alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - - Logs - Transientification is terminating. Memory reductions in progress.
memeka has quit [Quit: Connection closed for inactivity]
memeka has joined #panfrost
<alyssa> Blah, wifi adaptor kerplutzed, need to figure out how to connect otherwise
<alyssa> "Why is my computer so slow?" "Oh, right, forgot I was compiling a Linux kernel in the background"
<robclark> yeah, that would do it
<ezequielg> alyssa: try building a browser.
<ezequielg> :P
<HdkR> Is that a challenge?
<ezequielg> sudo create-cpu 16
<ezequielg> although, right now, being 10pm and just arriving from yoga, i'd like to do: sudo prepare dinner
<alyssa> ezequielg: I've built dillo before, yes :P
<HdkR> But have you built webkit?
<alyssa> Can't say I have
<HdkR> It's not a great experience
<ezequielg> alyssa: seriously, a great trick is to put your config on a nice diet.
<ezequielg> and only build the minimum.
<alyssa> Still
<ezequielg> should not take more than.... 3 minutes?
<alyssa> I essentially compile on-target
<alyssa> :P
<ezequielg> ah
<ezequielg> that won't cut it
<ezequielg> i guess you have a kevin
<ezequielg> hard to tftp
<alyssa> I do have a kevin
<alyssa> And I transfer kernels via flash storage
<alyssa> It's ugly
* HdkR times chromium build
<alyssa> HdkR: lawl
<HdkR> Give it ten minutes
<alyssa> HdkR: You've a TR, cheaterbun
<alyssa> :P
<HdkR> :D
<ezequielg> alyssa: hm. do you need to do this very frequently?
<alyssa> Not at all, no
<alyssa> I stick to userspace
<alyssa> Just need to recompile kernel when I need to enable a new device driver or whatever
<alyssa> ...Still don't have audio, I don't think
<ezequielg> mainline
<ezequielg> ?
<alyssa> 4.20.something, ye
<alyssa> "rockchip-i2s ff880000.i2s: COuld not register PCM is suspect"
<ezequielg> this is my 4.20 config
<ezequielg> tested with chromiumos userspace, but afaicr I do have audio.
<ezequielg> on my kevin
<ezequielg> alyssa: ^
<alyssa> :eyes:
<alyssa> Anything I'm supposed to be looking for in particular?
<ezequielg> you are supposed to just use it :)
<ezequielg> if you want...
<alyssa> But i already have a config working.. for other stuff.. aaa :p
<ezequielg> life is full of choices
<HdkR> Dang. ten minutes up and only around 1/3rd the way compiled
<memeka> HdkR: i built webkit
<memeka> have you built chromium?
<HdkR> I'm doing it right now :P
<memeka> oh oh better: have you build tensorflow?
<HdkR> I stay as far away as I possibly can from DL
<alyssa> ezequielg: Profound :p
<alyssa> alsamixer "cannot open mixer: No such file or directory" eh-heh
<alyssa> invali card index: 0
<alyssa> Also "no soundcards found..."
<alyssa> /dev/snd is indeed lonely
stikonas has quit [Remote host closed the connection]
<alyssa> ezequielg: CONFIG_SND_SOC_TS3A227E=y
<alyssa> ^ What's that do?
<alyssa> Wat.
<alyssa> This code was broken last night
<alyssa> It is now not broken
<alyssa> What even
<alyssa> You know, not complaining
<HdkR> Grabbed a blob trace of es2tri. Let's see what we are messing up :P
<HdkR> Interesting. it is still using job_descriptor_size of 1
<HdkR> kms mandates a display is connected to run? That's really obnoxious
<memeka> HdkR: yes, needs display :)
<memeka> HdkR: progress? :D
<HdkR> I've got a trace of the blob, trying to see what is different
<HdkR> no way
<HdkR> It's fricken core_req
<HdkR> Alright, peeling back changes until I can hit a minimum change list
<HdkR> memeka: So yes, progress
<memeka> :)
<memeka> what's core_req? :D
<HdkR> There are two job descriptor core requirement descriptions. Tells what the job is doing
<HdkR> compat_core_req and core_req
<HdkR> t8xx only requires core_req to work
<HdkR> t6xx requires that core_req is 0 and to use compat_core_req
<HdkR> panfrost sets both to the same thing atm
<memeka> heh
<HdkR> By request of me when testing /something/
<HdkR> But it would have been broken regardless so w/e
<HdkR> memeka: Dirty patch
<HdkR> I can't test rendering. It just makes the job submission failures go away
<HdkR> Does kmscube typically render to a screen?
<daniels> HdkR: yep
<HdkR> Maybe I need to kick the board to show the screen
<daniels> it should work ootb, but otoh one of the reasons we stopped working on exynos-drm is that every time we fixed it or did something, samsung would break it such that all commercially-available hardware on mainline would break
<HdkR> Nothing new then :P
<daniels> but unbuyable dev hardware running tizen worked great :\
<daniels> anyway, er, good luck
<daniels> back to bed
<HdkR> lol, I'm just here to fix t6xx. Samsung cruft can do w/e :P
<HdkR> black screen with kmscube after killing lightdm/xorg
<HdkR> ah, cursor vanishes though, interesting
<daniels> rk3288 is also t6xx iirc?
<HdkR> t7xx
<daniels> ah right, 768
<daniels> assuming exynos doesn't do implicit sync, do a few glFinish()es help at all?
<HdkR> Doesn't seem to
<daniels> also, if you just call gbm_bo_map() on the buffer you get back from gbm_surface_lock_front_buffer() do you see anything in there or just all black?
<HdkR> Dunno on that one :P
<daniels> might pay to dump the actual buffer content to narrow down a kms vs. gpu issue
<HdkR> good idea
<HdkR> I don't trust SFBD not being broken :P
<HdkR> alyssa: Do you still have an SFBD device around for testing?
<HdkR> kms is definitely getting the buffer. I memset it and I get output
<memeka> HdkR: yeah i get no errors, just black screen tho` :(
<HdkR> Alright, will look at this more later. Might be that the job handling just isn't happy still
hanetzer has joined #panfrost
<hanetzer> yus, rk3399-gru-bob is supported by u-boot. I'ma port the kevin nao :D
<alyssa> HdkR: See dmesg, probably says DATA_INVALUD_FAULT
pH5 has quit [Quit: bye]
NeuroScr has quit [Ping timeout: 240 seconds]
NeuroScr has joined #panfrost
BenG83 has joined #panfrost
raster has joined #panfrost
belgin has joined #panfrost
belgin has quit [Quit: Leaving]
<HdkR> alyssa: Nope. all faults have vanished :P
<rtp> ohh....
afaerber has joined #panfrost
raster has quit [Remote host closed the connection]
<alyssa> HdkR: ...Are you sure that's not just because you messed with core_req and now it's not submitting any work?
raster has joined #panfrost
BenG83 has quit [Quit: Leaving]
pH5 has joined #panfrost
stikonas has joined #panfrost
<Lyude> yay, serial adaptor for the RK3399 board I go sent just came in the mail today
<HdkR> alyssa: It could be the cause
raster has quit [Read error: Connection reset by peer]
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
afaerber has quit [Quit: Leaving]
cwabbott has quit [Client Quit]
cwabbott has joined #panfrost
* alyssa shrugs