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/
megi2 has joined #lima
megi1 has quit [Ping timeout: 480 seconds]
<anarsoul> ugh, we have a bug with varyings
<anarsoul> or rather with split_input
<anarsoul> vec2s must also be aligned
<anarsoul> also texture bias is broken.
<anarsoul> (and we already have 2 failures there)
<anarsoul> I suspect it's a register conflict
<anarsoul> texld lod argument conflict with subsequent mov.v1 $0 ^texture
<anarsoul> so if lod is in $0 it fails
mranostaj has quit [Remote host closed the connection]
mranostaj has joined #lima
<anarsoul> well, I'm wrong, it's broken the same on blob
<anarsoul> rellla: enunes: can I get a r-b for https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13131/ please? It's a vary nasty bug with very simple fix
mranostaj has quit [Quit: leaving]
mranostaj has joined #lima
chewitt has joined #lima
<enunes> anarsoul: I'm looking at the MRs now
<enunes> I'm still in sort of a vacation for the next couple of weeks so mostly offline
adjtm has quit [Ping timeout: 480 seconds]
adjtm has joined #lima
Viciouss7 has quit []
Viciouss has joined #lima
chewitt has quit [Read error: Connection reset by peer]
adjtm has quit [Remote host closed the connection]
adjtm has joined #lima
megi2 has quit []
megi has joined #lima
drod has joined #lima
<anarsoul> enunes: I see, sorry for bothering you on your vacation :)
<anarsoul> and thanks for reply!
<anarsoul> rellla: re: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13138 - if you really need disassembly because you're working on some compiler improvement, I'd suggest to capture shaders with MESA_SHADER_CAPTURE_PATH and then compile them with mesa-shader-db
<anarsoul> with drm-shim added by icecream95 you can even do it on x86 machine
<anarsoul> I'm really not sure if we need standalone compiler tool anymore :)
<anarsoul> (it never worked reliably for me anyway)
<rellla> anarsoul: the reason for this commit was, that i always found it unhandy, that there never was a shader disassembly in the dump file. you only can get it on-the-fly at stdout ... it's more a thing of have-a-complete-dump
<anarsoul> rellla: yeah, it's convenient to have disassembly in the dump
<rellla> i never used the standalone thing btw.
<anarsoul> I used it for compiler development, compiling on my arm machines is way slower than on my x86
<anarsoul> now I use mesa-shader-db + drm-shim
<rellla> i had this on my list a long time and now it was triggered by your MR, because i wanted to check, why this special test fails now. find out, whats different in an easy way.
<anarsoul> rellla: well, we don't use txp lowering anymore, that is what's different
<anarsoul> disassembly is pretty much identical to the blob and it also fails in this test
<anarsoul> btw, image comparison doesn't look that bad
<rellla> right
<anarsoul> difference is above threshold though
<rellla> maybe hw has some problems...
<anarsoul> we'll never know :)
<rellla> :)
<rellla> i'm sure, the remaining stencil tests fail due to a hw bugl, but they are really no-real-world tests, so...
<anarsoul> rellla: btw new failure used to pass if coords are coming from reg, not varying output
<anarsoul> so it could be something with lod bias not dealing with high precision coords properly?