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/
apol[m] has joined #lima
icecream95 has joined #lima
icecream95 is now known as Guest1832
mvnetbiz[m] has joined #lima
anarsoul has quit [Read error: Connection reset by peer]
anarsoul has joined #lima
megi1 has quit []
megi has joined #lima
drod has joined #lima
tlwoerner has quit [Read error: Connection reset by peer]
tlwoerner has joined #lima
<anarsoul>
rellla: hey
<anarsoul>
rellla: how did you find this unknown_1 bit that fixes lod bias tests failures?
<anarsoul>
I double-checked mali-offline compiler and it always has unknown_1 == 0
<rellla>
anarsoul: i tried all bits of unknown_0 and unknown_1
<rellla>
most of them have no effect, but some do. i think it's worth to check them all again.
<rellla>
bit 2 had the effect of changing the results of these tests.
<rellla>
mali-offline-compiler has a bug? so does the blob?
<anarsoul>
rellla: I doubt it, IIRC kusma said that there are no unused bits in mali400 isa
<anarsoul>
so it does something for sure, but I'm hesitant to use unknown bits
<anarsoul>
I really wish ARM released ISA docs :)
<rellla>
you mean, the blob compiler does respect every bit?
<rellla>
or do you mean, the blob does respect every bit in the correct way?
<anarsoul>
rellla: I mean that I doubt that they have a bug in code generation
<rellla>
if we don't try and without risk we won't find new bits that aren't set in the blob shader :)
<rellla>
and i doubt we will get any docs ;)
<anarsoul>
rellla: I suspect that unknown_1 can be absolute lod value
<anarsoul>
(5 bits, near other lod/bias bits)
<anarsoul>
maybe only some of these 5 bits are this constant
<anarsoul>
rellla: I wonder if we need something like src/panfrost/midgard/midgard_errata_lod.c
megi has quit [Quit: WeeChat 3.3]
megi has joined #lima
adjtm has quit [Remote host closed the connection]