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/
Leopold__ has quit [Remote host closed the connection]
Leopold_ has joined #lima
yuq825 has joined #lima
Daanct12 has joined #lima
Leopold_ has quit [Remote host closed the connection]
Leopold_ has joined #lima
Leopold_ has quit [Remote host closed the connection]
Leopold_ has joined #lima
Daaanct12 has joined #lima
Daanct12 has quit [Ping timeout: 480 seconds]
Daaanct12 has quit [Quit: WeeChat 4.0.4]
yuq825 has quit []
sunshavi has quit [Remote host closed the connection]
sunshavi has joined #lima
drod has joined #lima
Leopold_ has quit []
Leopold_ has joined #lima
megi has quit [Ping timeout: 480 seconds]
<enunes> anarsoul: I made some progress in the 4096 texture issue, by creating a smaller reproducer and being able to dump with the blob driver
<anarsoul> great!
<anarsoul> so what's the issue?
<enunes> I think it has to do with how we calculate BLOCK_STEP and BLOCK_STRIDE for that specific size
<anarsoul> oh
<anarsoul> so kusma was right
Daanct12 has joined #lima
<anarsoul> block_num is incorrect?
megi has joined #lima
<anarsoul> oh
<anarsoul> block_stride is also different
<enunes> shift_w too
<enunes> indeed he was in the right direction
Danct12 has quit [Ping timeout: 480 seconds]
<anarsoul> :\
<enunes> my smaller reproducer actually reads the texture in half so I guess it makes sense
<anarsoul> so I assume block_w: 0 is invalid
<anarsoul> and that's why we hit it
<anarsoul> maybe it makes sense to add an assert for that
<enunes> it doesn't seem to be possible to be 0 so I wonder how we end up with it
<anarsoul> overflow?
<anarsoul> well, we are shifting it right, so it can reach 0
<anarsoul> width = (width + 1) >> 1;
<enunes> with some prints I see fb->block_w 256 so I'm quite confused now
<enunes> wow #define PLBU_CMD_BLOCK_STRIDE(block_w) PLBU_CMD((block_w) & 0xff
<anarsoul> yet I get width: 256, height: 2, shift_w: 0, shift_h: 0
<anarsoul> ouch
<anarsoul> :)
<anarsoul> well, sounds like width cannot be 256?
<anarsoul> or can it?
<anarsoul> yeah, I think it's 8 bit
<anarsoul> so 4096 needs a special handling
<anarsoul> enunes: something like https://pastebin.com/U361wyHK ?
<enunes> seems like it, I was just coming up with something similar to test
<enunes> yep bug is gone
<anarsoul> great!
Leopold_ has quit [Ping timeout: 480 seconds]
Leopold_ has joined #lima
drod has quit [Remote host closed the connection]