<rperier>
(same kernel built with the same config , one is loaded from NAND via rk bootloader, the other one from sdcard via uboot)
<stdint>
pll_gpll
<rperier>
yeah, this is my feeling too, but as this is the first time I work on clk tree, I wanted another opinion :)
<wzyy2>
"I suppose it was bought from RK for designing a product and then someone decided to make some extra cash selling the “spares” when done"
<wzyy2>
- -i think it's the truth....
jelly-home has quit [Ping timeout: 255 seconds]
<wzyy2>
It would be interseting if someone can produce it..
<wzyy2>
It reminds me of my college time, me and my friend had sell many unauthorized home-made opensource boards to earn pocket money... repent ;-)
<wzyy2>
uhmm... The oneline documents have BOM list, gerber, sch, process-requirements.....
BenG83 has joined #linux-rockchip
BenG83 has quit [Client Quit]
BenG83 has joined #linux-rockchip
wzyy2 has quit [Remote host closed the connection]
wzyy2 has joined #linux-rockchip
nighty- has quit [Quit: Disappears in a puff of smoke]
<topi`>
the rk3399 has an interesting match of good single-threaded perf (the 2x a72 cores) and fairly good multithreaded perf because of 6 corew
<topi`>
the rk3368 is a nice product, but aside of some niche use cases with properly multithreaded workloads, the single-thread perf just isn't there
<topi`>
I'm not interested in running android, but running linux stuff, ./configure && make
<topi`>
perhaps dedicating a node for some cryptocoins, which demands good ST perf because the coin daemons are usually poorly multithreaded
<stdint>
well I think both of them having a hardware acceleration device for cryptography
<topi`>
the armv8a specifies aes instructions, and all the 64bit rockchip products implement them
<stdint>
but if you want cryptocoins, choose a GPU solution would be better
<topi`>
as opposed to amlogic s905 and others
<topi`>
I did not mean mining them; just running a full node which also consumes cpu cycles and lots of ram+disk
<topi`>
with a full node you can build sites like a block explorer, etc
afaerber has quit [Quit: Leaving]
cnxsoft has quit [Read error: Connection reset by peer]
cnxsoft1 has joined #linux-rockchip
cnxsoft1 is now known as cnxsoft
Aussie_matt has joined #linux-rockchip
descend-irc__ has joined #linux-rockchip
BenG83 has quit [Remote host closed the connection]
wzyy2 has quit [Ping timeout: 248 seconds]
afaerber has joined #linux-rockchip
wzyy2 has joined #linux-rockchip
jelly-home has joined #linux-rockchip
jelly-home has quit [Ping timeout: 258 seconds]
BenG83 has joined #linux-rockchip
<LongWork>
stdint: i have a question. Would adding AFBC to VPU something possible ? is that a lot of work ?
<phh>
LongWork: afaik vpu on rockchip is really hard-wired, it doesn't have a firmware or anything, so it definitely looks not possible
jelly-home has joined #linux-rockchip
<LongWork>
ah :/
<rperier>
mmind00: I think that the fix is mostly in uboot rather than the kernel itself. Because technically the kernel still has to run from the rk bootloader and uboot, right ? so I think that a first fix is to fix clk tree in uboot
<rperier>
(for rk3188)
<rperier>
that's what I am investigating
Aussie_matt has quit [Read error: Connection reset by peer]
<mmind00>
rperier: if the actual clock definitions are wrong in uboot yes ... but on the kernel side we should also make sure, clocks are setup correctly and not completely rely on the bootloader to set everything up
jelly-home has quit [Quit: ""]
jelly-home has joined #linux-rockchip
<rperier>
mmind00: that's something I wanted to check with you (I mean , confirmation). Can the kernel override pll rates and aclk/hclk/pclk stuffs completly ?
<mmind00>
rperier: yep (except the dpll rate used for dram of course)
<rperier>
yes, technically the kernel should be bootloader independant and works in all cases, that's just I have read and the default PLLs values in clk-rk3188.c (uboot) are wrong
<rperier>
mmind00: ok, good to know
<mmind00>
rperier: look at board dts files ... in the cru node we often have assigned-clocks sections setting core pll rates and such
<rperier>
looking
<rperier>
not for rk3188 and rk3xxx, only for rk3288
<rperier>
I will adapt the dt and investigate on this
jelly-home is now known as jelly
cnxsoft has quit [Quit: cnxsoft]
TheTrash has left #linux-rockchip ["http://quassel-irc.org - Chat comfortably. Anywhere."]
wadim_ has quit [Remote host closed the connection]
BenG83 has quit [Ping timeout: 276 seconds]
<rperier>
mmind00: it seems better with http://paste.ubuntu.com/25263590/, the clock tree looks more coherent, even if I still have blicking, but I think that's a part of the solution
<rperier>
:)
<rperier>
I still get weird stuffs on the cpll branch and but the rest looks good, compared to the working kernel