00:09
archetech has quit [Quit: Konversation terminated!]
00:19
<
macc24 >
alyssa: sometimes i wish kevin was smaller and had a 1280x800 screen :/
00:19
<
macc24 >
or had bifrost gpu
00:19
<
alyssa >
macc24: It's called bob.
00:20
<
macc24 >
alyssa: well i only have kevin, and i only use it for drawing, cadmium cross-device support and performance comparisons
00:21
<
alyssa >
---Why is the mesa build broken?
00:21
<
macc24 >
printing for 5h is forever...
00:21
<
HdkR >
mesa build is broken?
00:25
<
HdkR >
looks like it is fine on a fresh upstream/master for me
00:25
<
HdkR >
With an everything build, clang, release
00:26
<
alyssa >
I'm not jealous of your build speed or anything ;o
00:26
* alyssa
still needs to do a clean mesa build on m1 just to feel envy :p
00:26
<
HdkR >
Testing with gcc now
00:27
<
alyssa >
I rm'd my build and it's still unhappy
00:27
<
alyssa >
so unless it's a ccache bug?
00:27
<
HdkR >
Debug build maybe?
00:27
<
HdkR >
gcc release passed, checking debug
00:31
<
HdkR >
debug passed
00:31
<
HdkR >
gcc version 10.2.0
00:35
<
alyssa >
HdkR: culprit was an accidentally defined empty nir.h in src/compiler
00:35
<
alyssa >
because I can't vim apparently
00:36
<
alyssa >
also, question for someone who knows:
00:36
<
alyssa >
top shows "MiB Swap: 0.0 total, 0.0 free, 0.0 used."
00:36
<
alyssa >
yet I also see activity from kswapd
00:37
<
HdkR >
Might be worth adjusting swappiness
00:37
<
urjaman >
.... for what swap ...
00:38
<
macc24 >
alyssa: just yeet swap out of your kernel
00:39
<
HdkR >
Swap is nice to have for the case of a bad actor stealing all ram once in a while
00:39
<
alyssa >
but I will say
00:40
<
alyssa >
the fact that was handled gracefully enough (a small slow down and then back to normal) was...
00:40
<
alyssa >
Well, I shouldn't say impressive since that's the bare minimum
00:40
<
macc24 >
HdkR: earlyoom?
00:40
<
urjaman >
yeah i actually have 512M of swap on my C201, i usually dont run into OOM kills (and when they happen, they happen... reasonably)
00:40
<
alyssa >
but I was tensing up ready for the system to completely freeze up
00:40
<
HdkR >
macc24: Maybe I'd prefer a bit of slowness rather than murdering the app though :P
00:40
<
alyssa >
and it was perefctly fine.
00:40
<
macc24 >
HdkR: every time my ram got eaten on any of my machines i had to reboot
00:40
<
alyssa >
Looks like yeeting madvise might not be so tragic
00:41
<
urjaman >
oh geeze early oom ... that was frustrating on my Jolla when i was trying to use it as a linux machine (as opposed to a phone :P)
00:41
<
HdkR >
Definitely depends on machine. If you start swapping to eMMC or gods...microSD, you're dead
00:41
<
macc24 >
HdkR: altough the ssd on my desktop is still seen by linux as an ide hard drive
00:41
<
macc24 >
that might be part of problem xD
00:42
<
urjaman >
well atleast it's configurable too :P
00:42
<
macc24 >
HdkR: duet's emmc is not that slower than ssd on my xeon box
00:44
<
icecream95 >
Everyone who has problems with OOM: 1. Install zram 2. Setup 5GB zram swap with zstd 3. Never complain about OOM again
00:45
<
macc24 >
icecream95: damn i wish i knew this when i tried to run cities skylines on overclocked phenom 9650 with 4gb of ram
00:45
<
macc24 >
it ran for few minutes before freezing, i went as far as disabling my wm to save ram xD
00:45
<
HdkR >
macc24: Just make sure to stick your swap on a 32GB/s NVMe array. :P
00:46
<
icecream95 >
macc24: Is that sarcasm?
00:46
<
macc24 >
icecream95: nope
00:46
<
macc24 >
the mcp78 chipset absolutely sucked under linux, especially with usb input
00:47
<
macc24 >
i ran freebsd on it for a while for that reason
00:50
<
alyssa >
--Honestly the fact I can use Chromium at all without fear is an improvement :o
00:51
<
macc24 >
alyssa: please tell me that chromium runs on my mt8183 chromebook when it runs linux
00:52
stikonas has quit [Remote host closed the connection]
00:53
<
macc24 >
finally my 5h print stopped printing base...
00:54
<
icecream95 >
macc24: Five hours? Is that printer from HP? :P
00:55
<
macc24 >
icecream95: it didn't refuse to work because my filament is not from hp so i guess no
00:55
<
alyssa >
--So, why did my computer shut off randomly?
00:55
<
alyssa >
Time to go crawl /var/log...
00:57
<
icecream95 >
alyssa: An actual power off (as opposed to a freeze) sounds like running out of battery, runaway thermals or pressing the power button
00:58
<
alyssa >
Those are all possible!
00:58
<
macc24 >
icecream95: or broken resume after accidentally suspendind
00:58
<
icecream95 >
macc24: Hurry up and fix suspend on duet!
00:58
<
alyssa >
"Stopping session 1 of user alyssa"
00:59
<
macc24 >
icecream95: when i get bored of my new toy
00:59
<
macc24 >
and when i get debug cable to work for more than one second
01:00
<
alyssa >
rk_iommu page fault at 0x3ee5000 of type read
01:00
<
alyssa >
that seems.. bad
01:00
* macc24
laughs in mediatek
01:02
<
alyssa >
meh, probably not my fault (tm)
01:02
<
macc24 >
woah, chromium runs on duet
01:03
<
alyssa >
it -is- a chromebook..
01:03
<
macc24 >
nvm it crashed
01:04
<
alyssa >
what kind of crash?
01:04
<
macc24 >
more like freeze
01:04
<
macc24 >
didn't look at dmesg
01:04
<
macc24 >
just mesa-git-master
01:04
<
macc24 >
my spidey sense tells me that it's gpu faults
01:05
<
icecream95 >
macc24: Why not try running CrOS in a virtual machine?
01:05
<
macc24 >
icecream95: ?
01:05
<
icecream95 >
CrOS == Chrome OS
01:05
<
macc24 >
yeah but why?
01:06
<
icecream95 >
Uhhh... Android app support?
01:06
<
macc24 >
icecream95: anbox?
01:06
<
macc24 >
ok heres a tip for everyone: if you're trying to figure out why display freezes, starting ssh server is a good idea
01:07
<
alyssa >
macc24: anbox doesn't support microphone input
01:07
<
macc24 >
alyssa: cadmium doesn't support microphone input
01:07
<
icecream95 >
It doesn't support microphone output, either
01:08
taowa has joined #panfrost
01:11
<
icecream95 >
macc24: That might be the same as the Firefox faults
01:12
atler has quit [Quit: atler]
01:12
<
macc24 >
icecream95: if you say it it may actually be true
01:13
atler has joined #panfrost
01:17
<
icecream95 >
macc24: I can try testing Chromium, but first I need to find a distro package linked against compatible libraries
01:17
<
macc24 >
icecream95: apitrace?
01:35
icecream95 has quit [Ping timeout: 240 seconds]
01:48
raster has quit [Quit: Gettin' stinky!]
01:51
BenG83 has quit [Quit: Leaving]
01:57
icecream95 has joined #panfrost
02:01
atler_ has joined #panfrost
02:04
vstehle has quit [Ping timeout: 256 seconds]
02:04
atler has quit [Ping timeout: 256 seconds]
02:38
_whitelogger has joined #panfrost
02:50
<
icecream95 >
macc24: I've got the Debian package for Chromium working now
04:06
karolherbst has quit [Ping timeout: 272 seconds]
04:16
dstzd has joined #panfrost
04:17
jernej has joined #panfrost
04:19
rcf has quit [Ping timeout: 260 seconds]
04:20
rcf has joined #panfrost
04:46
yawniek has quit [Ping timeout: 256 seconds]
04:47
yawniek has joined #panfrost
05:00
yawniek has quit [Ping timeout: 256 seconds]
05:01
yawniek has joined #panfrost
05:24
davidlt has joined #panfrost
05:29
icecream95 has quit [Ping timeout: 246 seconds]
06:00
vstehle has joined #panfrost
06:02
icecream95 has joined #panfrost
06:11
davidlt has quit [Ping timeout: 260 seconds]
07:02
<
icecream95 >
alyssa: 96 bytes is an odd size for shader prefetch, don't you think?
07:02
<
icecream95 >
I'd expect it at least to be an even multiple of 32 bytes (i.e. multiple of 64 bytes)
07:03
<
icecream95 >
Indeed, increasing it to 128 bytes fixes the CI failure for your branch
07:04
<
icecream95 >
So why did I come up with the 96 bytes figure in the first place?
07:05
<
HdkR >
I know of another architecture that prefetches 96bytes at a time, so it isn't entirely unheard of
07:06
<
icecream95 >
HdkR: Shhh, I'm telling a story
07:06
<
icecream95 >
When I was testing with different sizes, it was off a branch without any of the scheduler stuff, so all of the clauses were 32 bytes
07:07
<
icecream95 >
If I read the code properly, with "pan/bi: Don't pad constants" some of the clauses will be only 16 bytes
07:08
<
icecream95 >
HdkR: Now you made me forget what I was going to write next
07:10
<
icecream95 >
alyssa: Anyway, moral of the story is that BIFROST_SHADER_PREFETCH should be replaced with 128 - sizeof(last_clause)
07:12
Elpaulo has joined #panfrost
07:17
<
icecream95 >
After seeing the end-of-shader clause, the GPU is going to stay with the 128 bytes it has prefetched (which includes said eos clause) and not prefetch any more
07:52
icecream95 has quit [Read error: Connection reset by peer]
08:06
icecream95 has joined #panfrost
08:12
icecream95 has quit [Quit: leaving]
08:20
icecream95[m] has joined #panfrost
08:21
icecream95[m] is now known as icecream95
08:21
rak-zero has joined #panfrost
08:42
davidlt has joined #panfrost
09:01
kaspter has joined #panfrost
09:22
jschwart has quit [Ping timeout: 264 seconds]
10:03
jschwart has joined #panfrost
10:12
stikonas has joined #panfrost
10:40
leah has quit [Quit: WeeChat 2.8]
10:54
karolherbst has joined #panfrost
11:09
atler_ has quit [Quit: atler_]
11:10
atler has joined #panfrost
11:33
enty has joined #panfrost
11:33
ente has quit [Read error: Connection reset by peer]
11:46
embed-3d has joined #panfrost
11:52
Elpaulo has quit [Quit: Elpaulo]
11:59
raster has joined #panfrost
12:18
alpernebbi has joined #panfrost
12:45
jernej has joined #panfrost
12:50
jernej has joined #panfrost
12:52
unoccupied has joined #panfrost
12:52
atler has quit [Quit: atler]
12:53
unoccupied is now known as leah
12:54
atler has joined #panfrost
13:05
<
kinkinkijkin >
still no idea why supet tux kart hangs the system if you try to open the story mode, but everything else is fine
13:06
dstzd has joined #panfrost
13:08
<
macc24 >
kinkinkijkin: story mode is weird
13:08
<
macc24 >
it hanged midgard for a while iirc
13:20
chewitt has quit [Ping timeout: 256 seconds]
13:21
chewitt has joined #panfrost
13:53
chewitt has quit [Read error: Connection reset by peer]
13:53
chewitt_ has joined #panfrost
14:44
<
alyssa >
icecream95: I sure love stories.
15:56
kaspter has quit [Quit: kaspter]
19:03
raster has quit [Remote host closed the connection]
19:12
alpernebbi has quit [Remote host closed the connection]
19:12
robertfoss_ has quit [Ping timeout: 256 seconds]
19:12
xperia64 has quit [Ping timeout: 256 seconds]
19:13
robertfoss_ has joined #panfrost
19:13
endrift has quit [Ping timeout: 256 seconds]
19:13
tchebb has quit [Ping timeout: 256 seconds]
19:13
endrift has joined #panfrost
19:14
xperia64 has joined #panfrost
19:14
karolherbst has quit [Ping timeout: 272 seconds]
19:14
tchebb has joined #panfrost
19:15
davidlt has quit [Ping timeout: 246 seconds]
19:17
Depau_ has joined #panfrost
19:17
alpernebbi has joined #panfrost
19:18
Depau has quit [Ping timeout: 240 seconds]
19:57
chewitt_ has quit [Read error: Connection reset by peer]
19:58
chewitt has joined #panfrost
20:27
stikonas has quit [Remote host closed the connection]
20:33
stikonas has joined #panfrost
20:47
jschwart has quit [Quit: No Ping reply in 180 seconds.]
20:52
jschwart has joined #panfrost
20:58
stikonas has quit [Remote host closed the connection]
20:59
stikonas has joined #panfrost
21:02
alpernebbi has quit [Quit: alpernebbi]
21:33
<
robclark >
alyssa: you need to play more stk :-P
21:35
<
robclark >
(you know, or s/play/test/ if you must.. it's the same thing :-P)
21:42
<
macc24 >
robclark: i play, err, test stk on g72 and t860 on mesa master fyi
22:03
stikonas has quit [Remote host closed the connection]
22:04
stikonas has joined #panfrost
22:29
<
macc24 >
robclark: fyi qcom chromebook releases soon
22:41
<
robclark >
I heard something about that :-P
22:49
warpme_ has quit [Quit: Connection closed for inactivity]
23:09
<
icecream95 >
kinkinkijkin: STK story mode is working fine here (except for cacti being purple), maybe you need to update Mesa
23:17
raster has joined #panfrost
23:25
archetech has joined #panfrost
23:55
<
alyssa >
icecream95: What colour are cacti supposed to be? ELI Canadian.