds]
dddddd has joined #lima
yann has joined #lima
<rellla> wtf?
<rellla> anarsoul, enunes: write.use.* works now with that fixes upon enunes branch: https://gitlab.freedesktop.org/rellla/mesa/commits/lima-vertex-fixes-2
<rellla> though about introducing "util_upload_index_buffer_aligned" should be thought about. i'd prefer to redo util_upload_index_buffer ...
<rellla> i'll rebase that branch on master/enable-tiling and give deqp a chance now...
<rellla> anarsoul: https://gitlab.freedesktop.org/anarsoul/mesa/commits/lima-vertex-fixes with the additional fixes breaks that tests again btw :p
<anarsoul> yay, tiling is merged :)
<anarsoul> rellla: :(
<anarsoul> I know that the tests are broken in my branch
<anarsoul> I'll look into it once Erico's changes merge
<anarsoul> let me have a quick look on what can be broken
<anarsoul> rellla: I don't see anything obvious in my changes
<rellla> i will also take a look into the differences.
<anarsoul> new-fails is likely just a flacky test
<rellla> yes, it's passing when executed as single test.
<rellla> so no regressions. i think i can give my rb :)
monstr has quit [Remote host closed the connection]
<anarsoul> IIRC it still needs fixes in gallium
<rellla> i also should drop write.use.index_array.* from the skip list, as it passes now.
<anarsoul> rellla: I don't think that your fixes actually fix underlying issue for this tests
<anarsoul> they're likely just flacky
<anarsoul> and in your cases timing is right for them to work
<rellla> hm...
<rellla> i will watch them...
<rellla> if they are flaky, your branch most likely is ok then
<rellla> in the meaning of having no regressions. will check that tomorrow again.
<anarsoul> my branch just changes location of gl_Position and gl_PointSize to the same BO as varyings
<anarsoul> all the alignments are honored
<anarsoul> sizes are the same as before
<rellla> i think we still have an issue with indexed drawing somewhere
<anarsoul> likely
Kwiboo has quit [Ping timeout: 252 seconds]