alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://freenode.irclog.whitequark.org/panfrost - <daniels> avoiding X is a huge feature
mias_ has quit [Ping timeout: 256 seconds]
stikonas has quit [Remote host closed the connection]
<icecream95> The format MALI_FORMAT_SPECIAL | 0xc is actually RGB233, not NV12
icecrea105 has joined #panfrost
icecream95 has quit [Ping timeout: 240 seconds]
icecream95 has joined #panfrost
icecrea105 has quit [Ping timeout: 256 seconds]
buzzmarshall has quit [Remote host closed the connection]
robert_ancell has quit [Ping timeout: 264 seconds]
cwabbott has quit [Ping timeout: 272 seconds]
davidlt has joined #panfrost
cwabbott has joined #panfrost
kaspter has quit [Ping timeout: 240 seconds]
kaspter has joined #panfrost
_whitelogger has joined #panfrost
kaspter has quit [Ping timeout: 264 seconds]
camus1 has joined #panfrost
camus1 is now known as kaspter
_whitelogger has joined #panfrost
<tomeu> alyssa: oh, cool idea, we didn't know much about those jobs back then
<tomeu> alyssa: guess that the header of those jobs gets also updated when the job starts and when it finishes
<tomeu> so if we don't see overwrites then that's probably not the problem
<tomeu> and instead it's some other IP block that is running wild
mias_ has joined #panfrost
anarsoul|c has quit [Quit: Connection closed for inactivity]
warpme_ has joined #panfrost
icecream95 has quit [Ping timeout: 250 seconds]
raster has joined #panfrost
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
stikonas has joined #panfrost
cwabbott_ has joined #panfrost
cwabbott has quit [Ping timeout: 246 seconds]
cwabbott_ is now known as cwabbott
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
robmur01_ has joined #panfrost
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
cwabbott has quit [Read error: Connection reset by peer]
kinkinkijkin has joined #panfrost
<kinkinkijkin> how's progress so far? mostly in general, but also specifically stability and ability to avoid X
cwabbott has joined #panfrost
yann|work has quit [Ping timeout: 264 seconds]
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
yann|work has joined #panfrost
kaspter has quit [Remote host closed the connection]
kaspter has joined #panfrost
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
yann|work has quit [Ping timeout: 264 seconds]
yann|work has joined #panfrost
kaspter has quit [Ping timeout: 250 seconds]
kaspter has joined #panfrost
indy has quit [Ping timeout: 265 seconds]
abordado_ has joined #panfrost
abordado_ has quit [Remote host closed the connection]
indy has joined #panfrost
cwabbott has quit [Remote host closed the connection]
cwabbott has joined #panfrost
<alyssa> icecream95: Hm interesting
<alyssa> tomeu: Mm..
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
<HdkR> kinkinkijkin: Panfrost stability depends on which Midgard GPU you're using and your use case. In general pretty stable though
<HdkR> Also which kernel you're running
<kinkinkijkin> ah
<kinkinkijkin> mali-T628 is of special interest to me, and latest LTS kernel
<kinkinkijkin> if you feel like letting me ask you instead of googling
<kinkinkijkin> and general 3D taskss
<HdkR> So ODROID-XU4 stuff? :)
<kinkinkijkin> yep
<kinkinkijkin> you know me, hdkr
<kinkinkijkin> :p
<HdkR> No idea what the LTS kernel version is, you need one of the 5.x versions
<kinkinkijkin> 5.4.x
<HdkR> Can't remember which number x though
<robmur01_> In general, a fair amount of stuff works on T628, but it's not quite as well supported as the newer models (T760+)
<kinkinkijkin> I believe it's 5.2 onwards? I've been following this passively for a while
<kinkinkijkin> ah
<HdkR> You could try adding the GPU to the whitelist to try things since your kernel is new enough
<kinkinkijkin> I might
<robmur01_> IIRC some kernel DRM scheduler stuff wasn't quite ironed out until around 5.5, but the relevant fixes should be/have been backported to 5.4 stable
<kinkinkijkin> need to set up an hdmi-accepting screen again first, since I sold my gaming monitor a while ago
karolherbst has quit [Quit: duh 🐧]
<HdkR> I'm also far too lazy to get one of my XU4s fully upgraded to test :P
<alyssa> Okay, time to start standalone vertex jobs from the standalone bi compiler
<HdkR> oop
chewitt has joined #panfrost
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
karolherbst has joined #panfrost
cwabbott has quit [Remote host closed the connection]
cwabbott has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
<alyssa> Uh how do I get an fd
<HdkR> fopen
<HdkR> :)
<alyssa> thanks :p
<HdkR> You didn't say what type of FD :D
<alyssa> drmOpenWithType it looks like
MistahDarcy has quit [*.net *.split]
bbrezillon has quit [*.net *.split]
Stary has quit [*.net *.split]
MoeIcenowy has quit [*.net *.split]
MoeIcenowy has joined #panfrost
Stary has joined #panfrost
<alyssa> Excellent, I have the device opened.
kinkinkijxu4 has joined #panfrost
<kinkinkijxu4> now on my xu4, hooked up into an hdmi>component adapter and displayed on my 720x480 tv which uh... i forget what it's called, it doesn't display the entire image
<HdkR> overscan
<kinkinkijxu4> yes that
<kinkinkijxu4> so once I have everything set up I'll try out some 3d stuff
<HdkR> https://gitlab.freedesktop.org/mesa/mesa/-/blob/master/src/gallium/drivers/panfrost/pan_screen.c#L761 Make sure to add your device to the whitelist, or make it always pass :P
<alyssa> why is my write value test failing
<alyssa> DIfing
<alyssa> it works on midgard
<alyssa> and it works in igt
<alyssa> so why do we DIF here?
<alyssa> tomeu: Something tells me we're getting closer to the mystery
<alyssa> Oh, nope
<alyssa> the BO I was writing to was just too small
<alyssa> somehow that led to a DIF
<alyssa> I.. don't know how that would happen
<alyssa> No, this is nearer.
<alyssa> robmur01_: Can confirm DIF happens from my in-mesa, out-of-GL demo
<alyssa> But IGT is fine
kinkinkijxu4 has quit [Quit: Leaving]
<alyssa> So I think we have something wacky going on deep in mesa
<alyssa> 9and the same test is fine on midgard -- write_value in both cases)
<alyssa> Wait! I can reproduce intermittent fails in igt too if I hammer it hard enough
<alyssa> Looks like I'll be doing some kernel hacking today..
<HdkR> Who needs scaling when you can just cap it out? :D
<alyssa> HdkR: What I wonder is if that's a performance issue, or an aml errata issue
tgall_foo has quit [Ping timeout: 264 seconds]
<anarsoul> oof indeed
Elpaulo has joined #panfrost
icecream95 has joined #panfrost
<daniels> brilliant.
icecream95 has quit [Remote host closed the connection]
buzzmarshall has joined #panfrost
davidlt has quit [Ping timeout: 240 seconds]
icecream95 has joined #panfrost
<alyssa> Okay, this is interesting
<alyssa> On runs that are successful, the magic loop executes once
<alyssa> on runs that won't be successful, it'll be multiple
<alyssa> if nothing else - confirmation this is the right track
<alyssa> narmstrong: ^^
<alyssa> Er no
raster has joined #panfrost
<alyssa> No, I stand corrected.
<narmstrong> Hmm why do you use the hardkernel kernel ??
<alyssa> narmstrong: I'm not, I'm trying to get mainline working
<alyssa> I have your PWR_KEY patch for probe but there are issues beyond that
<alyssa> and it looks like the aml/hk kernel is doing a lot more tricks..
<alyssa> (If you know what the rest of that file is about, I'd appreciate it)
<alyssa> looks like some amlogic-specific devfreq stuff though..?
tgall_foo has joined #panfrost
<narmstrong> Forget the devfreq stuff, it runs fine at max clock
<alyssa> Yeah, ofc.
<narmstrong> For the other stuff, i simplified it on the original kbase
<alyssa> link?
<narmstrong> Chewitt has a version working on 5.7
<narmstrong> I stopped updating it...
<alyssa> OK
<alyssa> Simpler indeed.
<narmstrong> I added some pm callbacks in the kbase core, but this works perfectly, we use it on our upstream based AOSP port
<alyssa> Alright
<alyssa> so the bug I'm up against is *that* subtle
icecream95 has quit [Ping timeout: 264 seconds]
icecream95 has joined #panfrost
raster has quit [Quit: Gettin' stinky!]
<icecream95> alyssa: I don't believe you that native NV12 ever worked, especially considering that MESA_FORMAT_NV12 doesn't exist
<anarsoul> just implement R8/R8G8 and mesa will handled missing yuv formats for you
<icecream95> That's exactly what it's doing, but there is allegedly native support for >=t760
<alyssa> icecream95: Hm, maybe kmscube got lowered to that format somewhere. Fair enough, r-b on that patch too then :)
<alyssa> icecream95: Native support for YUV -- not sure about NV12 specifically vs planar vs etc
robmur01_1 has joined #panfrost
robmur01_ has quit [Ping timeout: 258 seconds]
robmur01_ has joined #panfrost
robmur01_ has quit [Client Quit]
<icecream95> alyssa: panfrost_is_format_supported returns false for PIPE_FORMAT_NV12
<anarsoul> alyssa: mesa lowers yuv formats to multiple planes of r8 if these aren't supported natively
<anarsoul> and does colorspace conversion in shader
robmur01_1 has quit [Ping timeout: 246 seconds]