alyssa changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - - Logs - <daniels> avoiding X is a huge feature
<alyssa> shadeslayer: Are we sure we need to submit outstanding jobs on context destroy?
<alyssa> I don't see Freedreno doing it..?
<alyssa> or iris
gcl has quit [Ping timeout: 258 seconds]
davidlt has quit [Ping timeout: 245 seconds]
vstehle has quit [Ping timeout: 255 seconds]
gcl has joined #panfrost
<tomeu> alyssa: any comments on the wallpaper code?
<tomeu> I think we somehow are missing some of the stuff that should be rendered right after the wallpaper
<tomeu> for example, weston's background image
<tomeu> in kmscube it isn't visible with the current code, but we were missing the first vertex of a cube side when the wallpaper jobs were created after the app ones and then moved to the start
vstehle has joined #panfrost
Elpaulo has quit [Read error: Connection reset by peer]
Elpaulo has joined #panfrost
<tomeu> alyssa: if you have time, would be nice if you could check if there's anything interesting going on in the blob when submitting the first vertex job right after the wallpaper
stikonas has joined #panfrost
* HdkR loads up panfrost
<HdkR> ...What was I working on
<HdkR> alyssa: btw, you have a warning in midgard.h where you're assigning the location of atomic_umax twice. Line 550 and 551
stikonas has quit [Remote host closed the connection]
<HdkR> Interesting. Looks like Bifrost has a conditional discard instruction
<tomeu> oh, i think I know what's going on
<tomeu> when uploading the vertex data for the wallpaper, we're overwriting some of the vertex data that the app uploaded just before
yann has quit [Ping timeout: 258 seconds]
<tomeu> hmm, maybe I was incorrectly printing the contents of the vertex buffer at emission time
<tomeu> that calculates a size of the vertices to be uploaded of 56 (in my case), instead of 64 (4 vertices, each with 4 floats)
<tomeu> then when the vertex buffer for the wallpaper would be uploaded, it would be placed from offset 56 instead of 64
<tomeu> but that seems to be quite deliberate, and if it was wrong I would expect massive breakage everywhere in mesa
<ente> alyssa: seems your site's certificate has expired?
Elpaulo has quit [Ping timeout: 244 seconds]
Elpaulo has joined #panfrost
vstehle has quit [Ping timeout: 246 seconds]
vstehle has joined #panfrost
hlmjr has joined #panfrost
herbmillerjr has quit [Ping timeout: 244 seconds]
yann has joined #panfrost
raster has joined #panfrost
raster has quit [Ping timeout: 245 seconds]
raster has joined #panfrost
tlwoerner has quit [Quit: Leaving]
tlwoerner has joined #panfrost
afaerber has quit [Quit: Leaving]
afaerber has joined #panfrost
vstehle has quit [Read error: Connection reset by peer]
vstehle has joined #panfrost
<alyssa> ente: Oh, whoops, right, meant to fix that uhhh
<alyssa> ente: Try now..?
<alyssa> tomeu: Not sure....
Elpaulo1 has joined #panfrost
<shadeslayer> alyssa: but vc4 and v3d do it, so I'm not sure what to follow here
<alyssa> shadeslayer: Welcome to the 'joy' of mesa dev :-(
<alyssa> HdkR: Fixed.
Elpaulo has quit [Ping timeout: 246 seconds]
Elpaulo1 is now known as Elpaulo
<shadeslayer> alyssa: so how does one find out? just trial and error?
<alyssa> shadeslayer: Usually by this point I would've whined in #dri-devel, and somebody would be nice enough to give pointers :-)
<shadeslayer> heh
<HdkR> alyssa: Cool :)
<alyssa> shadeslayer: And you might hear "Oh, whoops, that's a bug in iris/freedreno" or maybe you'd hear anholt say "Yeah, that's probably unnecessary but just in case..?"
<shadeslayer> hahaha
<alyssa> shadeslayer: I'll be honest.. the command stream code (the pan_*.c files) is really bad. Not really sure how to fix that.. :(
<alyssa> Though the compiler isn't terrible, code quality-wise
<shadeslayer> alyssa: heh, I don't think it's just you, I'm reading the modetest code from libdrm and I'm having trouble there too
<alyssa> >_<
<shadeslayer> it's just .. not nice
<alyssa> pan_context.c is just
<alyssa> Not my best work, ok
<shadeslayer> :D
<shadeslayer> alyssa: not to mention that modeset just segfaults if you supply anything but a valid string
<shadeslayer> +for the modes
<shadeslayer> alyssa: so, perhaps if we just test out my parenting patch and see if it works :)
raster has quit [Remote host closed the connection]
stikonas has joined #panfrost
Elpaulo has quit [Quit: Elpaulo]
janrinze has quit [Ping timeout: 252 seconds]
yann has quit [Ping timeout: 268 seconds]
raster has joined #panfrost
pH5 has joined #panfrost
<alyssa> shadeslayer: Resend the patch?
<shadeslayer> alyssa:
* alyssa tries
<alyssa> Segfault
<alyssa> shadeslayer:
<shadeslayer> wooo'
<shadeslayer> I should acquire hardware
raster has quit [Remote host closed the connection]
<ente> alyssa: works!
afaerber has quit [Remote host closed the connection]
unoccupied has joined #panfrost
Elpaulo has joined #panfrost
<alyssa> ente: Thanks :)
stikonas has quit [Remote host closed the connection]
gregordinary has joined #panfrost
afaerber has joined #panfrost
hlmjr has quit [Quit: Konversation terminated!]
herbmillerjr has joined #panfrost
herbmillerjr has quit [Quit: Konversation terminated!]
<HdkR> alyssa: Lyude: Do you think it is possible to have a bifrost demo up and running at XDC 2019? :P
<Lyude> HdkR: that would be nice!
<Lyude> dunno if I can really speak on how possible it is though
<HdkR> Nice yes, I mean more of physically possible to get the code to the point that it can do it :D
<Lyude> I can try, I don't think I'll be getting any work time on this though for a while so I'll have to work on weekends to do it
<HdkR> hehe
<HdkR> This weekend for me is RA bug fixing, Texture IR op messings, and if I get enough time, messing with the ODROID-N2 some more to try and get the DRM on it. Failing that I'll be hooking kbase back up
<alyssa> HdkR: If I had a board in my possession right now, yes, very possible ...
<alyssa> Speaking of, it's mid-May :8
<HdkR> alyssa: Is that saying that you actually want a board now? :P
<Lyude> i still do not understand amd's obsession with 24-width tabs on line continuations
<HdkR> wow
<Lyude> yeah amd code is wild
<HdkR> I'm becoming comfortable with the two space lifestyle
<alyssa> HdkR: Didn't I say that a few weeks ago..? Or is my memory failing..? =P
<Lyude> they also always do this thing where instead of declaring a pointer to a subelement of a struct (say, struct *foo = &bar->baz->bay->bax;) they just type the full thing eery time
<Lyude> *every
<Lyude> even->if->theres->this->many->subelements
<HdkR> That sounds like an annoying amount of %s/ALL THE THINGS/ARE NOW THIS/g
<Lyude> HdkR: I would but supposedly we somehow came to the understanding that DC doesn't follow the code style the rest of the codebase does (which considering how we pull in acpi, is frankly bullshit, but I digress...)
<HdkR> Begrudging understanding
<HdkR> I can understand that
<Lyude> yeah, i'd understand a bit more if they like, actually stuck to their own codestyle
<HdkR> haha
<HdkR> I have that problem with variable naming
<HdkR> `[Freedreno] XDC 2019:` Side funny. XDC 2019 is the year of the Freedreno
<alyssa> Pretty much.
stikonas has joined #panfrost
stikonas has quit [Remote host closed the connection]
pH5 has quit [Ping timeout: 246 seconds]
anarsoul|c has quit [Quit: Connection closed for inactivity]