ChanServ changed the topic of #lima to: Development channel for open source lima driver for ARM Mali4** GPUs - Kernel has landed in mainline, userspace driver is part of mesa - Logs at https://people.freedesktop.org/~cbrill/dri-log/index.php?channel=lima and https://freenode.irclog.whitequark.org/lima - Contact ARM for binary driver support!
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #lima
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #lima
megi has quit [Ping timeout: 245 seconds]
BenG83 has quit [Ping timeout: 264 seconds]
mrueg has quit [Quit: http://quassel-irc.org - Chat comfortably. Anywhere.]
mrueg has joined #lima
kaspter has quit [Ping timeout: 250 seconds]
kaspter has joined #lima
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #lima
kaspter has quit [Ping timeout: 276 seconds]
kaspter has joined #lima
dddddd has quit [Remote host closed the connection]
<MoeIcenowy> anarsoul: why is it setting the freq?
<anarsoul> MoeIcenowy: because blob just does it?
Barada has joined #lima
<anarsoul> anyway, setting it to 384 doesn't fix blob for me
<anarsoul> so looks like I can't make mali dumps anymore :\
kaspter has quit [Ping timeout: 276 seconds]
kaspter has joined #lima
<anarsoul> rellla: can you make a dump of this one for me? https://github.com/yuq/gfx/tree/master/gbm-surface-part
<anarsoul> preferably on both mali400 and mali450
<MoeIcenowy> anarsoul: what happened on this test?
<anarsoul> MoeIcenowy: it's partial update
<anarsoul> I want to check how exactly blob does it for mali400 and mali450
<anarsoul> I suspect that dlbu in mali450 can generate partial pp stream
<anarsoul> for mali400 we may still need to generate it
<anarsoul> but it just looks quite unlikely to me that you have to generate 3k-8k entries for each frame to do partial update
<anarsoul> basically specify each 16x16 tile that you want to repaint
camus has joined #lima
kaspter has quit [Ping timeout: 240 seconds]
camus is now known as kaspter
<rellla> this is for 450, i had to do some gbm/egl tweaks to get it running, but GL code is the same.
<rellla> sadly i haven't set up mali400 with blobs anywhere and i don't know if i'm in the mood to do the horror once again :)
<rellla> at least not it needs some days :p
<rellla> s/not//
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #lima
Barada has quit [Quit: Barada]
adjtm_ has quit [Ping timeout: 240 seconds]
adjtm has joined #lima
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #lima
<MoeIcenowy> anarsoul: looks like lima now still sometimes fail to allocate memory...
<MoeIcenowy> anarsoul, bshah: BTW KWin GL2 (not ES) compositing is still severely broken on lima
<bshah|matrix> I haven't actually attempted that...
<rellla> anarsoul: sth like that should work: https://pastebin.com/ukbpBxY9
<rellla> it produces sth like https://pastebin.com/raw/S5YKtLaa
<rellla> i will do the same for plbu
megi has joined #lima
ecloud is now known as ecloud_wfh
adjtm has quit [Ping timeout: 246 seconds]
bshah|matrix has quit [Read error: Connection reset by peer]
Danct12[m] has quit [Remote host closed the connection]
asriel has quit [Write error: Connection reset by peer]
z3ntu has quit [Read error: Connection reset by peer]
romainmahoux[m] has quit [Read error: Connection reset by peer]
Danct12 has quit [Write error: Connection reset by peer]
insep[m] has quit [Read error: Connection reset by peer]
anarsoul[m] has quit [Remote host closed the connection]
danqo has quit [Write error: Connection reset by peer]
MartijnBraam has quit [Write error: Connection reset by peer]
dllud has quit [Ping timeout: 268 seconds]
<MoeIcenowy> we have bad GL_ARB_draw_elements_base_vertex
<MoeIcenowy> this is why kwin GL backend doesn't work
<MoeIcenowy> GLES backend works because there's no way to export glCopyBuffer on GLES2 with extensions
adjtm has joined #lima
<MoeIcenowy> and the KWin codepath requires both functions
<mripard_> anarsoul: can you send a patch to fix it?
dllud has joined #lima
<MoeIcenowy> mripard_: ?
<mripard_> MoeIcenowy: ?? :)
<mripard_> I was talking about the clock rate issue anarsoul reported
<MoeIcenowy> oh
<MoeIcenowy> for blob-based sunxi-mali?
megi has quit [Ping timeout: 265 seconds]
<mripard_> yep
dddddd has joined #lima
megi has joined #lima
<anarsoul> rellla: thanks
<anarsoul> MoeIcenowy: fix it?
<anarsoul> mripard_: will do tonight
<MoeIcenowy> anarsoul: fixed
<MoeIcenowy> but I still cannot get kwin running
<MoeIcenowy> and strangely for the same commit, I can get kwin running when built natively, cannot get when built cross
<anarsoul> :)
<MoeIcenowy> and a weird backtrace is shown...
<MoeIcenowy> which ends at `0x00000000fbff0000 n/a (n/a)`
<anarsoul> MoeIcenowy: that's why I usually don't cross compile
<MoeIcenowy> strange... to locate where the fail happens
<MoeIcenowy> it's located at a glDrawArrays
<MoeIcenowy> is mesa lazy to pass state to gallium driver?
<MoeIcenowy> but at least I'm sure the misrender by glDrawElementsBaseVertex is fixed
<anarsoul> submit an MR?
Elpaulo has quit [Quit: Elpaulo]
<MoeIcenowy> doing
<MoeIcenowy> !2430
<anarsoul> MoeIcenowy: this will likely conflict with enunes' changes, so please coordinate with him
<MoeIcenowy> anarsoul: do you mean the work beyond the vertices limitation?
<anarsoul> yes
<enunes> this is a small change, I think it can go first and I rebase
<MoeIcenowy> anarsoul: built this version natively, kwin can now run with GL backend, but looks some texture error is present
<anarsoul> what texture error?
Ntemis has joined #lima
<Ntemis> hello
<anarsoul> Ntemis: hi
<Ntemis> am here to report some issues we face while trying to use lima on RetroArch and the different cores it supports
<anarsoul> please open an issue at https://gitlab.freedesktop.org/lima/mesa/issues
<Ntemis> i really wish for Mesa 19.3 to be usable for broad usage by the public
<anarsoul> make sure you include apitrace
<Ntemis> For retroarch with lima, it still refuses to render the retroarch GUI.
<Ntemis> The animated ribbon in the background appears in greyscale and nothing else appears on the screen.
<Ntemis> Lima doesn’t seem to like some part of the modeswitching Process. Not sure exactly what part, whether it’s creating the gbm device or surface or setting the crtc mode.
<MoeIcenowy> anarsoul: seems random
<MoeIcenowy> but the fade animation is totally broken
<anarsoul> Ntemis: let me guess, 64 bit kernel and 32 bit userspace?
<Ntemis> can you guys please try to use retroarch with lima on 32bit userspace
<Ntemis> :)
<Ntemis> U got it
<MoeIcenowy> anarsoul: ah does Lima have bitness issue?
<anarsoul> Ntemis: I don't think that anyone is working on it
<Ntemis> so a known issues i must guess
<MoeIcenowy> I think Lima should be 32-bit only
<anarsoul> MoeIcenowy: that's likely display driver, not lima
<MoeIcenowy> oops... sun4i-drm is quite simple
<anarsoul> or kmsro
<anarsoul> anyway, out of my area of expertise and interest :)
<anarsoul> Ntemis: most lima devs work on lima in their spare time and as far as I know there's little to no interest in debugging issues of 32-bit userspace on 64 bit kernel
<anarsoul> it would be easier for you to switch to 64 bit userspace
<Ntemis> yeap but all current emu cores on retro distros are aarch32 due to dynarec support
<Ntemis> so we need this to work am afraid :(
<Ntemis> my self am a contributor for lakka and batocera and i would like to switch to mesa 19.3 and kernel 5.4 when they are out to all devices that support it
<Ntemis> allwinner H5 would be my 1st candidate following amlogic S905
<Ntemis> if i manage to make that work adequate
<Ntemis> and i wait for amlogic S812 to get hdmi out working so i can resurrect it too
<Ntemis> S812 is the most powerful 32bit soc i encounter so far on 32bit
<Ntemis> that Mali450 MP6+2 really shines when you use it for emulation purposes
<anarsoul> keep in mind that mali4x0 still has only one GP (one that runs vertex shaders)
<anarsoul> so MP6 means 1 GP core and 6 PP cores
<Ntemis> in paper it states 2 GP cores
<anarsoul> that's a lie
<anarsoul> (go check your dts for number of GP interrupts if you don't believe me)
<Ntemis> its official document by amlogic
<Ntemis> its officially documented
<Ntemis> hmm
<anarsoul> yeah, and official documentation from ARM never mentions that mali450 has 2 GP cores :)
<Ntemis> thank you for clearing this up for me
<anarsoul> no worries
<Ntemis> still is a powerful gpu though never the less
<anarsoul> well, not really
<Ntemis> can run mupen64 on 720p without losing a frame
<Ntemis> i call this a good gpu
<anarsoul> Ntemis: they beated most of competitors by cheating
<anarsoul> PP supports only fp16
<Ntemis> yes thats correct
<Ntemis> other on paper and other on real life and trust me S812 beats the S905 without a blink
<Ntemis> so to be clear with 32bit kernel this issue doesnt exist?
<Ntemis> lets say on Allwinner H3 lima will be "ok"
<anarsoul> Ntemis: should be better
<anarsoul> you may hit other bugs
<Ntemis> i see
<chewitt> fwiw, LibreELEC runs 32-bit userspace on 64-bit kernel for nearly all arm boards
<chewitt> for two reasons; a) libwidevine evil needed for Netflix/Amazon can only be found in 32-bit format for arm devices
<chewitt> and b) Kodi can run libretro emulators, which are mostly 32-bit only (as Ntemis mentioned)
<anarsoul> chewitt: well, then we need someone with experience in display drivers and kmsro to look into it
BenG83 has joined #lima
Tofe has quit [Ping timeout: 276 seconds]
Tofe has joined #lima
<enunes> is there a simpler way to reproduce these issues other than running these other distributions or emulator stuff?
<Ntemis> @enunes these other distributions you are referring is the perfect example how to use the lima driver to its fully
<enunes> Ntemis: well, what I mean is that the easiest it can be made to debug without requiring a lot of effort to set up, the most likely it can get fixed soon
<enunes> setting up one of these distributions and figuring out how to install some emulator core and run some game is quite a bit of effort
<enunes> some simpler application or reproducer would be helpful
z3ntu has joined #lima
z3ntu has quit [Ping timeout: 250 seconds]
z3ntu has joined #lima
warpme_ has quit [Quit: Connection closed for inactivity]
adjtm has quit [Read error: Connection reset by peer]
adjtm has joined #lima
danqo has joined #lima
<Ntemis> @enunes how about i offer you guys a ready to flash image that has them already included?
<Ntemis> the issue here is what board you guys have to test uppon
<enunes> Ntemis: I think apitrace is preferred
<enunes> actually I was able to run a 32-bit container on my 64-bit fedora board and run kmscube/glmark2/apitrace on it so I might be able to reproduce with that
<Ntemis> ow! perfect
<enunes> so I guess file a issue on https://gitlab.freedesktop.org/lima/mesa/issues like anarsoul mentioned and attach the apitrace (please trim it to as small as possible) to it