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: 265 seconds]
buzzmarshall has quit [Remote host closed the connection]
tlwoerner has quit [Excess Flood]
tlwoerner has joined #lima
yuq825 has quit [Ping timeout: 260 seconds]
Barada has joined #lima
Barada has quit [Client Quit]
yuq825 has joined #lima
dddddd has quit [Ping timeout: 272 seconds]
monstr has joined #lima
<rellla> anarsoul: another finding ... lima and blob does differ in the halffloat value of the uniform, which is used as color.g, see:
<rellla> http://imkreisrum.de/deqp/dsmasked/lima.dump.0001 -> grep for 230a9c0 and
<rellla> maybe this is the issue, because the test is very flaky depending on how many clears and how many draws i do
<rellla> my guess is, that stencil cmds are ok for rsw, but the value which is tested isn't right or accurate enough?
<rellla> i also wonder, what https://gitlab.freedesktop.org/mesa/mesa/blob/master/src/gallium/drivers/lima/lima_draw.c#L706 does and if it's correct, because that's another difference to the blob
Barada has joined #lima
yann has joined #lima
yann has quit [Ping timeout: 268 seconds]
<anarsoul> rellla: last one should be replaced with util_logbase2_ceil()
<anarsoul> I didn't know it exists when I implemented the code above :)
<anarsoul> rellla: IIRC blob packs uniforms, lima doesn't
<anarsoul> so we allocate vec4 per uniform regardles of how many components it has
<rellla> anarsoul: and why do we have 0x3632 in lima as fp16 value and 0x3631 in mali for the same value?
<anarsoul> rellla: likely blob rounds it differently?
<rellla> where is this done in our code?
camus1 has joined #lima
<anarsoul> in lima_update_pp_uniform()
<anarsoul> util_float_to_half()
<anarsoul> I doubt it really matters though
kaspter has quit [Ping timeout: 265 seconds]
camus1 is now known as kaspter
<anarsoul> that's very little difference:
<anarsoul> (gdb) call _mesa_half_to_float(0x3632)
<anarsoul> $1 = 0.387207031
<anarsoul> (gdb) call _mesa_half_to_float(0x3631)
<anarsoul> $2 = 0.386962891
<enunes> hey, I was out for a while with some travel after fosdem, now I finally have some time again; I'm planning to work on some optimizations and refactor in ppir, assuming nobody else is working on that
megi has joined #lima
<yuq825> enunes: nice, combine instructions?
<enunes> yuq825: yes that would be one thing, I had the thought of implementing a general combine pass and see what we can improve from there
<enunes> like changing scheduling or the earlier steps to favor combining
yuq825 has quit [Quit: Leaving.]
Barada has quit [Quit: Barada]
dddddd has joined #lima
yann has joined #lima
yann has quit [Ping timeout: 240 seconds]
yann has joined #lima
champagneg has quit [Ping timeout: 265 seconds]
champagneg has joined #lima
champagneg has quit [Ping timeout: 265 seconds]
champagneg has joined #lima
yann has quit [Ping timeout: 240 seconds]
monstr has quit [Remote host closed the connection]
<rellla> i have no more ideas :/
yann has joined #lima
Elpaulo has joined #lima
champagneg has quit [Ping timeout: 240 seconds]
champagneg has joined #lima
yann has quit [Ping timeout: 265 seconds]
<anarsoul> rellla: just drop this test for now?
<rellla> anarsoul: :p you told me that already ;)
<anarsoul> well, stencil is not widely used feature
warpme_ has quit [Quit: Connection closed for inactivity]
buzzmarshall has joined #lima
yann has joined #lima
yann has quit [Ping timeout: 272 seconds]