<KotCzarny>
hmm, what gpio number in uboot has button on opipc?
premoboss has joined #linux-sunxi
diego_r has joined #linux-sunxi
paulk-collins has joined #linux-sunxi
yann|work has joined #linux-sunxi
_massi has joined #linux-sunxi
<yann|work>
wens: anyone working on this soc, btw ? the situation for that board does not seem much better than that of the M2 ?
robogoat has quit [Ping timeout: 240 seconds]
robogoat has joined #linux-sunxi
<wens>
yann|work: vishnu is working on a83
<wens>
progress is somewhat blocked on clk work
<yann|work>
if only I had the time to play with the M2 I have here - but it's more motivating to use the more powerful ones :)
<KotCzarny>
well, if powervr would be supported it would be much more powerful for media related tasks
<KotCzarny>
also, it's also quad core a7, which one of yours is more powerful?
<yann|work>
well, the H3 is at least comparable ?
<KotCzarny>
both are 1.2ghz quad core a7
<KotCzarny>
process is the same 40nm too
<KotCzarny>
ahm, no
<KotCzarny>
h3 seems to by 28nm
<yann|work>
the wiki says 28
<KotCzarny>
also, a31 seems to be 2-channel ddr
<yann|work>
ah, only one for the H3 ?
<KotCzarny>
h3 has half of the l1/l2 cache compared to a31
<yann|work>
strange, the wiki page for H3 claims VFP4 support, but cpuinfo only reports VFP3
<KotCzarny>
i wonder how memory benchmarks compare on both
paulk-collins has quit [Quit: Quitte]
<KotCzarny>
Features : swp half thumb fastmult vfp edsp thumbee neon vfpv3 tls vfpv4 idiva idivt
<KotCzarny>
here on opipc
<KotCzarny>
so it might be your kernel
apritzel has joined #linux-sunxi
<yann|work>
oh no, it's just my eyes in fact... looks like if late at light my brain's pattern-matcher is not as greedy as it should :)
<KotCzarny>
use the grep padawan, the grep
<KotCzarny>
erm. h3 has twice cache of a31
<KotCzarny>
seems its a bit early for me too
<KotCzarny>
um
<KotCzarny>
scratch that
<KotCzarny>
its l2
<KotCzarny>
someone should tidy the features on wiki pages
<KotCzarny>
so it seems a31 to be twice better than h3 for multimedia tasks
<yann|work>
including video decoding ?
<KotCzarny>
as long powervr is supported
<yann|work>
ah yes, that was the catch ;)
<KotCzarny>
The Mali-400 if its the MP4 version is 2nd most powerful, but still only half as powerful as the SGX544MP2 at best. If the Mali-400 is the MP2 version, then its 1/4th the performance at best, and the Adreno 203 is close to it.
<yann|work>
wondering if for cedarx decoding+rendering we really need that much powervr support
libv_ has joined #linux-sunxi
libv has quit [Ping timeout: 276 seconds]
naobsd has quit [Quit: naobsd]
<MoeIcenowy>
but it's nearly not possible to support pvr on GNU/Linux
<MoeIcenowy>
the pvr driver architecture is a disaster
<KotCzarny>
:>
<KotCzarny>
dont tell me, i have 6 year old nokia n900 with pvr530
<yann|work>
hey, there's not even a page about powervr in the wiki :(
<KotCzarny>
wanna create it and say 'dont you dare to explore this area, dragons ahead' ?
<apritzel>
MoeIcenowy: yeah, but still 2.0 mm header and 1.8V, propably
<apritzel>
MoeIcenowy: and you don't know _which_ UART they have on the connector
<libv>
and you got some stale dead kernels left and right, some had already vanished again
<apritzel>
on the Hikey it wasn't the one with the boot messages and the console :-(
<libv>
and then images... endless images.
<MoeIcenowy>
apritzel: ah-oh
<libv>
and none of the people who seemed so active at the time still cared about anything much arm after that
<libv>
yeah, late 2012
<apritzel>
MoeIcenowy: also I don't see a DT in the kernel ...
<libv>
there was a fraction of interest from the same people in Q2 2013 for the A20 based boards, but that was pretty limited, and not very good in terms of ensuring that the acquired knowledge survived
<libv>
hence me whining to get linux-exynos off the ground
<libv>
and the exact same thing is replaying now between MoeIcenowy and apritzel :)
<libv>
which of the two is the better aarch64 platform.
<MoeIcenowy>
libv: hahaha
<MoeIcenowy>
interesting
<MoeIcenowy>
I think A64 is a fairly cheap solution
<libv>
it's a development board of qualcomm
<KotCzarny>
but seriously, what is so magic about 64bit at the moment?
<libv>
it has a shelf life of minutes.
<MoeIcenowy>
but Pine64 is terrible for it's bad source archive
<libv>
some corporates will buy and use it for ages, but that's it
<libv>
you cannot buy it anymore in 2017
<apritzel>
MoeIcenowy: I only care about the hardware
<libv>
olimex is going to use the a64
leviathancn has quit [Ping timeout: 240 seconds]
<ssvb>
apritzel: ok, I'll push it out in a few hours
<MoeIcenowy>
Some Allwinner boards do exists for a long time
<apritzel>
ssvb: thanks!
<MoeIcenowy>
for example, cubieboard series
<apritzel>
ssvb: I am tempted to quickly hack the RMR stuff on top of it
<libv>
whatever reproducible work you do on allwinner a64, like with anything reproducible that the sunxi community has done and documented, remains
<libv>
that is one constant of linux-sunxi
<MoeIcenowy>
but Allwinner A64 do not have a quite good GPU
<libv>
dragonboard will be catching dust faster than you think
<MoeIcenowy>
not performacely good
<MoeIcenowy>
neither
<libv>
MoeIcenowy: who cares?
<MoeIcenowy>
nor open-source-ly good
<libv>
and remember who you are talking to.
<MoeIcenowy>
Hmm
<MoeIcenowy>
But I still don't think a Mali400MP2 can be better than an Adreno 306
clonak has joined #linux-sunxi
<libv>
if a64 is indeed aarch64 in a cheap and relatively useful package like a10/a20, then it will be everywhere
<libv>
and sunxi will, once again thanklessly, be the canonical source of support for it
<MoeIcenowy>
but the tablet market is not as good as it is in the era of a10/20
<MoeIcenowy>
allwinner is gonna die in this market
<MoeIcenowy>
by being attacked by Mediatek
<libv>
MoeIcenowy: the ultra-cheap development board market on the other hand...
<MoeIcenowy>
libv: hmm
<MoeIcenowy>
It may still be the point
<MoeIcenowy>
but development board is not comsumer-level at all
<MoeIcenowy>
I don't it have a grand market
<MoeIcenowy>
it's only for us
<MoeIcenowy>
but not for terminal users
<libv>
where allwinner again, has to thank linux-sunxi for having provided the stability that allwinner itself fails or refuses to provide
<MoeIcenowy>
hhh
<MoeIcenowy>
allwinner do have no stability
<apritzel>
I guess they should give up on software at all ;-)
<KotCzarny>
MoeIcenowy: development boards started to be viable desktop/htpc replacements
<apritzel>
instead just contracting #linux-sunxi
<MoeIcenowy>
apritzel: good idea
<libv>
apritzel: none of those cheap chinese shops work like that
<MoeIcenowy>
maybe they should provide all technical documents
<MoeIcenowy>
and let linux-sunxi work in the mainline things
<libv>
also, almost none of the not so cheap western shops work like that
<MoeIcenowy>
:-)
<libv>
how many big open source names work for a freescale/nxp?
<libv>
heck, even samsung is having a hard time embracing these ideas
<libv>
there's intel and ...
<MoeIcenowy>
to be honest, the popularity of sunxi development boards are still not as high as the raspberry pi
<libv>
amd is almost dead anyway
<libv>
MoeIcenowy: but they are the first alternative
<MoeIcenowy>
libv: yes
<MoeIcenowy>
but I've never heard any educational usage of sunxi boards
<libv>
and to some extent, we should count ourselves lucky that every single one of those cheap board makers copies linux-sunxi work and documentation, and then sets up an idiot forum somewhere
<MoeIcenowy>
(maybe except CHIP?
<libv>
MoeIcenowy: educational use of the rpi?
<libv>
"getting kids to program assembler again"
<MoeIcenowy>
libv: seems to be python, not asm
<libv>
that translated into: "hey, here is a handful of guys REing the videocore, let's tell them that we are going to release our own bootloader in future so that they will give up"
<MoeIcenowy>
in addition, it's right that dragonboard has 1.8v uart
<MoeIcenowy>
For the board discussion
<MoeIcenowy>
sunxi have now no UEFI interface
<MoeIcenowy>
it may be a disadvantage
<jelle>
I remember thinking I could make a nice video camera project with the RPI years ago at uni
<wens>
no UEFI, some may say it's better :)
<apritzel>
MoeIcenowy: there are UEFI patches on top of U-Boot now ;-)
<MoeIcenowy>
apritzel: oh?
<apritzel>
by Alex Graf
tkaiser has joined #linux-sunxi
<libv>
who was very busy when a15 hit
<libv>
(ex)SuSE guy
<libv>
like me
<libv>
so interesting to see his name pop up now after my rant there before :)
<apritzel>
the H3 driver just uses apb1, apb2, ahb1, ahb2
<wens>
not the bus gates
<wens>
since each mod clock might have different parents
<wens>
apritzel: another way would be to boot their kernel, and dump /sys/kernel/debug/clk/clk_summary
<apritzel>
wens: "booting their kernel": you are asking for much ;-)
<apritzel>
wens: since you seem to be knowledgeable about this: doesn't the parent of a bus gate actually depend on the mux setting in the respective clock?
<apritzel>
(I guess I am perfectly confused in this matter ...)
domidumont has quit [Remote host closed the connection]
<wens>
apritzel: the "bus" clock enables access to the device over the bus
domidumont has joined #linux-sunxi
<apritzel>
my understanding it that it gates the clock just before it enters the device, is that right?
<wens>
for some peripherals, that doubles as the peripherals internal clock
<wens>
some blocks also have a second "mod" clock, like mmc, which can also be gated, and tuned
<wens>
and allwinner's code actually does a poor job of modeling bus clock gates and their parents
domidumont has quit [Remote host closed the connection]
<apritzel>
good to know ... ;-)
domidumont has joined #linux-sunxi
<wens>
devices with separate mod clocks don't get an ahb parent
<wens>
which sucks for people looking at it and trying to figure out the bus clock tree
<wens>
apritzel: anyway, please just follow the clock diagram in the datasheet for now :)
<apritzel>
wens: OK, thanks, will do
<KotCzarny>
hrm, uboot build system ignores CC variable (ie. ccache/distcc not working when not specified on make cmd line, but only in env)
Net147 has quit [Ping timeout: 264 seconds]
Net147 has joined #linux-sunxi
IgorPec has joined #linux-sunxi
<apritzel>
wens: too bad you are not as FOSDEM ...
<wens>
apritzel: well, you might not be able to find me there even if i did attend :)
<apritzel>
you have a point there ;-)
<apritzel>
but I could keep you from actually eating at the dinner ;-)
<apritzel>
ssvb: you mean you load U-Boot via FEL instead of loading it from the SD, right
<apritzel>
?
<apritzel>
sound worthwhile, indeed
<ssvb>
yes, you can load it via FEL and you can load it from the SD card too
<apritzel>
ssvb: like that message at the end of the libdram commit ;-)
hipboi has quit [Quit: This computer has gone to sleep]
<apritzel>
ssvb: by the way, I also pointed TLLim to libdram
leviathancn has joined #linux-sunxi
<libv>
MoeIcenowy: after lima had existed for like 2 years
<libv>
the fsf stopped making just REing powervr the priority graphics project
<libv>
lima is/was of course on top of that list
<libv>
guess how often i have been asked by anyone related to the fsf whether lima needed any help...
<MoeIcenowy>
libv: but none helped you?
<libv>
i got asked this 4 weeks ago, 5ys down the road, and 4 years after the big announcement, and today i cannot believe that i actually wasted my time answering this email, as of course this was not a serious request.
<libv>
MoeIcenowy: this was the second time i got to deal with the fsf
<libv>
i have been developing graphics drivers for 12.5ys
<libv>
i made display drivers independent of BIOS, i freed ATI, provided the first native display support for coreboot and opened up arm GPUs
<libv>
the fsf is simply irrelevant
<libv>
and incapable of seeing that they are, let alone them doing something about this
<libv>
the last time i dealt with them was when RMS was in .tw claiming that .tw vendors are soo much better than .us ones
<ssvb>
apritzel: now I'm going to finalize fel boot support, and then take care of the axp803 pmic
<wens>
hehe
<libv>
whereas VIA had just replaced the licenses on their xfree86 driver from MIT to VIA proprietary
<libv>
and i had just spent weeks trying to wring VIAs hands to get this fixed again
<libv>
the fsf does not live in the same dimension
<wens>
via is mostly irrelevent today
ssvb has quit [Quit: Leaving]
<libv>
wens: as said, 12.5 years
<libv>
wens: AMD is mostly irrelevant today
<libv>
VIA is just gone
<libv>
but VIA is why you have terms like "modesetting" and structured (mostly) display drivers
<libv>
it's where i cut my teeth
<libv>
but still, the fsf is irrelevant, whatever bullshit they write on their website is just that, and there is nothing real or sensible behind it
<MoeIcenowy>
VIA still produces some SoCs that contains GPU.
<libv>
MoeIcenowy: no-one buys them anymore
<libv>
no-one cares
<MoeIcenowy>
And recently VIA created a company in China mainland named "Zhao Xin" to produce them
<libv>
this was already true at the end of the last decade
<libv>
the only legacy VIA leaves is ITX
<wens>
intel is now introducing STX :|
<libv>
there is some weird relationship between the people behind HTC and VIA, and wondermedia and whatnot
<libv>
i actually bought the wondermedia based things a year ago
<libv>
as they have mali-200 and iirc uboot and kernel were thrown over the hedge at one point
<libv>
i should spend some time on them and make sure that i at least have all the basics in a wiki and on github
<libv>
but there are som many things i should be doing :(
* WarheadsSE
has a read the pile of libv truth/rant
* WarheadsSE
agrees
montjoie has quit [Ping timeout: 265 seconds]
* WarheadsSE
departs again
montjoie has joined #linux-sunxi
leviathancn has quit [Ping timeout: 250 seconds]
<KotCzarny>
libv: write email to fsf that you need workers team subsidised for you
<KotCzarny>
:>
<libv>
KotCzarny: "i cannot believe that i actually wasted my time answering this email"
<libv>
at 13:26
<KotCzarny>
yeah, still, if they follow up
<MoeIcenowy>
yes if they follow up
<libv>
they do not follow up.
<MoeIcenowy>
even when there's now both ISA and limare now
<libv>
they write irrelevant manifestos and hold rants, nothing more
<libv>
and this priority projects page is more of the same
HeHoPMaJIeH has joined #linux-sunxi
HeHoPMaJIeH has quit [Changing host]
HeHoPMaJIeH has joined #linux-sunxi
premoboss has quit [Ping timeout: 272 seconds]
<KotCzarny>
libv, how come you are not on a payroll from some mainstream distro to do graphics driver work?
matthias_bgg has joined #linux-sunxi
<libv>
KotCzarny: because i have a big mouth, and because i say unpopular things that very often turn out to be true.
<libv>
and i also do some things and create facts that forces people and company to change their behaviour and that is never popular.
<libv>
companies even
<KotCzarny>
right, management then
<MoeIcenowy>
libv: which code in limare uses libMali.so ?
<libv>
MoeIcenowy: my command stream code uses the binary shader compiler
<libv>
but a code to run on handgenerated shaders from connor also exists in the same codebase
<libv>
we had the simplistic gles1 q3a running well on that
<libv>
which only needs a handful of shaders anyway
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<apritzel>
so I use the a20 bus_gates driver, but only for MMC and the pinctrl
<jemk>
apritzel: to me your dt looks good, and all related clocks look the same in the h3&a64 datasheets (except the pll6x2 vs pll6 you already fixed)
<apritzel>
jemk: right, that what I checked yesterday, too
<apritzel>
jemk: thanks for taking the time to check this!
leviathancn has quit [Ping timeout: 272 seconds]
<apritzel>
(the a20 bus_gates driver because 4.4-rc8 didn't have the H3 one yet)
<apritzel>
and I changed the timer and the GIC node already, doesn't make a difference (old DT with new nodes still boots, new DT with old nodes still doesn't)
<apritzel>
maybe I should reformat both DTs to be easier comparable ...
<mripard>
apritzel: I might have an idea
<mripard>
2s, let me check
<wens>
where is your non-working clock?
<mripard>
hmm, no, nevermind
<wens>
non-working dt i mean
<mripard>
apritzel: with your working DTS, it would be great if you could simply make sure that all the clocks you define are registered
<mripard>
especially the bus gates parents
<apritzel>
wens: it's either <&bus_gates 112> or <&apb2> directly
<mripard>
what I'm guessing is that the bus gates driver doesn't register
<apritzel>
mripard: in the old one?
<mripard>
so the UART cannot get its clock
<apritzel>
I have printks in there and the H3 bus gates gets enabled
<mripard>
but since ít doesn't register, it isn't garbage collected by the clock framework either
<apritzel>
ah
<mripard>
which would also explain why it works with the 24MHz clockc
<mripard>
but without more information, like logs, clk_summary and so on, it's just a wild guess
<apritzel>
you mean that the final clk_register_gate() fails?
<wens>
apritzel: whole dts plz
<mripard>
yeah
<apritzel>
wens: the old DTS is in the gisthub link above
<apritzel>
mripard: thanks for your help, will try it later, the VGIC is calling again :-(
<mripard>
plaese give us some logs next times
<mripard>
at least the earlyprintk (or whatever equivalent we have in armv8) ones
<apritzel>
mripard: sure (sorry about that)
JohnDoe_71Rus has joined #linux-sunxi
<apritzel>
and yeah, arm64 does not have DEBUG_LL or earlyprintk
<apritzel>
but earlycon
<wens>
no earlycon support in 8250_dw yet
<apritzel>
wens: ???
<wens>
is very low on my todo list
<apritzel>
earlycon=uart,mmio32,0x01c28000 worked liked a charm for me
<wens>
the designware-specific 8250 driver doesn't have the earlycon OF macro to match "snps,dw-apb-uart"
<wens>
so no automagic earlycon afaik
<wens>
of course you can do earlycon= :p
<apritzel>
ah, got it
<wens>
clk_ignore_unused is also helpful if you want to kick some clk problems down the road
<wens>
though not recommended
<yann|work>
libv: as I read it, the priority list from the fsf is not about the fsf's priority, rather a hint to people looking for something to do for the community - the only problem being that most talented people already have something to do :>
paulk-collins has joined #linux-sunxi
premoboss has quit [Quit: Sto andando via]
<libv>
yann|work: no-one looks there and says: "let's go do that"
<yann|work>
libv: the question would rather be, would the fsf be willing to spend money to pay someone like you to get this priority off the ground (and obviously, would someone like you be available for this)
<yann|work>
right, it's rather a "state of the world"
<libv>
i know, i need to spend some time with sunxi-mali
<libv>
but tonight i am going to do graphics devroom stuff, same for all of thursday
<libv>
yann|work: done though, thanks for pointing this quick change out
afaerber has quit [Quit: Ex-Chat]
domidumont has quit [Read error: Connection reset by peer]
afaerber has joined #linux-sunxi
hp197 has quit [Ping timeout: 250 seconds]
hp197 has joined #linux-sunxi
Rambler44 has joined #linux-sunxi
<Rambler44>
So I was looking at a H3 and A64 board, it has about the 1 GHz I need and made with 28nm process. Was wondering if I should pick an earlier Allwinner device for support, I guess I'm too greedy looking at Cortex-A7+, maybe one with A20?
<Rambler44>
The thing is my basic understanding of how processors work reaches certain limits when it comes to cores, my computer right now has a dual-core, so does my phone. Now where lies the benefit exactly on having a quad-core H3 over a single-core 1GHz Broadcom CPU put in the RPi Zero for example
<plaes>
Rambler44: if you want "good support" then it's A20
<Rambler44>
When it comes to OpenSSL it will, for instance, appear to be hardcoded to run with a core so in essence it should run better on the Zero then on for example the A20?
<Rambler44>
plaes, I see, kind of expected. It's older (but not that old) and OSHW companies like OLIMEX have taken it in.
Netlynx has joined #linux-sunxi
Netlynx has quit [Changing host]
Netlynx has joined #linux-sunxi
<Rambler44>
I heard you should stay away from Banana Pi for adding crappy USB-to-SATA etc, but BPi M1 with A20 might be exactly what I'm looking for.
<plaes>
it's still missing proper display driver, though framebuffer works
<Rambler44>
Terrible USB-to-SAT or one of those models that had a real one?
<KotCzarny>
Rambler44: only a10/a20 have real sata
<KotCzarny>
and opipc has 4 real usb ports (each independently connected to chip instead going through internal hub)
<Rambler44>
Yeah, at the price tag of $15 I was also considering one of those. I mean it's unbeatable. But 1.6 GHz what have they been doing with their software?
<apritzel>
I guess terrible is relative here, having native SATA (even a crappy implementation) beats any other storage solution
<apritzel>
(we have in those kind of boards, that is)
<KotCzarny>
apritzel: tkaiser reached 120MB on usb-uas raid using 3 usb ports
<KotCzarny>
dont remember if it was on write or read
<apritzel>
KotCzarny: at what CPU load?
<apritzel>
seriously: sounds "USB-UAS RAID over three ports" sounds like a stretch to me
<apritzel>
also I guess you can only do this on the limited number of devices that actually have three USB controllers, right?
<KotCzarny>
apritzel: remember that native sata also eats your cpu
<Rambler44>
I probably don't need that much write/read that it's necessary with a native SATA, only issue is that if I do have an internal drive am I supposed to go around trying to find a dock for it.
<KotCzarny>
and mind you, with 3x usb you can offload them on different cpus, with sata you have only 1
mzki has quit [Quit: leaving]
<KotCzarny>
apritzel: yeah, only some boxes connect without internal hub
<Rambler44>
OPI PC does only have a 100Mbps NIC though
<Rambler44>
Depending on what you are doing with those drives there is a nice limit there.
<KotCzarny>
i think opi plus has gigabit
<Rambler44>
Might attach some crappy USB-NIC?
<apritzel>
KotCzarny: how can a hub give you any benefit in terms of speed?
<KotCzarny>
apritzel: hub gives your negative impact
<Rambler44>
KotCzarny Yeah but I see it has a lot of useless stuff on the PCB too
<Rambler44>
Still cheaper than picking an A20 boards of course
formruga has quit [Quit: Konversation terminated!]
<KotCzarny>
Rambler44: it all depends on what you are trying to build, for me i'm happy with bpi-r1, dont need stellar speeds, but having no middle man (usb-sata) makes all sata things work
<Rambler44>
I just don't like the NAND and Wi-Fi module they are putting on it.
<KotCzarny>
(ie. smart)
<Rambler44>
Of course I can see the benefit of an A20
diego_r has quit [Quit: Konversation terminated!]
<Rambler44>
And in a way, if this is the only development board with proper SATA if there are no hopes to see other ones in the distant future that's probably one of the CPUs one should have.
<KotCzarny>
Rambler44: even with native sata a10/a20 devices are limited to ~40-45MB/s write
<KotCzarny>
but read is ~110-200MB/s
<Rambler44>
Interesting. Assuming it's stuck with whatever old kernel Allwinner threw at it.
<Wizzup>
I get 130MB/s read on mainline
<Wizzup>
With a SATA PM even
<KotCzarny>
Rambler44: write cap is unknown, happens on both legacy and mainline, so probably something hw wise
iamfrankenstein1 has joined #linux-sunxi
<Rambler44>
I see. Seems like OPI has another advantage, we got a lot of MicroSDHC cards lying around and from the wiki it says it has that.
JohnDoe1 has joined #linux-sunxi
<Rambler44>
A20 would probably use the non micro ones.
<KotCzarny>
but what can you say about cheap box, if you want performance you dont look into 15$ range
vishnup_ has joined #linux-sunxi
enrico__ has joined #linux-sunxi
adj_ has joined #linux-sunxi
<KotCzarny>
if you want to support open source go for olimex lime2
<Rambler44>
So you would say an A20 would perform better? Its not like money matters that extremely much just there is a surcharge on import fees to EU here at a certain point.
iamfrankenstein has quit [Ping timeout: 240 seconds]
JohnDoe_71Rus has quit [Ping timeout: 240 seconds]
enrico_ has quit [Ping timeout: 240 seconds]
adj__ has quit [Ping timeout: 240 seconds]
iamfrankenstein1 is now known as iamfrankenstein
<Rambler44>
So it can fulfill the treaty on passing over items without taxation?
<KotCzarny>
do you have tme.eu store in your country?
avph has quit [Ping timeout: 240 seconds]
<Rambler44>
Yes
<KotCzarny>
check the price then
<KotCzarny>
they are direct olimex reseller i think
<Rambler44>
Ah, it seems like the A20 you sent me does micro SD?
avph has joined #linux-sunxi
<KotCzarny>
most a20 boards have micro sd
<Rambler44>
Well, assuming they were meant for tablets and mobile devices use those why would it have a regular SD one.
<KotCzarny>
original banana pi has fullsize sd slot
<Rambler44>
It costs about 10 dollars more than a Banana Pi to pick OLIMEX
<KotCzarny>
but you support real eu company
<KotCzarny>
with real support
<KotCzarny>
if you buy banana, all of the support you get is linux-sunxi community
<KotCzarny>
(at best)
<Rambler44>
10 dollars is easily wasted on four 0,5 L bottles of water anyway. The thing is that when I hit over 40 dollars I have to pay an import fee on 25% that is there to protect our industry.
<KotCzarny>
but you have tme in your country
<Rambler44>
What kinds of development boards they sell on our online stores? Hmm, I think they sell RPi 2
<Rambler44>
So theres that.
<KotCzarny>
and they probably sell it from local store/warehouse
<KotCzarny>
you can browse sbc category (dont know how is it translated to your language)
cptG_ has joined #linux-sunxi
<KotCzarny>
anyway, bbl
yann|work has quit [Ping timeout: 272 seconds]
cptG has quit [Ping timeout: 264 seconds]
nabblet has joined #linux-sunxi
avph has quit [Ping timeout: 245 seconds]
<nabblet>
hi, can you point me allwinner competitors that are more considerated towards the GPL?
<nabblet>
From wikipedia I get this list of competitors Allwinner Technology, MediaTek, Intel, Actions Semiconductor, Spreadtrum, Leadcore Technology, Samsung Semiconductor, Qualcomm, Broadcom, VIA Technologies and Amlogic
<nabblet>
also there is Rockchip but as I understand they're even worse than Allwinner
<nabblet>
I know this is probably off-topic here but I lack all that insider knowledge and might miss things during my research
<nabblet>
Rambler44: I know, that's why I am looking for competing products that are (more) compliant.
<Rambler44>
Also, whatever old inefficient Mali-200 they are shipping I'm pretty sure Allwinner at least would give you proper files to be able to hardware accelerate video... ...
<Rambler44>
Ah, no, that's ARM.
matthias_bgg has left #linux-sunxi ["Leaving"]
book` has quit [Ping timeout: 260 seconds]
IgorPec has joined #linux-sunxi
<nabblet>
Rambler44: "proper files"?
paulk-collins has quit [Ping timeout: 260 seconds]
<Rambler44>
I have a personal grudge as to why I stay away from Samsung CPUs, going by the way they treat people on their phones I doubt they would care much about whatever company buys their CPUs to use on development boards.
GeneralStupid has quit [Ping timeout: 260 seconds]
avph has joined #linux-sunxi
GeneralStupid has joined #linux-sunxi
<Rambler44>
They literally pushed patches on public channels that bricked the phone
<Rambler44>
Made it dead beyond easy repair
lemonzest has joined #linux-sunxi
<Rambler44>
Then they pushed a "fix" for their Samsung apps which involved disabling important security functions on my phone.
<Rambler44>
Going by the popularity of their phones I do hope they have improved something over the years.
book` has joined #linux-sunxi
<Rambler44>
If you care for GPU I think Broadcom deliver true open-source drivers for Videocore IV now.
<Rambler44>
nabblet
<nabblet>
Rambler44: i see that there are several instructions on how to build kernels for samsung devices/cpu ... but building such a kernel may involve binary blobs and the instructions don't highlight that info
paulk-collins has joined #linux-sunxi
<nabblet>
Rambler44: ok, I'll look that up
reinforce has quit [Quit: Leaving.]
reinforce has joined #linux-sunxi
akaizen has joined #linux-sunxi
<nabblet>
Rambler44: However, only a minor part of the driver was actually released as open source and all of the actual video acceleration is done using a firmware coded for its proprietary GPU, and which was not open sourced; https://en.wikipedia.org/wiki/VideoCore#Linux_support
<nabblet>
:(
nove has joined #linux-sunxi
<nabblet>
argh!
<Rambler44>
Well, at least you can use GPU compared to with Mali.
<nabblet>
The free licensed 3D graphics code was committed to Mesa on 29 August 2014
<Rambler44>
All GPU companies do this to protect their technology and there are many people which despise open-source
<Rambler44>
Mali got some really hardcore technology to protect by the way
avph has quit [Ping timeout: 245 seconds]
<Rambler44>
From my basic understanding I think they are using the older model on how GPUs are operating
<nabblet>
Rambler44: hardcore technology to protect <- as in hardware that is protect-worthy or hardware that is made to protect other hardware?
<Rambler44>
No, I was being sarcastic. I don't see what they have to protect there.
<Rambler44>
As a matter of fact it could have turned into a pretty big selling point if they opened up.
<Rambler44>
They are old GPUs
<Rambler44>
You think it's going to push 4K? Haha, whoever marketed that what a good joke.
<nabblet>
Rambler44: didn't the linux kernel just now include native support for raspi (and it's 3D stack therefore)?
<Rambler44>
I hope the eyes are enjoying 20MHz on HDMI standards not up to 4K.
<Rambler44>
20Hz*
<Rambler44>
As far as I can see the Arch ARM packages have headers for ODROID-C1 3.10.80-20, XU 3.4.104-2 and XU3 3.10.92-5. http://odroid.com/dokuwiki/doku.php?id=en:odroid-xu4
<Rambler44>
nabblet Not sure, I use mine as a headless server on kernel 4 and so far it does its job.
<Rambler44>
For support your best bet is RPi or OLIMEX at least from what I've heard.
<nabblet>
Rambler44: jep that's also my knowledge ... limited knowledge though ;)
<Rambler44>
I could pickup a RPi2 for $51 which is a good choice but I don't really like their power supply let alone the compontents, ethernet yaddi ya all routed onto the bandwidth of 1x USB 2.0?
<nabblet>
Rambler44: jep, that's why my choice was olimex ;)
<nabblet>
especially the "put it all on the usb bus" thing
avph has joined #linux-sunxi
<Rambler44>
That A20 model I got sent with SATA looks so promising I'm not sure I can say no to it even if I end up paying 20% fees and so on.
<nabblet>
Rambler44: I can understand you... also don'
<nabblet>
don't forget to order the power cord for the sata disk :)
<Rambler44>
But with my RPi I've been a little spoiled kernel wise
<nabblet>
I forgot it and had to forge my own cable :
<Rambler44>
4.3 didn't bother to update it as it runs stable.
<Rambler44>
Running much older kernels on production servers
<Rambler44>
They were never really meant to be powered off so
<nabblet>
Rambler44: That A20 model I got sent with SATA looks so promising <- promising in what aspect? what do you plan to do with it?
<Rambler44>
To use more than 100Mbps at least
<Rambler44>
I understand that a MicroUSB can't power a HDD, how expensive is the power cords I wonder.
khuey|away is now known as khuey
<nabblet>
Rambler44: the cord is not as cheap as you'd expect :P
<nabblet>
Rambler44: I can give you my documentation on making that wire (i think i should have it somewher) - at some point I think they swapped the red and black wire....
leviathancn has joined #linux-sunxi
<nabblet>
Rambler44: no, you should use a proper power supply if you're powering a hdd too
<Rambler44>
External drives with cabinet and everything are priced equally.
<Rambler44>
But the quality of whatever drives they put inside might be questionable.
<Rambler44>
Might also want to go into sleep mode to save power, could override with dhparams and so on maybe.
<nabblet>
Rambler44: I bught just the case and went with one of the low power western digital disks
<nabblet>
but as I said: proper power supply plus the cabel to power the disk from the olimex board is required
<Rambler44>
Why do I got a feeling putting it inside a server cabinet and powering the drive inside of there to connect it to the ARM server defeats the whole purpose of having it.
<Rambler44>
:)
<nabblet>
:P
<tkaiser>
Rambler44: You talked about 28nm process when mentioning H3 and A64. What's the 'source'?
rolo has joined #linux-sunxi
<rolo>
hello
<Rambler44>
Ah, the ugly 15K drives we got long ago in 2000-something. Just slide them in and they run.
<rolo>
i am trying to get audio working in A13 olinuxino WIFI board. I have enabled the in kernel ALSA modules (also have tried build in) but without success. I am using the sunxi 3.4.104-gd47d367 kernel compiled with gcc-linaro-arm-linux-gnueabihf-4.7-2013.04-20130415
<nabblet>
Rambler44: it was helpful talking to you. Now that I know what GPU to look out for I can focus on finding a compliant CPU :) Thanks!
<Rambler44>
tkaiser sorry if I am throwing false information in here. Might land at 40nm.
<Rambler44>
As far as I know the ARM Cortex-A7 which is also used in A20 was at that manufacturing process.
<Rambler44>
The one I have on the Broadcom CPU I have here that is.
<tkaiser>
Rambler44: We also have the information in our wiki: H3/A64 = 28nm. But in the meantime I have doubts
<Rambler44>
Hmm, I wonder...
<tkaiser>
Since Allwinner never mentioned the process for these SoCs.
<Rambler44>
According to an engineer I talked with BCM2836 goes with 40nm
<tkaiser>
According to TL Lim (one of the Pine64 guys) the A64 is 40nm. He asked Allwinner's PM for the SoC.
<Rambler44>
So it might be that the wiki contains false information then?
<tkaiser>
Yes, at least for A64 I assume I've been wrong writing 28nm in the wiki (based on some googling before)
<Rambler44>
Those CPUs are, however, a little stronger than the BCM CPU with 40nm.
<Rambler44>
In general the rumor has been that Allwinner CPUs appear to be stronger but when it comes to action it performs worse to other CPUs.
<Rambler44>
I think its due to the overheating throttling the CPU
<tkaiser>
Hmm... it depends. A83T with active cooling might be pretty fast even if it's just Cortex-A7.
<tkaiser>
H3 with a good heatsink and enough airflow can run constantly at 1.2 GHz on all CPU cores (cpuburn)
<Rambler44>
Since most get in touch with Allwinner through cheap Chinese tablets. By which I also did, then yeah.
<tkaiser>
When I did some measuring regarding integer performance both seem to perform identical when clocked identical
<tkaiser>
So me as a complete NOOB am thinking that the process (28nm vs 40nm) might make a difference how hoch clockspeeds might be and heat generation?
<Rambler44>
tkaiser, 1.2 GHz seems like a killer price for $15. If they just didn't screw up everything else.
<tkaiser>
Is there a way to judge from the outside in which process a SoC is made?
<Rambler44>
I believe all the Cortex A7 gets a 40nm for now since most of the newer CPUs such as the Broadcom and the A64 seem to have been confirmed 40nm through people working for them.
<Rambler44>
I see no reason A20 has 28nm when it was made before that model.
<Rambler44>
tkaiser Not much you can do when looking at the chip.
<Rambler44>
I tried staring at CPUs but no matter how much I look I can't quite put my finger on what process was used.
<Rambler44>
Much easier to just go by the SoC on CPUs that can be physically taken out
<Rambler44>
A64 looks like its going to be a horrible marketing mess and yet another kickstarted I'm not backing
<rolo>
anyone to give me an advice ? on the audio problem ?
<tkaiser>
Someone from Xunlong said they use a different voltage regulator on the One/Lite that provides 1.1V-1.3V and they chose 654-1200 MHz. I would believe these "1.6 GHz" are chinese marketing (or at least something I don't understand due to intercultural issues)
enrico__ has quit [Quit: Bye]
<Rambler44>
So they did come to sense
<tkaiser>
Maybe, until I've seen the fex stuff for the 2 new H3 boards from Xunlong I've no idea
<Rambler44>
Oh look at that board without source, all those Cortex-A7 with 40nm, but no - wait! Suddenly H3 has 28nm...
_massi has quit [Quit: Leaving]
<Rambler44>
A20 even gets a 55 nm rating here, but you know how much to trust wikipedia.
<Rambler44>
It's fun to see that wikipedia and sunxi-linux wiki gives completely different numbers. Maybe people are making them up.
domidumont has joined #linux-sunxi
<Rambler44>
tkaiser, well, I assume nobody cares as long as you can see the power efficiency / frequency. At least these values are true http://linux-sunxi.org/A80
avph has joined #linux-sunxi
<Rambler44>
I didn't realize Allwinner ever made anything without Mali
<Rambler44>
Oh, if you didn't get what I said about those values being accurate. That wiki page got a few question marks :)
akaizen has joined #linux-sunxi
vishnup has quit [Quit: Leaving]
vishnup_ is now known as vishnup
domidumont has quit [Remote host closed the connection]
lamer14538351316 has joined #linux-sunxi
domidumont has joined #linux-sunxi
domidumont has quit [Remote host closed the connection]
domidumont has joined #linux-sunxi
tkaiser has quit [Ping timeout: 245 seconds]
lamer14538351316 has quit [Client Quit]
<KotCzarny>
Rambler44: i bought meanwell rd-65a to power my sunxi device(s) + hdd + monitor + whatever
<KotCzarny>
also, source code wise i think broadcom is one of the worse offenders (blobs everywhere)
<KotCzarny>
Rambler44: also nice thing to buy is li-ion battery (i bought 18650 and a holder with cable and plug)
<KotCzarny>
those boxes have power regulators capable of charging/using li-ion without additional circuitry
<KotCzarny>
insta UPS
<Rambler44>
For how long?
<KotCzarny>
they eat ~1.5-3W == ~0.2-0.5A
<KotCzarny>
which means 2600mAh would last for a few hours
<KotCzarny>
0.2-0.5A @ 5V, li-ion is 3.7V so cut ~1/3 of the calculations for rough estimate
<KotCzarny>
also i have mounted my device in pc case (along with hdd, battery, gigabit switch and power brick)
<Rambler44>
What if you really don't use all these resources
<KotCzarny>
then you can put it in acrylic case and have tiny computer with blinkenlights
vishnup has quit [Quit: Connection closed for inactivity]
<Rambler44>
Like tkaiser said here, he could have it running with all cores max at 1.2 GHz but with good air flow
<Rambler44>
Let's say I put a heatsink on it, would it defeat the purpose to have an acryllic case over it? Hmm.
<KotCzarny>
let me put it this way, i have bpi-r1 (a20 dual core), when it was running in the acrylic case i unscrewed the screws a bit to leave ~1mm space and put it vertical
<KotCzarny>
downclocked to 800mhz it was going to ~37C max under full load
<KotCzarny>
(it also has riny radiator on cpu)
<KotCzarny>
and you still didnt say what purpose you have in mind for the box
<Rambler44>
Oh I experiment a lot with them, that's why microsd was perfect, a lot of different stuff it can load up.
<KotCzarny>
well, keep in mind uboot can boot from network
<KotCzarny>
i use mine as home nas/router/audio box, and some wifi on demand
sunxi_fan1 has joined #linux-sunxi
a|3x has quit [Ping timeout: 256 seconds]
<Rambler44>
Well, at home I got my 300 MHz router still going strong, running Tomato by Shibby.
<KotCzarny>
it replaced my earlier power hungry 40W x86 box
<Rambler44>
I tried compiling OpenWRT multiple times on my Latinq 500 MHz box but the horror was unbearable.
<KotCzarny>
i think there is openwrt for a20
<Rambler44>
Well, what do I expect, support for a Latinq CPU?
<Rambler44>
At least I should be happy I got it booting and somewhat functioning.
<Rambler44>
unstable
<Rambler44>
KotCzarny why run OpenWRT when you can run Linux though?
<Rambler44>
Unless, of course, it's purely for router.
<KotCzarny>
openwrt is for people who cant setup linux as a router
<Rambler44>
Which doesn't make sense to me because whatever Wi-Fi amp you can give it that's going to be a dumb idea.
<KotCzarny>
or just want ready made product
<Rambler44>
Nothing bad about OpenWRT, I wished so strongly to have it run on my 500 MHz latinq host but Tomato on this 300 MHz has to cut it for now.
<KotCzarny>
keep in mind that on allwinner cpus most disk/network data eats cpu
<KotCzarny>
which means gigabit speeds arent quite reachable
<Rambler44>
I know a guy was discussing how CPU was being eaten up on a video streaming site
<Rambler44>
Well, if I were to use it at home for example to play with.
<KotCzarny>
if you plan on using it as a webcam encoding box, rpi is a batter choice, but as a storage a20 wins
<TheLinuxBug>
KotCzarny: how did you wire up the battery on that box and what is the general cost of one? I wouldn't mind being able to set something like that up, I assume by what I read it can be powered from AC or the battery?
<Rambler44>
My router at home is only 100Mbps, end of choice. That is because I wanted to save money and I can't get my gigabit Latinq up and running stable yet and it's hard to get help (largely on my own)
<Rambler44>
So if I were to use if as for example NAS
<Rambler44>
I doubt it would bottleneck at sending files at 100Mbps
<KotCzarny>
TheLinuxBug: 18650is ~5usd, holder is ~2usd
<TheLinuxBug>
The rd-65, sorry
<Rambler44>
A RPi 1 might do it
<TheLinuxBug>
or did I misunder stand
<KotCzarny>
TheLinuxBug: rd-65 is just power adapter with 5 and 12V outputs, i just cut microusb cable and wired it there
<TheLinuxBug>
KotCzarny: does that battery come with the circutry needed for passive charging? or was it the rd-65 that did?
<KotCzarny>
and the battery is connected to bpi
<Rambler44>
I remember I tested out the 11 network cables I got wired up at home, even if they were cat 5 all of them passed 1 Gbit/s test except one
<Rambler44>
I somehow have a cable that doesn't fit Cat 5 standards and has only 2 pairs, it handles only 100Mbps
<Rambler44>
Labeling stated "FOXCONN"
<TheLinuxBug>
KotCzarny: directly connected to it? Where did you wire it on the BPi
<KotCzarny>
TheLinuxBug: keep in mind that charger on the board is quite dumb in the sense it keeps battery at 100% all the time, didnt find the way to set charging thresholds
<Rambler44>
Sounds like the perfect way to kill batteries
<KotCzarny>
yes
<KotCzarny>
still, 5usd is not much and i needed it mostly for those power loss events
<Rambler44>
It's weird, back in Russia power went off like a daily occurrence. Been living here for three years and only power loss we had was because the line got overloaded so they had to upgrade it or something.
<KotCzarny>
yes, cables in my flat are old, i often trigger overloads because of it
<KotCzarny>
(ie. washing machine running and i forget about it and try to boil water etc)
<KotCzarny>
another use would be just switching cables/rooms without reboots
<KotCzarny>
or even going full mobile (with added touchscreen)
<KotCzarny>
TheLinuxBug: keep in mind i havent tried powering bpi-m1 via that bat-lead, dont know if its wired properly
<Rambler44>
Got an old house too, some have been upgrade to switches others are those with tiny cables that can be cut at overload and you need to actually plug in a new one for it to work.
<KotCzarny>
TheLinuxBug: 18650 holder with a cable saves you from soldering the battery (which also might be fatal for it)
<Rambler44>
A lot of the power plugs have no uhm earth? Not sure how to translate.
<TheLinuxBug>
ground
<TheLinuxBug>
ya
<Rambler44>
grounding yeah
<TheLinuxBug>
KotCzarny: make sense
<KotCzarny>
and also makes it easier to keep it in place
<TheLinuxBug>
(thumbsup)
<TheLinuxBug>
thanks!
<KotCzarny>
update the wiki if those bat solder points work
<KotCzarny>
(also, usual disclaimer: i didnt try it on bpi-m1 and i guarantee nothing, andcan't be held responsible for all injuries/damages)
<KotCzarny>
my m1 is used only as an audio box at gf's place
<Rambler44>
Be like me and short circuit, struggle to desolder your ugly work and so on.
akaizen has quit [Remote host closed the connection]
<KotCzarny>
another cool idea is tiny audio amp that run on 5V with 2x2W output power
<Rambler44>
I remember one time I became so mad I put both my hands with Phillips Jewelers screw to short to circuits and used my foot to push a power button over doing all that solder work.
<KotCzarny>
with added battery and some buttonsand external speaker could replace all those bt speakers
<Rambler44>
two*
ssvb has joined #linux-sunxi
<Rambler44>
That reminds me, there is a lot of bad DAC/amp out there. With pretty decent cheap Realtek chips going around isn't it possible to make a cheap but good enough USB-OTG/WASAPI etc external one?
<Rambler44>
Most of the time internal audio picks up processor noise, while you never notice it I have done many times. On this PC the Realtek Linux drivers picks up a great deal of noise (almost none on Windows)
avph has quit [Ping timeout: 240 seconds]
<Rambler44>
It might be noise cancellation technology or bad drivers, who knows. Maybe download an un-free driver.
<Rambler44>
When I asked my friend this was what he came up with as external audio
<Rambler44>
Yeah, he showed it to me. I heard the superb music quality he had on his equipment. And the tube? Oh, that was extremely hot. Would never work for any low-power solutions.
<yann|work>
any idea what could cause glmark2 score to be half the expected value (on h3)
alexxy has quit [Ping timeout: 250 seconds]
<yann|work>
since it's a dual-core gpu, i'm wondering if it woudn't be using only a single core
<Rambler44>
Isn't H3 quad-core?
<Rambler44>
Ah, ueah, GPU.
<jelly>
Rambler44: a cheap c-media CM106 does better wrt noise levels than most onboard audio cards
<KotCzarny>
Rambler44: i had a problem when both my a20 box and pc were connected via common ground, and with sata drive connected lots of noise was in the signal
<KotCzarny>
yan: bad compilation/detection? wrong driver?
<Rambler44>
jelly I saw those floating around, I have one audio place that picks up a heck ton of audio.
<Rambler44>
I saw two cheap Chinese cards that were interesting
<Rambler44>
That one and another cube shaped one with a DAC from Texas Instruments
<Rambler44>
Well, cuboid, with USB 1.0
<KotCzarny>
grounding matters
<Rambler44>
USB 1.0 does not sound tempting.
<KotCzarny>
and metal case
<KotCzarny>
usb with digital (optical) output?
bonbons has joined #linux-sunxi
<KotCzarny>
but onboard audio also isnt bad as long you dont have anything noisy connected/close
<Rambler44>
So now I've also seen ceramic heatsinkgs.
<KotCzarny>
AND, remember you have hdmi output with audio
<Rambler44>
Ah, no, this PC works fine on Windows. It's a ALC3232
<jelly>
bought it to work around bad thinkpad audio, but it works well enough with a cubieboard as well
<Rambler44>
Bad thinkpad audio?
<ssvb>
apritzel: did you have a chance to test my U-Boot patches on your Pine64?
<Rambler44>
ALC3232 is what they put on this thinkpad.
<apritzel>
ssvb: sorry, not yet
<apritzel>
I need to fix that DT first
<Rambler44>
The amp appears to be derpy though
<apritzel>
ssvb: you took the DT from my repo for U-Boot as well, right?
<Rambler44>
Sometimes it happens that it takes 5 seconds to kick in before the audio gets louder
<Rambler44>
I wouldn't use Lenovo if it wasn't for certain conditions.
<Rambler44>
Ah, so you guys got a Pine64. I heard that mainline was something you wouldn't expect anytime soon.
<Rambler44>
Was thinking about ordering but it looked like they had no clue what they were doing
<ssvb>
apritzel: yes, but U-Boot does not rely on DT too much, so even broken DT may work without problems
<KotCzarny>
Rambler44: for now there is not even legacy kernel running linux
<KotCzarny>
(on pine64)
<Rambler44>
Exactly
<Rambler44>
But might prove to be a nice Android Lolipop box?
domidumont has quit [Ping timeout: 246 seconds]
<apritzel>
KotCzarny: ??
<Rambler44>
hEHE
<KotCzarny>
apritzel: is there?
<Rambler44>
Didn't Allwinner ship their kernel like always?
<apritzel>
yes, they have an Android kernel, 3.10.65
<apritzel>
plus Linaro stable + Android + Allwinner patches
<Rambler44>
Oh no, the horror of that kernel.
<Rambler44>
At least its not overclocked to 1.6 GHz
<ssvb>
apritzel: I would say that it's probably best to make things work fine in 32-bit mode first (a sane DT and basic drivers) and only then move to 64-bit experiments
<apritzel>
but if I find this bloody DT issue you will see mainline support still this week
ricardocrudo has quit [Remote host closed the connection]
<apritzel>
ssvb: I agree
<apritzel>
ssvb: actually a good test for my DT ;-)
<Rambler44>
Wasn't 64-bit ARM just a marketing thing. Well, the RAM support which isn't really being taken advantage off anyway is there.
<ssvb>
apritzel: if you want to boot the kernel, just disable PSCI in the DT because U-Boot does not provide it yet
<apritzel>
Rambler44: -ENOPARSE
<ssvb>
apritzel: and this UART thing is a mess :-)
<apritzel>
ssvb: in what respect? With the clocks?
<ssvb>
apritzel: whatever is the problem, it would be great if the serial console just worked :-)
<apritzel>
ssvb: do you plan to enable the PSCI support in U-Boot? Or can you wait for ATF?
<Rambler44>
Marketing guys would love that somebody fixed them mainline
<Rambler44>
It's not really a good kernel they are advertising with at the time being you see
<ssvb>
apritzel: PSCI does not give us anything at the early stage of development
<apritzel>
agreed, but UP is just annoying for that kernel compile test :-D
vagrantc has joined #linux-sunxi
<Rambler44>
At least they gave away development samples+
<Rambler44>
?
<apritzel>
why does actually everybody thinks that the hardware vendor provides the kernel?
<KotCzarny>
apritzel: because most people considering those boards are consumers and not oems?
<ssvb>
Rambler44: "the RAM support which isn't really being taken advantage of" - I think somebody is confusing the data bus width with the address bus width, and the virtual address space is an entirely different thing too
akaizen has joined #linux-sunxi
<ssvb>
Rambler44: A64 is just fine
<Rambler44>
Oh, yeah, I forgot. Wasn't it similar to H3
<apritzel>
arm64 is quite a different architecture
<ssvb>
Rambler44: and FYI, DRAM accesses are done in large bursts (the whole cache line is allocated or evicted) on both 32-bit and 64-bit processors nowadays
<apritzel>
the advantages of 64-bit are:
<apritzel>
a) big virtual address space (no highmem trickery)
<apritzel>
b) more and wider registers
<apritzel>
c) future proof
<apritzel>
you wouldn't wonder how many people don't take 32-bit serious anymore
* apritzel
had to learn the hard way when working for Calxeda
<apritzel>
*would* wonder, of course
<ssvb>
apritzel: BTW, both Cortex-A53 CatA errata seem to be only affecting the 64-bit mode according to the errata list
<Rambler44>
Well, 64-bit ARM only took off after Apple put it to use.
akaizen has quit [Read error: Connection reset by peer]
<Rambler44>
Then Snapdragon made a range of CPUs with overheating issue in a rush to match marketing demand.
akaizen_ has joined #linux-sunxi
<apritzel>
Rambler44: arm64 hasn't taken off at all yet ;-)
<apritzel>
but the Pine64 is a good sign that that could change
<ssvb>
IMHO having a 32-bit userland makes perfect sense for the end users right now, especially if the RAM size is small
<Rambler44>
These boxes never usually come with over 2 GB RAM
<apritzel>
ssvb: from a pragmatical point of view I totally agree
<apritzel>
ssvb: it's just this I'd love to see the Pine64 as an arm64 enabler
<apritzel>
so that more people can test our patches here ;-)
<apritzel>
hardware availability is the most common question I got about arm64
<Rambler44>
My 1.7 GHz snapdragon phone got 2 GB RAM, all powered with 32-bit 28 nm ARMv7
<ssvb>
I have a Playstation3 and run Linux on it with a 32-bit userland, even though the CPU supports 64-bit mode too :-)
<libv>
apritzel: as i said earlier today, this smells like arm virtualization all over again
Netlynx has quit [Quit: Leaving]
<ssvb>
this reduces the memory footprint quite a bit, which is very important when you only have 256MB of RAM :-)
<libv>
wait and see how long this goldrush lasts
<jemk>
Is there something like x32 possible with arm64?
<apritzel>
ssvb: oh yes, in this case
<apritzel>
jemk: oh oh
<jemk>
32bit pointers but still more/wider registers, could be useful for low ram
<apritzel>
jemk: it's work in progress
<apritzel>
version 28 or so on the list
<apritzel>
has been discussed for ages
<apritzel>
google for ILP32
<Rambler44>
I remember people were running 32-bit with virtualization when the virsual servers had 2 GB RAM or so maybe less
<WarheadsSE>
Sorry, but ROFL: "15:41:29 Rambler44 | Well, 64-bit ARM only took off after Apple put it to use."
<Rambler44>
Said something about it performing better than 64-bit templates, so providing both 32 and 64-bit ones because people want both.
<jemk>
ah, ok
<Rambler44>
WarheadsSE so you got it.
<apritzel>
jemk: the API isn't fully agreed on
<Rambler44>
It made me wonder if the host machine is 64-bit why do people want to run 32-bit on virtual instances.
<apritzel>
and the benefit is probably smaller than on x86
<apritzel>
and you see where x32 is today ;-)
<jelle>
hmm speaking about arm virtualization, the orange pi should have that right?
<Rambler44>
Almost bought Pine64 when it came out, as a matter of fact it was so interesting this got me interested in Allwinner again.
<apritzel>
Cortex A7 has virt
<jelle>
apritzel: ko
<apritzel>
but the firmware may play bad on enabling it
<jemk>
apritzel: just wondered whether there is anything like that
<Rambler44>
Since H3 got cheap devices like OPi PC hanging at it was just thinking software support from community-side would have been wider.
<ssvb>
apritzel: BTW, with an assembly coder hat on, I'm not very thrilled about x32 and various big endian offshoots
<ssvb>
apritzel: it means having additional configurations to test and support...
<apritzel>
ssvb: oh, yeah, I see what you mean
* apritzel
has to run now
<jelle>
apritzel: hmm so the orange pi pc should have it, hrrm I should try that some day
<Rambler44>
It comes with nice overclocking too
<apritzel>
jelle: yeah, ping me again, KVM on ARM/ARM64 is my actual day job ;-)
<jelle>
apritzel: oh haha
<jelle>
well first want to finish the pmic u-boot code for the orange pi pc now
<jelle>
my collegue uses lxc on his bananapi and it works awesomely
paulk-collins has quit [Ping timeout: 256 seconds]
apritzel has quit [Ping timeout: 248 seconds]
<ssvb>
jemk: do you have the right USB cable for using FEL boot on your Pine64?
paulk-collins has joined #linux-sunxi
<jemk>
ssvb: no, have to solder one first i guess. who had this silly idea...
<Rambler44>
Same guys which wrote "4K HDMI" on the specifications?
<jemk>
fortunately i have enough old usb cables laying around
<Rambler44>
“This board one of the most impressive, inexpensive options for anyone looking to implement high resolution video in their project.” – MAKE Magazine
<Rambler44>
hmm...
<Rambler44>
Mali-400-MP2 sure managed to impress
<ssvb>
Rambler44: what is your problem exactly?
<Rambler44>
It's because I really wanted to buy it, I really did. But I didn't back it in the end because I didn't like how they advertised it do have a good GPU.
<ssvb>
Mali is only used for the user interface eye candy in media players (XBMC/Kodi being a notable one), the video decoding heavylifting is done by a different hardware unit in the SoC
<ssvb>
the other media players don't even need Mali at all for the HD video playback
<Rambler44>
CPU extension?
<Rambler44>
No, wait, GPU extension.
<ssvb>
jemk: do we have a good introductory page in the linux-sunxi wiki about cedar?
Nacho has quit [Remote host closed the connection]
<Rambler44>
I'm not quite sure how video works when all comes to all. I thought that if it had a chance to do 4K on HDMI 1.4 it would be on Android and lower MHz
<Rambler44>
What I do know is that it needs CPU for playback
<Rambler44>
Most ARM devices die at H.265 and/or 10-high-bit-depth or so.
Nacho has joined #linux-sunxi
<Rambler44>
I think my problem is that I've been using RPi am I right?
<jemk>
h3 decodes h.265 4k perfectly fine with near zero cpu load
<ssvb>
10-high-bit-depth is another thing, the hardware accelerator either supports it or not
<Rambler44>
So because Videocore IV did everything, VPU and all. I'm thinking Mali does everything on Pine64 but that's not a fact.
<libv>
Rambler44: i got to lart one guy at my current customer today for labelling vivante as "graphics driver"
<ssvb>
jemk: maybe you or rellla could create an introductory wiki page about video decoding acceleration for the end users?
<Rambler44>
jemk What, that sound too good to be true- Bit-starved or? And for higher bit-depth?
<Rambler44>
Is that H3 running Linux
<libv>
people are too used to their image of a discrete graphics card doing everything in one big module
<Rambler44>
A wiki page sounds like a good idea.
<jemk>
it would be a good idea, but i'm really bad at writing such things
<Rambler44>
When I started thinking about graphics I got my first brand-new 256MB NVIDIA card.
<Rambler44>
Otherwise it was more like oh it displays that's good enough right.
<ssvb>
jemk: otherwise people will keep coming and have all the weird ideas about it, especially considering that the Kodi camp keeps spewing anti-allwinner propaganda and confusing the users
<jemk>
Rambler44: 10bit isn't possible as far as i know, but 8bit works well with libvdpau-sunxi in linux
<Rambler44>
I have gotten stuffed a lot of Anti-Allwinner in me.
<ssvb>
Rambler44: does RPi support 10bit?
<Rambler44>
Inefficient manufacturing process, Violated and abandoned GPL with Team Kodi, CPU performing much worse than its theoretical clock and so on.
<Rambler44>
ssvb No, haven't bothered to try on the Videocore IV, I assume it would go CPU.
<Rambler44>
Which it doesn't have enough of.
<ssvb>
so how is it H3 fault?
<jemk>
there is a small hint that they might planned to support 10bit, but at least the current drivers don't support it
<Rambler44>
No, it goes back to my first statement where ARM wasn't designed to playback 10-bit video. And why would you need 10-bit video in the first place?
<Rambler44>
If anything H.265 is much better to have supported.
<Rambler44>
If it can playback that then there is not much more to ask from ARM SoC
<ssvb>
Rambler44: "Violated and abandoned GPL with Team Kodi" - Team Kodi just has some personal grudge against Allwinner, so you need to take their statements with a grain of salt
<ssvb>
Kodi is not the only media player around, there are other alternatives which just work
<Rambler44>
Well, let's say I order myself a H3 to play video. I might very well do it but I have a 1 GB ATI card sitting there to do that already.
<Rambler44>
Would I by any chance pick Linux and try to playback with mpv, or Android and attempt to playback with MX Player.
<Rambler44>
From the standpoint I had before Android should have been the most logical pick if you want to playback video.
<nove>
recent rockchip socs have hardware h264 10bit decoding
<Rambler44>
Not that I'm very fond of using Android to playback video.
<Rambler44>
What a bad idea.
<Rambler44>
nove looks like its stating even h265 10 bits
avph has quit [Ping timeout: 245 seconds]
<Rambler44>
Well, one last question if anyone got some time before I go. SSL library (AES), RC4-MD5, salsa 20, chacha20. If I want to run a VPN with at least (some) encryption. Would it be smart to check if the CPU had crypto extensions for either of these before picking one?
<Rambler44>
At least I managed to clear out the fact that I thought h.265 would kill ARM CPUs
<nove>
if you are look for 10bits, because you like to have smooth gradients without be forced to use dithering, h265 8bit will not help
avph has joined #linux-sunxi
<Rambler44>
Well, h265 isnt a good idea since x265 isn't as well polished. But they should be praised for improving it as much as they have already in the course of last year.
<Rambler44>
I fear it won't be that adapted though as the patent claims are ready.
<Rambler44>
People will be staying on the standards for a time to come.
<Rambler44>
What seemed interesting was the space-saving potential.
<Rambler44>
But is it a good idea to bit-starve your encodes at all, that is the question.
ricardocrudo has joined #linux-sunxi
<yann|work>
libv: wrt to your comment on why adding mali r4p0 (or newer) support: looks like r5 would add arm64 support, and r6 Mali-470
paulk-collins has quit [Quit: Quitte]
<libv>
yann|work: when was this comment made?
<libv>
mali utgard has led an extraordinarilly long life
<libv>
yann|work: even then, this was before a64 also got a mali utgard
<ssvb>
yann|work: supporting new blobs needs hours of development and testing time, and nobody is lining up, readily willing to pay $$$$ for this work
<ssvb>
yann|work: people keep coming and making all kinds of unreasonable demands without giving anything in return
<ssvb>
yann|work: let me look up a mailing list post about this...
<KotCzarny>
ssvb: maybe because people think linux-sunxi is some magic allwinner department?
<Rambler44>
But the blobs work all fine with their Android build right? That's what they intended it for in the first place.
<rellla>
ssvb: wiki page is a good idea. i thought about a introducing page, too, including
<Rambler44>
Allwinner couldn't ever have dreamed of affording to pay for mainline support
<yann|work>
I don't know how much different they put between android and linux drivers, and the fact the userspace part is a blob does not help to get an idea...
<rellla>
What is cedarx, cedrus/re_..., mali, g2d, display ...
<KotCzarny>
cedrus has its page
<rellla>
ssvb: and if/why one needs that for kodi, mplayer and the whole media family.
<rellla>
ssvb: there is my kodi-page redo/update outstanding still...
mosterta has joined #linux-sunxi
<rellla>
KotCzarny: sure. People lurking around here know that, but read up the opi/bpi/cubieboard forums, you get sick by reading...
<KotCzarny>
rellla: after being here for some time, i've got to conclusion that opi/bpi forums are superfluous and if you really need/want something sunxi related its just linux-sunxi wiki or irc channel
<Rambler44>
So, if Mali does nothing - where does the video decoding happen on H3.
<rellla>
People don't know the difference mali and cedar. How should they get the difference between cedarx and cedrus?
<mosterta>
yes, unfortunately, there is less to little feedback on my fork.. or shall I say fortunately, since I discovered some bad bugs... and there could be still an issue with memory bandwidth, which I have to look at it again..
<rellla>
I'm away for some time now, but i hope to find some time for wiki work after that, if the task is still open then.
<Deskwizard>
mosterta: if you want a hand with testing, I've got an A20 thats isnt used atm
<rellla>
mosterta: nah. it would be better if a bright mass had found that issues :p
iamfrankenstein has joined #linux-sunxi
tmeyer has joined #linux-sunxi
<mosterta>
rella,I agree that would have been better..
<tmeyer>
mosterta: Hi, can I ask you some stuff about your kodi fork?
<rellla>
first we should make the firk more present in the wiki. and drop/hide that misleading actual kodi site.
<mosterta>
Deskwizard: I always can need some one to work at the issues? Do you want to test or can you develop as well..
<Deskwizard>
mosterta: well, developpement wise I won't lie to you, I'm a bit (read: very) limited ... I've played around with kernel stuff and android stuff mostly, nothing too involving I'm afraid
<mosterta>
rella: I also discovered a bug in sunxi-vdpau yesterday when watching a stream of ARTE, there are some artefacts in the AVC H264 profile.. But I didn't have the mood to look at it closer..
<rellla>
^jemk :)
<Deskwizard>
mosterta: with a bit of learning I could probably manage, but I'd have to have a deep look at the sources to see how it looks
<tmeyer>
mosterta: I got your code compiled but kodi crashes when playing any video
<mosterta>
Deskwizard.. which A20 do you have? I have a bananapi and could provide the required pieces in binary, otherwise you would have to compile by yourself..
<tmeyer>
somewhere in your vdpau-sunxi fork
<mosterta>
tmeyer: do you have a backtrace, or could you create one for me?
<Deskwizard>
mosterta: olinuxino-a2-micro-4gb, compiling it isn't usually an issue though
<Deskwizard>
/s/it/
<tmeyer>
Not at hand right now, but I can create one...
* rellla
looks at the most viewed wiki pages. It's clear that sth has to be done, to push the cedrus category pages up. And the 14. place of kodi makes me feel, that people are generally in it.
<Rambler44>
A while back I remember trusting the Kodi wiki when it said all Allwinner SoCs have no or question mark to support video decoding.
<rellla>
*intetested in
<Rambler44>
So if there are links to that site on here I would reconsider them.
<Deskwizard>
rellla: I agree, I wouldn't have searches for that long if the Kodi port was on wiki
<Deskwizard>
and what Rambler44 said as well
<Deskwizard>
since they are the first link that pop up, a lot of people wont dig further and just consider no Kodi on sunxi
<Deskwizard>
I saw you guys talk about it here, so I digged ;) but yeah, what I said
<Deskwizard>
links*
<rellla>
Unfortunately the wrong solution atm. Binary based. People get satisfied and drop their interest in the main goal. We have to drop that site or archive it somewhere.
<Deskwizard>
rellla: yeah I see what you mean, I totally agree with you.
<Rambler44>
I remember that somebody said I was false at no 10-bit support for ARM. This from Kodi wiki might have had something to do with why I thought so
<Rambler44>
H.264 Hi10P - No hardware decoders for Hi10P (High 10 Profile, also known 10-bit H.264) currently exist for any ARM hardware. Most H.264 Hi10P videos will not play smoothly as most ARM CPUs are just not powerful enough to software decode it fast enough.
<mosterta>
tmeyer: thanks, I will have a look at it
<tmeyer>
Thanks
<Rambler44>
Not that its a good idea to software decode it
<Rambler44>
But it works
<tmeyer>
Btw, there seems to be a kodi port for H3 but it's based on AW mediacodec stuff
<tmeyer>
came across this in the orangepi forums
<rellla>
tmeyer: that is, why i said "sadly" ...
<Rambler44>
Orane Pi forums, where the community builds are still OC at 1.6 GHz
<Rambler44>
Must contain a lot of reputable people going by that fact.
<tmeyer>
I tried to backport this to A20 but it didn't work
<tmeyer>
it's based on disp2 driver, too different from original disp
Rambler44 has quit [Quit: Page closed]
premoboss has joined #linux-sunxi
<mosterta>
tmeyer: have you switched "off" VDPAU mixing in kodi? it should be switched off, since only video surfaces are currently supported.
<tmeyer>
let me check...
ricardocrudo has quit [Ping timeout: 250 seconds]
<mosterta>
i can send you also my guisettings.xml file
<tmeyer>
VDPAU and Mixing was on. Should I switch off Mixing or both?
<mosterta>
you should keep VDPAU on, but mixing should be off.
<GeneralStupid>
i use 1.1 Mhz and a fan (temp controlled)
<GeneralStupid>
Ghz
<GeneralStupid>
kodi on orangepi: do you talk about android or linux?
<tmeyer>
Yes! It's working now!
<tmeyer>
Many thanks @mosterta!
<mosterta>
very well :)
<mosterta>
tmeyer: as formats also XVID, VIDX should be supported, but this currently not working as I just recently discovered.. but you can try the rest..
bonbons has quit [Quit: Leaving]
afaerber has quit [Quit: Ex-Chat]
<tmeyer>
Also, HD stuff doesn't work for me (1080p)
<GeneralStupid>
tmeyer: iam using a combination of mpv, scripts and tvheadend...
<tmeyer>
crashed with "could not allocate ump buffer!", then segfault
<mosterta>
what does "doesn't work" mean? which format H264 or MPEG4 or isn't shown anything or is there just a lag in the movie?
<GeneralStupid>
where can i get it?
<tmeyer>
maybe my own fault, I'm using a custom kernel...
<Deskwizard>
tmeyer: wasnt that like in red in his post? :P
<mosterta>
ups, my kernel patch had 190MByte for UMP, so there must be lots of UMP buffers.. unfortunately there is no statistics in UMP driver (yet)..
<mosterta>
tmeyer: so which format are working for you? 720p?
<tmeyer>
mosterta: codec for 1080p is h264, 720x576 h264 works...
<tmeyer>
let me check 720p...
<mosterta>
what hardware is it? how much memory do you have?
<tmeyer>
Cubietruck, 2G of ram
<mosterta>
1080p should work in general es well, I have tested it here with the big bunny video(s) and some others.. Which resolution do you have set up (1920x1080) or (1280x720)? The first requires much more memory bandwidth and is much more trickier than the later..
<mosterta>
it is not optimized for 1920x1080..
<GeneralStupid>
tmeyer: do you have a link to the allwinner board?
<tmeyer>
The original Cubietruck, not the plus model
<GeneralStupid>
tmeyer: i meant to the orangepi board
<tmeyer>
mosterta: kernel log has "Failed to allocate needed CMA memory 1597440 bytes" in it after trying 1080p
<tmeyer>
I should check my kernel config...
<mosterta>
yes, please check it how much you have configured..
<GeneralStupid>
tmeyer: you said: Btw, there seems to be a kodi port for H3 but it's based on AW mediacodec stuff, came across this in the orangepi forums
<GeneralStupid>
i would be interrested in that link…
<tmeyer>
CONFIG_CMA_SIZE_MBYTES is set to 192
p1u3sch1 has quit [Ping timeout: 276 seconds]
<mosterta>
tmeyer: CONFIG_CMA_SIZE_SEL_MBYTES=y should be set as well..
<mosterta>
then there is another issue.. you could try to increase a bit more, but more than 256MB does not make sense, Cedar HW is restricted to physical addresses < 256MB.
afaerber has joined #linux-sunxi
<GeneralStupid>
tmeyer: looks fine, did you extract Kodi from that openelec built?
<tmeyer>
mosterta: okay 720p h264 plays also fine, do I need to allocate more RAM for CMA?
<tmeyer>
GeneralStupid: Not really, I took the patches that add the Allwinner codec stuff from the source then adapted it for Cubietruck
<mosterta>
as I have written.. you could try to increase to 256MB.. more is not supported by Cedar HW.. But I think there is a different issue.. I have testet with 1080p and didn't see this issue..
<tmeyer>
GeneralStupid: but it didn't work well...I'm not very good at coding
ricardocrudo has joined #linux-sunxi
<mosterta>
But with 1080p you currently have the issue with dropped frames.. probably the memory bandwidth is limiting and I need to support output surfaces in vdpau-sunxi.. hope that will improve the situation..
<tmeyer>
mosterta: I'll try your kernel, if it works, then I know that my kernel build has some issues...
<GeneralStupid>
tmeyer: but you try to fork that for A20... i have an H3
<tmeyer>
mosterta: can you post your kernel boot parameters?
<mosterta>
tmeyer: here they are: console=ttyS0,115200 consoleblank=0 console=tty0 disp.screen0_output_mode=1920x1080p50 hdmi.audio=EDID:0 root=/dev/mmcblk0p2 rootfstype=ext4 elevator=deadline rootwait
<tmeyer>
GeneralStupid: Yes, that's what I did. If you have a H3, then you should be able to use that port without major hacking
<GeneralStupid>
but i dont want to use openelec
<tmeyer>
mosterta: Thanks a lot. I think I have an idea whats wrong. I'll test it when I have some more time.
<tmeyer>
GeneralStupid: Why not? I really like it compared to all those TV boxes running Android.
<GeneralStupid>
i like my debian :)
<tmeyer>
Oh, okay :)
<GeneralStupid>
which kodi version is it?
<mosterta>
tmeyer: welcome.. get in contact with me if you have further problems..
<rellla>
mosterta: i needed ~120MB at most with VDR at 1080i...
<tmeyer>
mosterta: Thanks for the help. Finally we can kick out the ancient and broken AW binaries and get something decent :)
<mosterta>
I tried to get the tool from ARM with kernel statistics working (DS-5 + gavor), but did not succeed.. But I didn't spend that much time up to now..
<rellla>
mosterta: look into my staging_mem branch. it's based on jemk code
<rellla>
who wonders ;)
<tmeyer>
GeneralStupid: looks like 16rc1
<mosterta>
rella: which staging_mem? at sunxi-vdpau or kernel?
<mosterta>
I can't find in sunxi-vdpau..
<GeneralStupid>
thanks
<GeneralStupid>
iam already trying to get it to work
<mosterta>
so you calculate from the dumped bytes the memory consumptiion?
<rellla>
Sadly not for ump :p
reinforce has quit [Quit: Leaving.]
<rellla>
Yes, everytime ve_malloc allocates, it should output the upper end of the reserved cma
<rellla>
The max. used cma block.
<mosterta>
ok.. then I did get you wrong.. you are talking about the highest memory consumption, while I thought you are talking about the memory bandwidth ~)
<rellla>
Yeah, consumption.
<rellla>
In times you needed to reserve mem, mem is worthy...
<mosterta>
I would prefer to add some sysfs variables into UMP driver and have a current/max value + maybe reset values..
<mosterta>
CMA only reserves the memory when it is actually needed.. if it it not needed, then kernel can use it for other purposes..
tmeyer has quit [Quit: Need some sleep now.]
<rellla>
Yes, but on module probe time, cedar_dev takes hardcoded 80MB in sunxi-3.4 atm.
<rellla>
That is too less if you buffer a few surfaces with libvdpau-sunxi
<mosterta>
only is you dont't switch to CMA in cedar_dev..
<rellla>
I do, but i use 190mb atm, which is too much still
<mosterta>
there is a kernel config parameter to switch to CMA mode..
<rellla>
Either this one, or without cma but the ve_reserve_mem kernel param
<rellla>
don't remember, if the cma value is somehow adjustable in recent sunxi3.4 now, but i don't think so
<mosterta>
right, but the 190MB are not pre-allocated, but only when the memory is really allocated piece by piece via CMA..
<mosterta>
sorry, I am wrong from that code snippet.. I have to look to my code..
nove has quit [Quit: nove]
<rellla>
If i'm not wrong, the whole block is taken from cma on module load an released on unload.
<rellla>
The value itself was discussed here on irc with ssvb some longer time ago.
<mosterta>
you are right.. good point I should change that in my kernel.. the 80MB I still have allocated are useless.. only when I want to switch back in vdpau-sunxi to local memory management, which I don't want to do..