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/
adjtm has joined #lima
adjtm has quit [Quit: Leaving]
<anarsoul>
I lowered dram clock on my rock64 to 667MHz (from 800), let's see if it improves stability
<anarsoul>
btw glmark issue is something different, it looks like it just crashes the kernel :)
<anarsoul>
I looked into spec@nv_texture_barrier@blending-in-shader failure and it seems to be precision-related
<anarsoul>
this test does 3 passes and in each pass it accumulates error
<anarsoul>
if I simplify the test to do a simple multiplication instead of sqrt it passes, yet it still samples the texture which is also a rendering target, so texture barrier works correctly
<anarsoul>
if I replace texture barrier with nop simplified test fails
<enunes>
anarsoul: thanks for looking into it, though is there anything we can do or do we need to keep the test in the fail list?
<anarsoul>
I can submit an MR that simplifies the test, but I'm not sure if it's acceptable since piglit targets desktop GL which mandates 32-bit precision
<anarsoul>
technically it'll be a workaround for GLES2-only GPU
<enunes>
yeah seems like overkill, and now we know that it is not an actual failure
<enunes>
I could improve the commit message adding that test to the list to mention this
drod has quit [Remote host closed the connection]
camus has quit [Remote host closed the connection]