ChanServ changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://oftc.irclog.whitequark.org/panfrost - <macc24> i have been here before it was popular
soreau has joined #panfrost
hyrc_ has joined #panfrost
hyrc has quit [Ping timeout: 480 seconds]
camus1 has joined #panfrost
camus has quit [Read error: Connection reset by peer]
megi has quit [Quit: WeeChat 3.3]
WoC has quit [Remote host closed the connection]
WoC has joined #panfrost
camus has joined #panfrost
camus1 has quit [Ping timeout: 480 seconds]
davidlt has joined #panfrost
macc24 has joined #panfrost
davidlt has quit [Ping timeout: 480 seconds]
camus1 has joined #panfrost
camus has quit [Remote host closed the connection]
davidlt has joined #panfrost
camus has joined #panfrost
camus1 has quit [Ping timeout: 480 seconds]
indy has quit [Read error: Connection reset by peer]
indy has joined #panfrost
camus1 has joined #panfrost
camus has quit [Read error: Connection reset by peer]
rasterman has joined #panfrost
pendingchaos has quit [Ping timeout: 480 seconds]
pendingchaos has joined #panfrost
camus has joined #panfrost
camus1 has quit [Read error: Connection reset by peer]
camus1 has quit [Read error: Connection reset by peer]
floof58 has quit [Ping timeout: 480 seconds]
<CounterPillow>
It's a rockchip SoC so they're putting in everything but the kitchen sink
<CounterPillow>
see: hardware gzip decoding for some reason, and RGA for some reason, and 10-bit H.264 decoding for some reason (fansubbed anime?)
<daniels>
MTK still has it as well
dschuermann_ has quit []
dschuermann_ has joined #panfrost
dschuermann_ has quit []
dschuermann has joined #panfrost
<macc24>
CounterPillow: "fansubbed anime?" you can make youtube run on h264 too
<macc24>
enable v4l2 video decoding in chromium
<macc24>
and bam
<macc24>
hardware accelerated youtube
<macc24>
on arm
<macc24>
theoretically
<CounterPillow>
I meant specifically 10-bit H.264
<CounterPillow>
which is a profile nobody outside of anime fansubs uses
<macc24>
and how do you know that? ;)
<CounterPillow>
Spending too much time around weebs and multimedia developers (they have a large overlap)
<CounterPillow>
the v4l2 requests stuff sadly doesn't expose the 10-bit support in rkvdec yet last I checked :(
<macc24>
would be funny if there are weebs in rockchip that made rockchip put hardware accelerated 10bit h264 video decoding in there
<robmur01>
chances are it's still just a 3rd-party video IP that does whatever the vendor could easily throw in to make the spec sheet longer, which in turn serves to make the licensee's SoC spec sheet longer ;)
* robmur01
picks i.MX8Q datasheet at semi-random... howzabout that "RealVideo 8/9/10 decode" :D
<macc24>
robmur01: "3rd-party video ip" hmm
<robmur01>
see drivers/staging/media/hantro/
<robmur01>
the rkvdec is supposedly in-house, but they said that about the suspiciously-Cadence-looking RK3399 PCIe too...
davidlt has quit [Remote host closed the connection]
<macc24>
what next? in-house rkgpu? xD
<macc24>
wait
<macc24>
why is hantro still in staging?
f11f12 has quit [Quit: Leaving]
floof58 has joined #panfrost
<robmur01>
I believe it's still waiting for user ABIs for the newer formats to stabilise
Guest8553 has quit [Ping timeout: 480 seconds]
<CounterPillow>
Yep, e.g. HEVC and VP9 controls aren't stable yet I think
<CounterPillow>
And since they've now added AV1 to that, well
<macc24>
ugh
<cphealy>
IIRC, Hantro VP9 and VP9 controls landed for 5.17 kernel.
rasterman has quit [Quit: Gettin' stinky!]
floof58_ has joined #panfrost
floof58 has quit [Ping timeout: 480 seconds]
<jernej>
VP9 controls skipped staging and went to uapi directly
<jernej>
so only due to HEVC
<macc24>
oh wow
<macc24>
firefox runs faster on veyron minnie than i remember
<macc24>
with webrender
<macc24>
no more manually forcing opengl acceleration :D
rasterman has joined #panfrost
<alyssa>
CounterPillow: Oh come on, more Mali drivers to write? how terrible! :-p
<macc24>
alyssa: but imagine the hardware :D
<robmur01>
macc24: more than that, imagine the firmware :P
<robmur01>
sadly CSF will not please the libre-everything crowd... :(
<alyssa>
robmur01: speaking of, is the CSF blob available anywhere yet?
<macc24>
alyssa: CSF?
<alyssa>
macc24: v10 mali (g710, etc) needs a firmware blob called CSF
<alyssa>
well
<alyssa>
it's more complicated than that but yeah
<alyssa>
(g310, g510, g610, g710 and anything newer all blobby)
<alyssa>
G68/G78 are the last blobless malis
<alyssa>
That doesn't mean someone can't r/e CSF and write a FOSS firmware instead ;-)
<robclark>
any idea how big the CSF fw is? Maybe you get lucky and arm just threw a cortex-m at it?
<macc24>
alyssa: oh right
<macc24>
forgot
<alyssa>
robclark: Arm dropped GPL kernel sources but not the firmware
<macc24>
alyssa: i remember seeing it
<alyssa>
so I don't think we can know anything until the first Android phone shows up on the market
<macc24>
will drop link to u when im in state to do so
<alyssa>
...or that, I guess :0
<alyssa>
`mali_csffw.bin` is the string to grep for
<alyssa>
maybe that's floating around somewhere
<macc24>
i'm kinda busy with undisclosed stuff lol
<alyssa>
oh hey
<macc24>
i should've quit irc before doing it
<alyssa>
So it is floating around now ;)
<macc24>
quit as in quit irc client
<alyssa>
robclark: 236kib, grumble
<robclark>
ouch
<daniels>
robclark: I can't imagine a universe in which it is not a Cortex-M
<alyssa>
Cortex-A has entered the room
<macc24>
daniels: what about a mali 400 running firmware xD
<daniels>
alyssa: 240kB yes, but the latter ~73% is almost exclusively NUL
<anarsoul>
macc24: 240kb? nope
<robclark>
a cortex-m is much closer to what you'd want for a GPU babysitter than cortex-a (let alone another gpu).. the SQE thing in adrenos is even somewhat more specialized.. I defn wouldn't call it a general purpose CPU (it barely has load/store instructions.. it is more like a FIFO processor)
<macc24>
there are also cortex-r cores iirc
<robclark>
the way cortex-m handles interrupts would make it a much better candidate than cortex-a.. I don't remember about cortex-r but I thought it does interrupts the "cortex-a way"
<robclark>
cortex-m is actually kinda neat for a micro-controller
<macc24>
i've got a stm32 board with a cortex-m, fun little mcu
tlwoerner has quit [Ping timeout: 480 seconds]
tlwoerner has joined #panfrost
<alyssa>
daniels: Ah yes so it is
<alyssa>
robclark: It's indeed a cortex-m, how dreadfully boring :-p
<robclark>
I guess that saves you from having to figure out a new instruction set, at least.. not really sure how feasible writing a replacement fw would be (I'm not inclined to bother trying for adreno/sqe) but it has been useful to be able to look at what it is doing (and audit certain parts)
<alyssa>
Not sure either
<robclark>
if fw is handling switching pgtables when switching between cmdstream from different processes, that is probably an area you might want to audit.. and in general understanding how the protections against malicious userspace generated cmdstream is worthwhile.. but completely replacing blob fw seems of little benefit
<alyssa>
Noted
<robclark>
basically just make sure userspace can't install or modify it's own pgtables.. 'cause if that happens, it's going to be a bad day ;-)
<alyssa>
lolol
pendingchaos has quit [Remote host closed the connection]
pendingchaos has joined #panfrost
pendingchaos has quit [Ping timeout: 480 seconds]
pendingchaos has joined #panfrost
nicegoals has joined #panfrost
nicegoals has quit [Excess Flood]
<daniels>
robclark: … hypothetically
<macc24>
do yall ever look at your own code after few months and think "the author is an idiot"?
<macc24>
i got kinda sidetracked into re-writing cadmium's rootfs installation xD
pendingchaos has quit [Read error: Connection reset by peer]
pendingchaos has joined #panfrost
camus1 has joined #panfrost
camus has quit [Remote host closed the connection]
rasterman has quit [Quit: Gettin' stinky!]
pendingchaos has quit [Read error: No route to host]