<alyssa>
Lyude: cwabbott: We need to get copyright stuff sorted sooner than later for sending this stuff upstream
<alyssa>
atm basically everything is GPL headered, but mesa's going to want MIT (or is it BSD? I forget) so we'll need to relicense
<hanetzer>
MIT iirc
<alyssa>
Panwrap aside, I've gotten rid of the "Panfrost Community" stuff to try to replace with actual authors to make this easier
<alyssa>
So, again outside panwrap:
<alyssa>
cwabbott: You have changes to include/mali-job.h
<Lyude>
alyssa: uhh-I'd leave that there, we don't really need to get rid of that
<Lyude>
alyssa: nouveau has the same copyright notice
<urjaman>
MIT.... and i'm late as f
<urjaman>
also, TIL.
<Lyude>
(also: you have my explicit permission to change the license of any code I've contributed to/written to whatever you need)
<alyssa>
Lyude: Even before said change it was a hodge podge of who-wrote-what-where
<Lyude>
i mean, we have git history don't we?
<Lyude>
tbqh: it might be a better idea to just get rid of the license headers entirely and just use SPDX ones
<alyssa>
Stuff has moved between repos no fewer than 11 times
<Lyude>
true, but the only people who ever would have really contributed thus far are the ones in this room
<alyssa>
that's fair
<Lyude>
so just ask everyone here and I think we should be fine
<alyssa>
And uh, to be totally honest, some of it's a vanity thing for me \shrig/
<alyssa>
*shrug
<Lyude>
yeah, but they may want SPDX headers anyhow, I'm not sure if mesa has moved over to using those
<Lyude>
I know the kernel will want that at least
<alyssa>
That's an easier problem to solve :p
<alyssa>
(Vanity meaning naming vs generic "The Authors")
<alyssa>
freedreno is explicitly "Rob Clark"
<alyssa>
Etnaviv is "Etnaviv Project" with an authors listing at the bottom
<alyssa>
videocore is "Broadcom" but that's totally corporate so that's diff.
<Lyude>
alyssa: i'm not terribly picky about where my name is; it's nice to have in places but if you forget it somewhere i'm not going to complain too much
<Lyude>
especially seeing as my face is the thumbnail for the canonical panfrost presentation at XDC
<alyssa>
Hehe, alright :p
<Lyude>
which is realllllllly fucking funny to just see keep popping up on random forums
<alyssa>
Hehe
<alyssa>
Sorry 'bout that ;)
<Lyude>
LOL it's ok
<Lyude>
it's not a problem at all, that's like the third presentation i've done
<alyssa>
I guess your face would have been safe if we didn't do such a spectglamolous presentation
<Lyude>
although most people won't associate the first two presentations I did with me
<Lyude>
but i don't mind either way
<hanetzer>
alyssa: 'staging' is the correct branch for now, right?
* urjaman
heard 'master' :P
<urjaman>
i didnt check to lol
<urjaman>
yeah master has been touched 16 minutes ago, staging 7 months ago :P
<hanetzer>
gotcha.
<urjaman>
alyssa: maybe point the default branch to master on mesa? ..
<Lyude>
alyssa: also btw, this is minor but someone will likely say something when we submit patches, we should probably cleanup the indenting in the meson.build files as well. I'll try to remember to do that when I go to submit the pull requests to make us use pkg-config for building against X
<alyssa>
urjaman: oops forgot to do that
<alyssa>
Lyude: indentation everywhere has to be changed, just need to run astyle or something
pH5 has quit [Ping timeout: 272 seconds]
<hanetzer>
src/gallium/drivers/panfrost/pan_screen.c:583:1: error: no return statement in function returning non-void [-Werror=return-type]
* alyssa
looks
<alyssa>
Oh, interesting. I guess the different compiler versions decide whether that's an error or... silent
<alyssa>
hanetzer: Fixed
<urjaman>
nah, that's all based on compiler flags
<hanetzer>
yep. unsure why -Werror is getting passed but it is (I just cloned the current mesa-9999.ebuild and added panfrostian use flags to it
<hanetzer>
)
<urjaman>
like atleast i think any recent gcc can find that :P (if the warnings are enabled... then if -Werror is set, it's an error.)
<alyssa>
urjaman: We're on the same flags
<alyssa>
The issue is that function body was "assert(0)"
<alyssa>
My compiler interpreted that as an implicit "return $DONTCARE" I guess
<alyssa>
hanetzer's did not
<alyssa>
Werror on that is a mesa thing, I've got used to it :p
<urjaman>
now i'm guessing debug vs non-debug build
<alyssa>
gcc version 8.2.0 (Debian 8.2.0-9)
<alyssa>
you?
<alyssa>
Is it weird that 3 space style makes me anxious?
<alyssa>
("...what?")
<alyssa>
I wonder if I can vim to make it look bigger
<urjaman>
3 space? wat.
<alyssa>
It's a mesa thing
<urjaman>
(i thought that was a joke style... somewhere uses it?)
<alyssa>
Most of mesa, yeah
<urjaman>
i'm somewhat amazed and disgusted at the same time
<urjaman>
i had to check, browsed to a random source file ... i see-e...
<alyssa>
seriously how do I make the 3 spaces look like more than three
* urjaman
doesnt know enough about vim things to know...
<alyssa>
#vim advices me to just use 4-space tabs (or hard tabs for that matter) for panfrost and let mesa deal with it :^)
<urjaman>
running everything through astyle both ways really isnt workable for contributing either
<urjaman>
hard tabs and "just set your tab size to 3 if you want it" :P
<alyssa>
Yeah
<alyssa>
So I realise that my "nice defaults" from vim that are so comfortable to me... is 8 columns, not 4
<alyssa>
I'm not going to subject people to that so I guess hard tabs are the compromise between "Alyssa's visual obsessions" and "literally everyone else"? :P
* urjaman
uses 8 quite often ...
<alyssa>
I mean, Linux is 8-spaces and it's arguably the most successful free sw project in history
<alyssa>
Coincidence? I think not1
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
cwabbott has quit [Client Quit]
cwabbott has joined #panfrost
<alyssa>
urjaman: spot interrogation, do you display line numbers in your editor? ;P
<urjaman>
alyssa: right now, yes
<alyssa>
k
<alyssa>
Is it atypical I don't?
<hanetzer>
prolly
<alyssa>
I'm out of touch with what people do these days
<alyssa>
It contributes too little value to justify the clutter imo
<urjaman>
for a long time i used to not do it, but instead have continouous display of the current line somewhere (but not line numbers on all lines)
<alyssa>
I mean, I have that
<urjaman>
and yeah i'm a weirdo who codes in either SciTE or nano :P
<alyssa>
but I don't pay attention to that. I mostly don't know how to disable it
<urjaman>
(and no, not nano as you get it shipped... nano with an "enable all the fancy things" nanorc :P)
<hanetzer>
alyssa: found another -Werror issue. pan_context.c:translate_tex_filter:640:1 error: control reaches the end of non-void function [-Werror=return-type], pan_context.c:translate_tex_wrap:631:1: error: control reaches end of non-void function [-Werror=return-type], pan_context.c:panfrost_translate_texture_swizzle:620:1: error: control reaches end of non-void function [-Werror=return-type];
<hanetzer>
midgard_compile.c:midgard_tex_format:1296:1: error: control reaches end of non-void function [-Werror=return-type], midgard_compile.c:midgard_ra_select_callback:1473:1: error: control reaches end of non-void function [-Werror=return-type], midgard_compile.c:quadword_size:2208:1: error: control reaches end of non-void function [-Werror=return-type], midgard_compile.c:component_from_mask:1716:1: error:
<hanetzer>
control reaches end of non-void function [-Werror=return-type]
<hanetzer>
ouch. didn't realize how big that would get :P
<hanetzer>
:set number relativenumber
<hanetzer>
so, only the current line's number is shown, and all other lines are shown as an offset to the current line.
<hanetzer>
alyssa: also, sorry if it just seems like I'm popping in just to pick apart the code. I'm setting it to compile via emerge with more or less stock gentoo config and doing homework untill something breaks :)
<urjaman>
... that sounds even less useful than the line numbers (to me), because i mostly use them for locating error messages, and knowing that i've written too many things into one file :P
<alyssa>
hanetzer: no worries, this is helpful and will save me later flaming! :P
<alyssa>
I want to get this code upstreamed asap so
<hanetzer>
urjaman: heh. I use those to do stuff like d14j to delete downward, or yank, or whatever.
<hanetzer>
or sort-motion, gs14j
<alyssa>
urjaman: "knowing that i've written too many things into one file :P" Pff
<alyssa>
hanetzer: Pull
<urjaman>
i like to keep things at ... well, practically under 2000 lines per file :P (the mental thing is that significantly over 1000 is when i start going "hmm, mhh, ehh, umm..." but i rarely do anything about it lol)
<alyssa>
Oddly that doesn't bother me nearly as much as lots of little files
<alyssa>
I don't mind living in a file
<hanetzer>
alyssa: doing :)
<urjaman>
i get lost in files with like 5k lines or more
<alyssa>
Don't you like living in the code? :P
<hanetzer>
code and coursework (some of which *is* code) is my life rn
<urjaman>
i've kinda burned out myself with work code a year (or two) ago, so ... (or that's what it felt like atleast)
<urjaman>
so i reallllly dont want to end up living in any code lol
<hanetzer>
alyssa: sw_helper.h:sw_screen_create_named:66:16: error: implicit declaration of function `panfrost_create_screen`; did you mean `pipe_panfrost_create_screen`? [-Werror=implicit-function-declaration], midgard_compile.c:midgard_tex_format:1298:1: error: control reaches end of non-void function [-Werror=return-type]
<alyssa>
Oy
<alyssa>
hanetzer: the latter one I just fixed
<alyssa>
Someone had the same issue as the first before and I'm trying to remember what the issue was
<hanetzer>
kk.
<urjaman>
i had the first issue (i think), but i never figured out why... i thought you'd have fixed it loong ago :P - i just disabled the code in question...
<Lyude>
alyssa: that was probs me!
anarsoul|2 has quit [Ping timeout: 268 seconds]
<alyssa>
urjaman: I'm just trying to figure out why it works fine here
<alyssa>
and I'm trying to figure out why my setup is special
<alyssa>
my git status is clean afaict so
<hanetzer>
well, it is a warning turned into an error.
<hanetzer>
so, you may not be -Werror'ing
<hanetzer>
I don't know why I am, but like I said, I just use the gentoo ebuild with a new video_cards_panfrost use flag to flip the meson option :)
<alyssa>
hanetzer: Again, I don't even get a warning. If I did, I would've fixed it :p When I do get that notice at all it's an error per the flag
<hanetzer>
¯\_(ツ)_/¯
<hanetzer>
what compiler you on?
<urjaman>
04:14 <@alyssa> gcc version 8.2.0 (Debian 8.2.0-9) |
<urjaman>
04:14 <@alyssa> you?
<urjaman>
... ehh, i've done better pastes :P
<alyssa>
urjaman: Pardon?
<urjaman>
hanetzer asked a question you had already answered... (and i copied the answer including the border inserted by screen since i was using it not-fullscreen, ugh)
<hanetzer>
hmm?
<urjaman>
hanetzer: ^^ the freaking compiler version you asked for, was already provided about 2 hours before you asked :P
<urjaman>
(and you never answered the question on what compiler gentoo is using ... i assume it might be one revision older, dunno tho... gentoo tended to be slightly conservative on the compilers way back when i used it...)
<urjaman>
and no, i dont know why i am awake at 8 in the morning with no sleep
<alyssa>
I'm so confused
<alyssa>
oh i see
<hanetzer>
urjaman: oh gotcha. I was a bit confused with you 'answering'
<hanetzer>
the gentoo version i'm using is 8.2.0-r4 p1.5
<hanetzer>
what sort of fps should I be getting with glxgears if I actually have panfrost working right?
BenG83 has joined #panfrost
BenG83 has quit [Remote host closed the connection]
BenG83 has joined #panfrost
BenG83 has quit [Ping timeout: 268 seconds]
BenG83_ has joined #panfrost
BenG83_ has quit [Read error: Connection reset by peer]
BenG83 has joined #panfrost
anarsoul has quit [Ping timeout: 268 seconds]
<urjaman>
s/glxgears/es2gears/
anarsoul has joined #panfrost
BenG83 has quit [Remote host closed the connection]
BenG83 has joined #panfrost
BenG83 has quit [Quit: Leaving]
BenG83 has joined #panfrost
<HdkR>
should be >100fps at 720p
<HdkR>
blob struggles with 60
afaerber has joined #panfrost
<thefloweringash>
picking through my CI builds, panfrost @ 3113f26 works, but with the new mesa @ 283a2a86d3, my es2gears no longer move
<HdkR>
I really need to get that CI system running
<thefloweringash>
Mine only builds, and doesn't naturally extend to proper testing. It's convenient to quickly see what works and what doesn't, but I haven't managed to find the time to actually contribute anything. =\
<HdkR>
Sadly this business trip lasts until sunday. So it will eat two weekends in all D:
TheKit has quit [Ping timeout: 252 seconds]
TheKit has joined #panfrost
BenG83 has quit [Quit: Leaving]
BenG83_ has joined #panfrost
BenG83_ has quit [Read error: Connection reset by peer]
tgall_foo has quit [Quit: Leaving]
BenG83 has joined #panfrost
<alyssa>
cwabbott: Alright, cool, thank you!
<alyssa>
thefloweringash: doesn't... move?
<alyssa>
I was messing with viewport stuff and I think that broke a lot since my gears don't show up right now, pah
<alyssa>
will sort out this evening
<thefloweringash>
They show up after I resize the window.
<alyssa>
Yeah the stuff that changed since then is viewport
<alyssa>
I guess I'll revert that code for now until I can sort this the right way
<alyssa>
hanetzer: and yeah, minimally "faster than llvmpipe" :P
<alyssa>
hanetzer: thefloweringash: Pushed fixes for both of your stuffs
Dr_Who has joined #panfrost
Dr_Who is now known as tgall_foo
pH5 has joined #panfrost
<hanetzer>
so, how can I know for certain I'm actually using panfrost and not llvmpipe ?
anarsoul|2 has joined #panfrost
<urjaman>
well, the two windows should make it quite obvious
<urjaman>
and the debug info...
<urjaman>
(and you can ask es2info with your calling style if you wish...)
afaerber has quit [Quit: Leaving]
<HdkR>
hanetzer: es2_info? :D
<Lyude>
hanetzer: jfyi: the panfrost driver doesn't work with the latest kernel driver at all
<Lyude>
you will need a kernel driver that still supports UMP; the old ARM mali UAPI
<Lyude>
which is the big reason i want to focus on that first
<hanetzer>
hrm. well, looks like I'm on llvmpipe for some reason. mali_kbase.ko is loaded and I've compiled mesa/master from the panfrost repo, system install even
<hanetzer>
for reference my gfx login path is like so: systemd starts lightdm, I log into mate from there
<Lyude>
I, think that would work? I have no idea ._.
<Lyude>
maybe try a basic startx test with this to eliminate the possibility it's got something to do with your gfx login path
<hanetzer>
should mate be ok, or should I go with something else for now?
<Lyude>
it's definitely finding panfrost; but yeah the dri errors come right after the dri file gets loaded for panfrost
<Lyude>
hanetzer: anything should be fine
<Lyude>
i've just been using X -retro
<Lyude>
if that doesn't work, I'd make sure it doesn't have anything to do with having dri overridden for the whole system
<Lyude>
i probably can check that on amethyst
<Lyude>
oh right-I forgot I don't have things setup yet so i can do vpn to my house so I can mount amethyst's sysroot over nfs, which i need to do the cross compile D:
<hanetzer>
my machine's build is self-hosting :P
<Lyude>
i mean you can do local prefixes without cross compiling
<hanetzer>
yeah.
<hanetzer>
so X -retro will launch a really old-school de/thing?
<urjaman>
well, it doesnt launch anything but X
<urjaman>
but it gives the stippled pattern so you can see that X launched
<hanetzer>
so how do you launch es2gears/etc from that?
<urjaman>
dont ask me :P (except yeah i'd just get a second VT and set DISPLAY, or run it with & and set DISPLAY ... )
<hanetzer>
and I shouldn't need anything like xf86-video-blah right? just stockish xserver and mesa with the panfrost patches?
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
<Lyude>
hanetzer: just set DISPLAY in the environment for another terminal and then run what you want there
<Lyude>
X should say what display it defaulted to; and if not you can also just specify it manually on the cli like X -retro :0
cwabbott_ has joined #panfrost
cwabbott has quit [Ping timeout: 252 seconds]
cwabbott_ is now known as cwabbott
<hanetzer>
Lyude: once my rebuild finishes I'll give straight X and startx a try, no lightdm
<Lyude>
hanetzer: alright; lemme know if you still have issues
<Lyude>
if things work correctly with the newer kernel driver; then you should be expecting to see an abort9)
<Lyude>
*abort()
<hanetzer>
I just can't kill it right now because I'd lose valuable compile time on this weak rig (really need to get that usb ethernet adapter)
<Lyude>
ahh, makes sense
<Lyude>
you see why it's taken me so long to get everything setup locally :)
<hanetzer>
all I've got left is net-libs/webkit-gtk-2.22.2 :)
<Lyude>
btw: if you end up switching to cross compiling or something like that give me a poke; you will possibly need meson/panfrost patches from me to make it work
<hanetzer>
prolly use distcc :)
<Lyude>
eek
<Lyude>
god speed
* Lyude
is sad her icecream setup broke out of nowhere :(
<Lyude>
so now I have to do all my compiling in the office on my blazing fast, dual core ultrabook w/ ht (it's not fast)
<hanetzer>
yeah. I need to look into distcc :P
<hanetzer>
erm, s/distcc/icecream/
<hanetzer>
yeah. my main machine is plenty powerful (ryzen 7 1700)
<HdkR>
It can always be faster. Epyc 2 was announced with 64 cores per CPU ;)
<HdkR>
per silicon substrate?
<HdkR>
Per socket, there we go
<hanetzer>
oof. 2hr estimated finish time.
<Lyude>
*roblox oof*
<urjaman>
it had been so long since i last rebooted the C201 that i had forgotten which password it uses...
<urjaman>
but yay 4.19.0 seems to run :P
<hanetzer>
yeah. that's what I'm on
<hanetzer>
also ccache is useful
<hanetzer>
so, should this be using dri{1,2,3}? something else entirely?
<hanetzer>
HdkR: yeah, but if I'm going to buy some massively powerful machine like that I'll just save and get a talos :)
<HdkR>
hanetzer: If the Talos could do a fresh compile of my workflow in less than this `real 1m54.305s` Then I would strongly look at it
<HdkR>
:D
<hanetzer>
well, considering it has two sockets maxing out at 22 cores with 4 threads a core, it could be pretty powerful ;P
<HdkR>
Although that's also a $12,000 machine with two of those D:
<hanetzer>
but muh foss tho
<HdkR>
lol
cwabbott has quit [Quit: cwabbott]
<HdkR>
Considering I'm pairing it with a Turing GPU, I'm terrible :)
cwabbott has joined #panfrost
cwabbott has quit [Client Quit]
cwabbott has joined #panfrost
<hanetzer>
do I need gles1/2 ?
<HdkR>
You'd want to provide it anyway
<Lyude>
hanetzer: I don't know; I don't think you can use panfrost as the actual ddx
<hanetzer>
is there some xorg.conf bits I should set?
<Lyude>
you just need to make sure it does the thing with client application
<Lyude>
hanetzer: no; you can't use it as a ddx
<Lyude>
ddx = xf86-video-whatever
<hanetzer>
so, I need what, fbdev, kms, ?
<Lyude>
hanetzer: so: the way panfrost renders things on screen (correct me if I'm wrong alyssa ) atm is kind of hackish (which is one of the other reasons we're doing the kernel work stuff): we make the GPU render into an X framebuffer, which is provided by the X server
<Lyude>
so you really are limited to just using it with things that render into X, not X itself
<Lyude>
using it as a DDX would mean we'd have to teach X how to allocate fbos using the mali driver
<Lyude>
or eventually once we have a real kernel driver: just using modesetting through the kernel driver which would support GBM
<urjaman>
hanetzer: yeah it should run with any X driver, since it currently just makes a window with Xshm, but i wouldnt even dare trying to run it as a system-wide mesa ...
<hanetzer>
xshm. maybe I lack that.
<urjaman>
nope, that's a feature of pretty much every X driver
<Lyude>
DRI2 error, then MALI_IOCTL_GET_VERSION fails
<Lyude>
after that is an abort()
<hanetzer>
and this is with the r27 mali driver?
<Lyude>
hanetzer: r26, but I wouldn't expect different behavior from r27
<hanetzer>
and what command is this? it doesn't look like whawt I got
<Lyude>
I have X -retro :0 on one terminal, other DISPLAY=:0 LD_LIBRARY_PATH=:/home/lyudess/prefixes/panfrost-mesa/lib64 es2_gears
<hanetzer>
ah
<Lyude>
I need to thank chewitt again for getting me one of these amlogic boards, this thing is super useful now that I've got it running moderately fast w/everything working
<hanetzer>
I'm rather looking forward to getting a kevin
<hanetzer>
does libpanwrap.so play into this at all?
<Lyude>
hanetzer: sort of, let me show you the inside of prefixes/panfrost-mesa