alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - - Logs - <daniels> avoiding X is a huge feature
gregordinary has quit [Quit: - Chat comfortably. Anywhere.]
vstehle has quit [Ping timeout: 252 seconds]
vstehle has joined #panfrost
_whitelogger has joined #panfrost
pH5 has joined #panfrost
_whitelogger has joined #panfrost
pH5 has quit [Ping timeout: 252 seconds]
pH5 has joined #panfrost
_whitelogger has joined #panfrost
pH5 has quit [Ping timeout: 258 seconds]
stikonas has joined #panfrost
TheKit has quit [Ping timeout: 258 seconds]
TheKit has joined #panfrost
TheKit has quit [Ping timeout: 250 seconds]
TheKit has joined #panfrost
_whitelogger has joined #panfrost
anarsoul|c has joined #panfrost
MoeIcenowy has quit [Quit: ZNC 1.6.5+deb1+deb9u1 -]
MoeIcenowy has joined #panfrost
<HdkR> Lyude: You've done kernelly bits. What sort of effort is required to get eGPU hotplugging to not hardlock the system? :P
yann has joined #panfrost
<Lyude> HdkR: reporting a bug :P, I haven't looked much at all at the eGPU stuff but I've always been curious how well it actually worked
<alyssa> Kek
<Lyude> HdkR: what's the problem exactly
<HdkR> Lyude: Kernel locks up on surprise hot unplug of eGPU on TB3 connection :P
<Lyude> HdkR:for which driver though?
<HdkR> ah, radeon
<Lyude> not amdgpu?
<HdkR> ...Whichever is the open one, I can never remember which is which
<Lyude> well there's two :p, that's why I asked. lsmod will tell you which one
<HdkR> amdgpu, had to check kernel logs
<Lyude> ahhhh, cool because if it was radeon I'd have no idea if we could even fix it :)
<Lyude> HdkR: I guess just show me the crash dump when it happens
<Lyude> erm, *backtrace
<HdkR> Gets reallllly angry
<HdkR> and since I have these wiggly ports, the idea of kernel crash means I don't wanna use it :P
<Lyude> those ports sound awful :(
<HdkR> 0/10, wouldn't purchase again
<Lyude> oh
<Lyude> May 17 18:10:28 ryanh-Blade kernel: [ 1298.825008] [drm:amdgpu_pci_remove [amdgpu]] *ERROR* Device removal is currently not supported outside of fbcon
<Lyude> sorry :(
<alyssa> Ook.
<Lyude> it's definitely a wip though
<HdkR> I guess that still matters if it has no display attached, running X on i965 situation, prime with the eGPU?
<Lyude> I guess so? You'd have to ask #radeon, but that's what I'd imagine...
<Lyude> wonder if it registers any display connectors
<HdkR> Not sure, piping the video back over worked surprisingly well with both GL and Vulkan. I was very surprised
<MoeIcenowy> alyssa: btw I think Allwinner provided 64-bit blob for Mali T720
<MoeIcenowy> (Allwinner H6
<MoeIcenowy> can it be used to develop unified 64-bit mode of panfrost userspace?
<MoeIcenowy> if you want, you can contact TL Lim, he is very willing to give Pine H64 (H6 board) to Panfrost developers
unoccupied has quit [Quit: WeeChat 2.4]
<alyssa> HdkR: I just pushed an updated panwrap verified to work with Dvalin. This is the modern panwrap, that just does a memory dump but no decode.
<alyssa> For decode, use the pandecode utility built with upstream mesa.
<alyssa> You should probably use this exclusively from now on =P
<alyssa> Starting Soon (TM), I should fix a few issues preventing pandecode from Bifrost decoding well (e.g. hooking up the disasm)
<alyssa> Got the disassembler hooked up :)
<HdkR> lol, 13 minutes is /very/ soon
<alyssa> What's your point
<HdkR> More soon than expected :P
<HdkR> Oh no! Rain!
<alyssa> HdkR: I thought you knew me well :(
<HdkR> sure sure. Different time domains is all
<alyssa> Hm?
<HdkR> My time domain has soon as longer than 13 minutes :P
<alyssa> HdkR: Like I said -- you don't know me well enough it seems :-(
<alyssa> =P
<HdkR> Where does panwrap live now? :P
<alyssa> Uncomment #define dvalin in include/
<alyssa> LD_PRELOAD the built thing, that'll dump to /dev/shm
<alyssa> So then just `pandecode /dev/shm`
<HdkR> Maybe I'll update eventually :D
<hanetzer> hehe, man you guys are hella cool :)
<HdkR> Hella cool would be day zero support for new architectures. Not quite as cool as Intel though :P
<alyssa> HdkR: I don't work quite -that- fast :)
<HdkR> We'd need real ISA documentation for starters. Shader compiler takes a boatload of time :P
<alyssa> Command stream too! :V
<HdkR> I haven't touched that yet, so for me it is zero time :P
<HdkR> Documentation on command stream changes would be cool as well
<hanetzer> HdkR: agreed, but that just makes the stuff you guys do even more hella cool
<hanetzer> "yeah, I wrote a driver for this without any vendor support, nbd"
<HdkR> I look up to the Nouveau team for coolness
<HdkR> Because a company actively working against the userspace is always interesting
<Lyude> HdkR: are you running a mainline kernel on your odroid btw?
<HdkR> no
<HdkR> I tried, got annoyed, gave up
<HdkR> I was basically copy and pasting memeka's commands from the hardkernel forum and it was failing
<HdkR> and 30+ minutes between kernel compilation times means iteration period is slow
<HdkR> Oh, apparently latest image includes OpenCL drivers. fancy
<HdkR> This might help me when I try again :P
<HdkR> I'll try the "regular" kernel first just to see if I can get an update working at all, then switch over to the nearly upstream branch
<HdkR> Cross-compiling will reduce iteration time as well when I can compile the kernel in a minute or something >>
<Lyude> i can try giving it a shot, i already have cross compiling setup for such things
<HdkR> Would be cool. I probably won't get to it this weekend after messing with RA and texture ops for a bit :P
_whitelogger has joined #panfrost
<alyssa> ^ List of low-hanging fruit tasks for someone wanting to contribute to Panfrost mesa (no prior graphics knowledge needed, just C)
<alyssa> Code cleanup stuff
<HdkR> :P
_whitelogger has joined #panfrost
<alyssa> HdkR: You know, something to point to when someone says "How can I help?", living list
<alyssa> More helpful than "fix some dEQP tests" :)
<alyssa> (Helpful to the asker, I mean)
<HdkR> right
jeez_ has joined #panfrost
jeez_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]