ChanServ changed the topic of #etnaviv to: #etnaviv - the home of the reverse-engineered Vivante GPU driver - Logs https://oftc.irclog.whitequark.org/etnaviv
chewitt has joined #etnaviv
pjakobsson_ has joined #etnaviv
pjakobsson has quit [Ping timeout: 480 seconds]
JohnnyonFlame has quit [Read error: Connection reset by peer]
pcercuei has joined #etnaviv
<mntirc> now building latest mesa with !25756 and !26096 on imx8mp
<f_> oh hi mntirc
<mntirc> hi f_
<f_> Why isn't your nick `minute` on oftc?
<mntirc> f_: not sure, maybe it was taken
<mntirc> marex: chromium 119 works for me with the clear patch, but has lots of corruption still
<mntirc> [12374.587555] etnaviv-gpu 38000000.gpu: MMU fault status 0x00000002
<mntirc> [12374.593668] etnaviv-gpu 38000000.gpu: MMU 0 fault (page not present) addr 0xfe133000
<mntirc> [12375.101393] etnaviv-gpu 38000000.gpu: GPU not yet idle, mask: 0x7fffbf9a
<mntirc> 20fps on webgl blobs, that's not nothing
<mntirc> austriancoder: anything that can be done about these mmu faults with chromium? do i need a newer kernel than 6.4? (this machine has only 4GB ram btw)
<austriancoder> mntirc: kernel version is fine. I only need to run the wegl blobs demo - or? Will have a look tomorrow.
<marex> mntirc: is that with egl or angle
<marex> ?
<marex> mntirc: btw I have yet to try to get the v4l2 stateless decoder going in chromium, but it's low prio and on the backburner
<mntirc> marex: angle
<marex> mntirc: and if you use egl, the gl/rendering/compositor process crashes, right ?
<mntirc> marex: gl=egl-angle,angle=opengl
<mntirc> marex: not sure if i was ever able to use egl
<marex> mntirc: that's how I used it until like chromium 109 or so, then it stopped working
<marex> mntirc: is angle somehow preferrable ?
<mntirc> marex: i think egl requires GLES 3?
<mntirc> marex: in my experience angle is faster
<marex> angle is gles3, egl is gles2 , no ?
<marex> huh
<marex> then maybe ignoring egl is just fine too
<mntirc> i think so
<marex> :)
<mntirc> marex: i have rmader in a chat and he's asking about hw rotation. lcdif can't do that, right? rotated scanout?
<mntirc> on imx8mp we always have to blit on a 90deg display?
<marex> mntirc: yes
<marex> I think there is CSC in it, but not ROT
<mntirc> ok
<marex> mntirc: wait a minute !
<mntirc> oh
<marex> 13.3.2.1 Graphics and data format
<marex> The output
<marex> pixels are passed onto the LUT and rotation engine for further processing.
<marex> but ... that is the only mention of rotation in that entire lcdif section in mx8mprm
<mntirc> weird
<marex> mntirc: considering the CSC is pulled out of PXP (pixel pipeline, from MX6UL and I think MX28/23), not really
<marex> PXP had rotator engine, but the LCDIFv3 only has the CSC part of PXP
<marex> I guess this was copy-paste hardware design here
<marex> and copy-paste documentation design
<mntirc> marex: the RM mentions rotation in the 2D GPU part
<mntirc> austriancoder: for me just googling "test" is the most reliable reproducer
<mntirc> austriancoder: i.e. searching for "test" on google.com
<mntirc> i also have really interesting/weird issues with images in chromium, will upload some screenshots
<mntirc> austriancoder: marex: pictures on our website are... green http://dump.mntmn.com/screenshot-2023-11-26-18-12-51.png
<mntirc> (and include scaled up parts of the window decoration)
<mntirc> and pictures on wikipedia are filled with gradients http://dump.mntmn.com/screenshot-2023-11-26-18-12-01.png
<marex> mntirc: well 2D GPU can do just that, 2D fixed pipeline transformations, sure
<marex> mntirc: but that's memory-to-memory operation, i.e. consumes extra DRAM bandwidth
<marex> mntirc: weird, very weird
<marex> mntirc: I have not seen that
<marex> but then, my testing was generally limited
<marex> mntirc: mesa version ?
<mntirc> marex: git from today, with MRs !26096 !25756 and not sure if !26089 atm
Surkow|laptop has quit [Remote host closed the connection]
Surkow|laptop has joined #etnaviv
pcercuei has quit [Quit: dodo]