alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - - Logs - <daniels> avoiding X is a huge feature
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
<alyssa> Holy moley
<alyssa> I'm doing some research into Mastodon and... yikes
<ente> hm?
<alyssa> ente: Let's just say I don't believe in federation anymore
<ente> I found you on mastodon a while ago but I've deleted my account since
<ente> is there new drama or something?
<alyssa> I've deleted it too
<ente> high five :D
<alyssa> ente: Not drama
<alyssa> I've been researching the accumulation of power in federated systems
<ente> oh, that's a good point
stikonas has quit [Remote host closed the connection]
<alyssa> Of the list "Email. XMPP, Tor" -- it's obvious how in each case, a major corporation or govt is at the center, even though they can technically be self-hosted
<ente> it was all centralized around and pawoo, similar to how email is centralized around and yahoo/outlook/gmx or whatever, and xmpp around
<alyssa> The obvious opposition is Mastodon, of course -- something that hasn't been 'hijacked' by the proprietary world
<alyssa> ente: It still is -- I ran the numbers
<alyssa> Just 3 instances account for 50% of user accounts
<alyssa> ^ I think that graph speaks for itself
<alyssa> Calling this a power law distriution would be too generous
<ente> thanks, I didn't give this too much thought so far
<ente> I mean, I would assume there are ways to work around it
<ente> like having a default "population cap" and defederating any instance that goes above it in the default codebase
<ente> was this the only reason for you to delete your account?
<ente> because I'm kind of lost on what the alternative would be. centralized services?
<alyssa> ente: This had nothing to do with why I deleted my account :P
<alyssa> If there's a population cap... people could operate multiple instances?
<ente> right, the truth is that there is nothing realistically preventing overly large instances managed by corporations
<ente> (or not corporations, that's not the point)
<ente> (but if there weren't pawoo already, nothing would prevent a "gmail of mastodon" to appear eventually)
<ente> alright, I'm gonna go curl up and play dead
rhyskidd has joined #panfrost
shenghaoyang has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
<alyssa> urjaman: My C201 is compiling Mesa now
<shenghaoyang> how long does it actually take for a clean build?
<alyssa> shenghaoyang: Haven't timed it in a while, since I rarely do clean builds
<shenghaoyang> *pondering whether to kill the scaleway VPS and save some cash*
<alyssa> Also, depends a ton of what machine you're building on (of course)
<alyssa> But -- not too bad if you don't need to do it too often
<alyssa> (Certainly under an hour on-target for what I've done. Usually a lot quicker.)
<alyssa> Usually I go off and do something else; unlike a lot of projects, mesa builds aren't terribly "needy"
<shenghaoyang> I have a C101PA, so it seems it's doable
<alyssa> shenghaoyang: Yeah, definitely not a big deal
<alyssa> It's _annoying_ but it's still way easier than cross builds ;)
<shenghaoyang> I'll just copy the half built tree from the VPS and hope I never have to rebuild
<alyssa> urjaman: You'll want in any case
<alyssa> urjaman: (To work around a pair of bugs)
<alyssa> shenghaoyang: Not sure why you spun up a VPS...?
<shenghaoyang> alyssa: scaleway has aarch64 servers, I took the easy way out and just got more grunt
<shenghaoyang> alyssa: PTSD from compiling things on a RPi...
<alyssa> shenghaoyang: Heh, well, we're WAY faster than the Pi ;)
<alyssa> urjaman: Build just finished. Latest upstream mesa is fine on the C201
<alyssa> Apply the patch I sent and you should be ok
<alyssa> Performance is quite nice, too
<alyssa> (Obviously it's a slower SoC than RK3399, but there's also WAY fewer pixels to push so frame rates are actually much better on some stuff)
<tomeu> hrm, I don't see a way of expressing an ORDER dependency with gpu-sched
<tomeu> guess the ABI should allow expressing it, and we can implement those later
paulk-leonov has quit [Ping timeout: 258 seconds]
paulk-leonov has joined #panfrost
<urjaman> i mean, now the error is something else :P
<tomeu> alyssa: with pre_deps it looks much nicer :)
Elpaulo has joined #panfrost
shenghaoyang has quit [Remote host closed the connection]
<tomeu> glmark2 on wayland it's looking quite nice, until the machine OOMs :)
shenghaoyang has joined #panfrost
shenghaoyang has quit [Remote host closed the connection]
shenghaoyang has joined #panfrost
pH5 has joined #panfrost
raster has joined #panfrost
shenghaoyang has quit [Remote host closed the connection]
<narmstrong> tomeu: add more memory !
<tomeu> was quicker to fix it :)
raster has quit [Remote host closed the connection]
<robertfoss> alyssa: I haven't posted them, but will do now
<robertfoss> alyssa: thanks for the reminder
raster has joined #panfrost
<tomeu> robertfoss: how it went?
<robertfoss> alyssa: it was ok, a bit underprepared.
<robertfoss> alyssa: I would have loved to demo the driver during the talk, but i wasn't even allowed to use my own laptop
<robertfoss> alyssa: in fact, I had to use powerpoint, which did terrible terrible things to the presentation template :p
<davidlt> robertfoss, was it recorded?
<tomeu> alyssa: is the whole of glmark2 working for you on kbase? I get some js faults at [desktop] blur-radius=5:effect=blur:passes=1:separable=true:windows=4:
<raster> glmark dies for me part way thru
<raster> i never see the jellies :)
<raster> (i can run that test on its own...)
<raster> and --off-screen is definitely b0rked
<tomeu> ok, then it's not the DRM driver's fault :)
<raster> nope - still kbase :)
<raster> not sure the new drm driver would be better... ? would it?
<tomeu> raster: I was asking alyssa because those errors are happening here with the DRM driver
<tomeu> I wanted to know if it could be related to the DRM driver
<tomeu> but if it happens with kbase, then I guess chances are they are not related to the DRM driver
<raster> well it may be my fault for having a weird system
<raster> or just being me
<raster> as things often break around me :)
BenG83 has joined #panfrost
<tomeu> raster: oh, I fixed today something that should get us further in glmark2
<tomeu> we were leaking big chunks of memory associated with a context
<tomeu> though with kbase those may be lazily allocated so maybe that wasn't your problem
<raster> tomeu: oh i noticed we had leaks too :) maybe it was the leaks actually
<raster> tomeu: i have noted apparent leaks...
<raster> what i'm wondering now tho is
<raster> panfrost_set_framebuffer_state() seemingly is implemented
<raster> but fbo's don't work....
<raster> porque...
<raster> hmmm
* raster digs
cwabbott has quit [Remote host closed the connection]
cwabbott has joined #panfrost
<gcl_> HdkR: Looks like Charbax to me
<tomeu> he was there last year
<hanetzer> l
cwabbott_ has joined #panfrost
<alyssa> robertfoss: Cool, thank you!
<alyssa> tomeu: No, -bdesktop and -bterrain are both broken right now. So js faults is expected on desktop, unfortunately
cwabbott has quit [Ping timeout: 258 seconds]
cwabbott_ is now known as cwabbott
<alyssa> robertfoss: Some bits for FBOs do work
<alyssa> I have patches sitting on the list from a few nights ago that get us further (so about half of glmark --off-screen works?)
<alyssa> It's just.. intensely buggy and needs a lot of code cleanup
pH5 has quit [Quit: bye]
pH5 has joined #panfrost
shenghaoyang has joined #panfrost
stikonas has joined #panfrost
shenghaoyang has quit [Remote host closed the connection]
raster has quit [Remote host closed the connection]
BenG83 has quit [Quit: Leaving]
BenG83 has joined #panfrost
paulk-leonov has quit [Ping timeout: 258 seconds]
raster has joined #panfrost
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
BenG83 has quit [Quit: Leaving]
raster has quit [Remote host closed the connection]
* alyssa wonders where everyone is today
<alyssa> tomeu: should probably be reverted -- it's fixed the right way in a patch I have upstream (er, it's on the list, still need to merge), so removing it here will just add one more merge conflict
<alyssa> tomeu: How does this work for e.g. a depth-only framebuffer?
<alyssa> (See glmark2-es2 -brefract)
<alyssa> Overall looks great tho :)