00:03
vstehle has quit [Ping timeout: 260 seconds]
01:00
atler is now known as Guest51579
01:00
Guest51579 has quit [Killed (weber.freenode.net (Nickname regained by services))]
01:00
atler has joined #panfrost
01:33
WoC has quit [Remote host closed the connection]
01:50
taowa has quit [Ping timeout: 245 seconds]
01:57
taowa has joined #panfrost
02:55
zkrx has quit [Ping timeout: 240 seconds]
03:07
zkrx has joined #panfrost
03:09
urjaman has quit [Read error: Connection reset by peer]
03:09
davidlt has joined #panfrost
03:10
urjaman has joined #panfrost
05:00
vstehle has joined #panfrost
05:47
<
tomeu >
ayaka: that kernel repo was abandoned once panfrost got merged into mainline
05:48
<
tomeu >
ayaka: you probably want to ask with question in #linux-rockchip
05:59
archetech has quit [Quit: Konversation terminated!]
06:05
adjtm has joined #panfrost
06:07
adjtm_ has quit [Ping timeout: 260 seconds]
06:36
adjtm_ has joined #panfrost
06:38
adjtm has quit [Remote host closed the connection]
06:56
alpernebbi has joined #panfrost
07:01
nlhowell has joined #panfrost
07:18
nlhowell has quit [Quit: WeeChat 3.1]
07:19
nlhowell has joined #panfrost
07:32
vstehle has quit [Quit: WeeChat 3.0]
07:33
vstehle has joined #panfrost
07:35
chewitt has joined #panfrost
07:37
davidlt has quit [Ping timeout: 260 seconds]
07:40
<
ayaka >
tomeu, then that is not necessary
07:40
<
ayaka >
tomeu, as linux rockhip is not an official channel also I am sure the backport for lts 4.4 is closed
07:45
<
chewitt >
if you wanted to backport the driver onto Rockchip BSP kernel(s) .. that would be something for Rockchip kernel maintainers to handle
07:45
<
chewitt >
and I assume you know who those people are :)
07:55
<
ayaka >
I just want to make some commit that when someone using vendor kernel could adopt this driver easily
07:55
<
ayaka >
using a upstream kernel would be a big deal for them
07:56
<
ayaka >
now I know there is not a proper place for that, I know should forget this kind of though now
08:32
raster has joined #panfrost
09:41
davidlt has joined #panfrost
09:48
stikonas has joined #panfrost
10:04
davidlt has quit [Remote host closed the connection]
10:14
davidlt has joined #panfrost
10:33
warpme_ has joined #panfrost
11:58
<
chewitt >
ayaka I would expect the current driver to depend on memory things that are not present in the BSP codebase
11:59
<
chewitt >
better option would be to work on porting the nand and camera drivers :)
12:00
<
chewitt >
the stateless/vdec drivers are quite good shape now - although the userspace side (ffmpeg) is not really upstream yet
12:02
<
chewitt >
there's a V4L2 love-fest/bring-up party happening in LE slack #raspberrypi channel for the last couple of months
12:02
<
chewitt >
I think you still have access, so feel free to ask Q's on status there
12:41
nlhowell has quit [Quit: WeeChat 3.1]
12:41
nlhowell has joined #panfrost
12:57
WoC has joined #panfrost
13:22
<
ayaka >
chewitt, well I was working on the that stateless driver, I am sure its performance is worse than vendor one
13:22
<
ayaka >
besides the gstreamer plugins bad has a v4l2 plugin for those stateless driver
13:22
<
ayaka >
I don't think it would be worst to port that
13:23
<
chewitt >
from what I understand the current stateless support is v1.0 .. enough for actual implementations to start
13:24
<
chewitt >
but once people implement, all the performance and niche cases come up, so there will be changes/improvements
13:24
<
ayaka >
sorry, I didn't notice who you are
13:25
<
chewitt >
then there is the "coordinated ballet" (aka cat-herding) exercise of aligning all the different implementations
13:25
<
chewitt >
ahh.. hello :)
13:35
danboid has joined #panfrost
13:36
<
danboid >
Is panfrost running on a Mali T720 capable of 4K 60 fps video playback, when using cedrus for hw decoding?
13:37
<
danboid >
I've got a HDMI 2.0 120 Hz TV that needs exploiting :)
13:37
<
HdkR >
The GPU doesn't decode video
13:37
<
HdkR >
It's a 3D accelerator only
13:38
<
HdkR >
There are other hardware blocks on your SoC that does video decoding
13:38
<
danboid >
Does the xserver not matter?
13:38
<
danboid >
xorg driver
13:39
<
HdkR >
Panfrost doesn't even do display output unless you're compositing something
13:39
<
HdkR >
That's another hardware block
13:40
<
danboid >
I want to use mpv on a Orange Pi 3, do you know if it would be able to handle 60 fps h265 decoding as promised on the spec sheet?
13:40
<
ayaka >
Mali T720 won't be able to render a UHD 30HZ output
13:40
<
ayaka >
it would be luckily that it could reach 24fps
13:41
<
danboid >
So it can decode a 4K but ony output 1080?
13:43
<
danboid >
I've read that the H3 / H6 socs can decode 4K h264 and h265 in hardware
13:43
<
danboid >
but thats not much use if it can't display 4K smoothly
13:44
<
Werner >
decoding a video stream does not necessarily mean to display it through a video output
13:44
<
danboid >
Thats what most people want it for
13:46
<
danboid >
So is there no accelerted 2D X server for Mali T7XX
13:47
<
HdkR >
Mali is only a 3D accelerator
13:48
<
danboid >
OK so is glamor known to work with panfrost?
13:49
<
HdkR >
glamor will work
13:50
<
danboid >
OK, sounds like the Tegra is still king of the ARM GPU hill then
13:50
<
HdkR >
Sure, if you get a Xavier then it'll stomp Mali
13:51
<
danboid >
but are you saying on of the Mali GPUs is better than the GPU in the nano?
13:51
<
HdkR >
Ones in high end SoCs yes
13:51
<
HdkR >
But those also typically don't go in to SBCs
13:52
<
danboid >
Which soc is this?
13:52
<
HdkR >
Kirin 9000, Samsung Exynos 2100
13:54
<
HdkR >
Comparing a low-end SoC to a flagship SoC from a company known for making overpowered hardware isn't really fair :)
14:06
<
danboid >
Maybe not, I'm just looking into the state of hardware accelerated video decoding for ARM Linux
14:06
<
danboid >
It seems the NVIDIA boards are the only game in town really
14:07
<
danboid >
Someone says something good about NVIDIA SHOCKER!!
14:12
<
amonakov >
I don't quite understand why the level of developer support for Mali and Adreno has not improved in the years of booming Android sales
14:12
<
amonakov >
(don't tell me it would have been even worse otherwise, because I don't see how it possibly could be worse)
14:13
<
HdkR >
It could be worse
14:13
<
adjtm_ >
I don't see how
14:14
<
HdkR >
PowerVR could be dominant
14:15
<
adjtm_ >
they are waiting to the right moment to write open source drivers for their gpus
14:16
<
adjtm_ >
I don't understand why took so long for the companies using the mesa drivers instead of the closed source ones (because I assume that new qcom phones use freedreno)
14:16
<
amonakov >
ah. and Vivante I guess. but one dominant player is not worse for developers than two+ shiity players!
14:17
* alyssa
isn't sure this is on-topic here...
14:17
<
adjtm_ >
alyssa, sorry
15:32
<
chewitt >
if that works, you know where to steal all the patches for whatever project you have
15:33
<
chewitt >
it's about as bleeding edge as you'll get for Allwinner H/W decoding
15:33
<
chewitt >
and further discussion can move to #libreelec channel
15:40
mdrjr has joined #panfrost
15:47
atler has quit [Quit: atler]
15:49
atler has joined #panfrost
16:38
danboid has quit [Remote host closed the connection]
17:25
_whitelogger has joined #panfrost
17:50
<
macc24 >
i think terraria does something that kills mali gpus
17:50
<
macc24 >
on g72 it's slow too
17:51
<
anarsoul >
is it slow on intel graphics?
17:51
<
macc24 >
my duet's gpu outperforms ironlake by a great margin, but it can't play terraria at speed that ironlake can do
17:52
<
macc24 >
it's under half frames per second than ironlake could co
18:04
felipealmeida has quit [Ping timeout: 260 seconds]
18:04
felipealmeida has joined #panfrost
18:06
paulk-leonov has quit [Read error: Connection reset by peer]
18:15
paulk-leonov has joined #panfrost
18:33
popolon has joined #panfrost
19:00
davidlt has quit [Ping timeout: 260 seconds]
20:45
indy has quit [Ping timeout: 265 seconds]
20:59
indy has joined #panfrost
21:05
alpernebbi has quit [Quit: alpernebbi]
21:11
nlhowell has quit [Ping timeout: 252 seconds]
21:18
warpme_ has quit [Quit: Connection closed for inactivity]
21:51
<
alyssa >
I wonder how much benefit we could get from disabling clean_pixel_write_enable when it's safe to do so
21:52
<
alyssa >
I guess the problem is it's not usually safe to do so on Midgard due to a hw erratum...
23:05
indy has joined #panfrost
23:15
bbrezillon has quit [Ping timeout: 240 seconds]
23:16
bbrezillon has joined #panfrost
23:50
popolon has quit [Quit: WeeChat 3.1]