ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
ourdumbfuture has joined #asahi-gpu
odak_ has quit [Quit: odak_]
odak_ has joined #asahi-gpu
possiblemeatball has joined #asahi-gpu
nimprod3l has joined #asahi-gpu
possiblemeatball has quit [Quit: Quit]
possiblemeatball has joined #asahi-gpu
turo has joined #asahi-gpu
skmp__ has joined #asahi-gpu
yamii_ has quit [Ping timeout: 480 seconds]
yamii_ has joined #asahi-gpu
odak_ has quit [Quit: odak_]
odak_ has joined #asahi-gpu
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
nimprod3l has quit [Ping timeout: 480 seconds]
nimprod3l has joined #asahi-gpu
ourdumbfuture has joined #asahi-gpu
nimprod3l has quit [Ping timeout: 480 seconds]
nimprod3l has joined #asahi-gpu
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
nimprod3l has quit [Ping timeout: 480 seconds]
aafeke_ has joined #asahi-gpu
mcint has joined #asahi-gpu
WindowPa- has joined #asahi-gpu
WindowPain has quit [Ping timeout: 480 seconds]
odak_ has quit [Quit: odak_]
odak_ has joined #asahi-gpu
cylm_ has quit [Ping timeout: 480 seconds]
aafeke_ has quit [Quit: aafeke_]
ourdumbfuture has joined #asahi-gpu
alyssa has joined #asahi-gpu
<alyssa>
new hypothesis that maybe there's some cache bit I'm setting wrong
<alyssa>
two texture loads, first from read/write image, second readonly
<alyssa>
third argument goes 13->gather
<alyssa>
seems potentially interesting
<alyssa>
otherwise no difference
<alyssa>
texture descriptors are the same up to compression, so that rules out any cache controls there
<alyssa>
13->12 on q3 which is 43:5
<alyssa>
so really that corresponds to setting bit 43 for the read/write case
<alyssa>
maybe
<alyssa>
or could be unrelated
<alyssa>
unclear, isnt that exciting
alyssa has quit [Quit: leaving]
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
ourdumbfuture has joined #asahi-gpu
mlp has quit [Read error: Connection reset by peer]
mlp has joined #asahi-gpu
alyssa has joined #asahi-gpu
<alyssa>
OK!
<alyssa>
If I set bit 43 on the image loads, the test passes
<alyssa>
WHYYYYY
<alyssa>
oh but now the compute shader version is failing because of course it is
<alyssa>
other debug junk, sure
<alyssa>
TellowKrinkle: to follow up on our previous chat, the reason I was getting flaking but your test with Metal wasn't is because Apple is correctly setting bit 43 of texture_load when loading from read-write images
<alyssa>
It's definitely some sort of cache control bit
<alyssa>
not sure exactly what
ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<alyssa>
I've been putting off figuring out cache bits but ugh
ourdumbfuture has joined #asahi-gpu
<alyssa>
At any rate that gets KHR-GLES* down to a single fail
<alyssa>
(due to eMRT)
<alyssa>
which means I now finally have a feature complete image implementation*
<alyssa>
* except atomics are still technically broken i know i know i'm getting to it
<alyssa>
also I broke texelFetch but yknow, details