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/
drod has quit [Remote host closed the connection]
freemangordon has joined #lima
drod has joined #lima
drod has quit [Remote host closed the connection]
drod has joined #lima
Danct12 has quit [Quit: A-lined: User has been AVIVA lined]
Danct12 has joined #lima
Leopold_ has quit [Remote host closed the connection]
Leopold_ has joined #lima
<anarsoul|2> enunes: I checked the shader for https://gitlab.freedesktop.org/mesa/mesa/-/issues/9323
<anarsoul|2> it looks like it is heavily scalarized
<anarsoul|2> and atan is lowered instead of using native instruction
<anarsoul|2> looks like nir doesn't have support for native atan
<anarsoul|2> and we really could use a copy propagation pass in ppir :)
<anarsoul|2> It doesn't look good: mov.v1 $0 $1, stop
<anarsoul|2> if I replace atan with a multiplication, instr number goes down to 36
<anarsoul|2> from 67
<anarsoul|2> and looks like we don't disassemble atan correctly :(
<enunes> anarsoul|2: cool, I hadn't looked into the actual reason yet before posting there
<enunes> so looks like at least atan can be added to nir and ppir for some improvement, maybe even relatively not too complicated work
megi has quit [charon.oftc.net liquid.oftc.net]
tanty has quit [charon.oftc.net liquid.oftc.net]
rellla has quit [charon.oftc.net liquid.oftc.net]
mmind00 has quit [charon.oftc.net liquid.oftc.net]
megi has joined #lima
rellla has joined #lima
tanty has joined #lima
mmind00 has joined #lima
Leopold__ has joined #lima
<anarsoul|2> enunes: yeah, shouldn't be too complicated
Leopold_ has quit [Ping timeout: 480 seconds]
Leopold__ has quit [Remote host closed the connection]
Leopold_ has joined #lima