<ssvb>
I think that the instructions are pretty much detailed
specing has joined #linux-sunxi
<ssvb>
KotCzarny: about this aliexpress thing, I think that this is an ordinary SD card socket, which is intended to be used in the SPI mode
<ssvb>
KotCzarny: it has nothing to do with SPI NOR flash
<KotCzarny>
ssvb, yes, but i wonder if it would boot via spi too?
<ssvb>
no, it would not
<KotCzarny>
shucks :/
jernej has quit [Ping timeout: 268 seconds]
<KotCzarny>
what's the difference between those two then? (sdcard on spi vs nor on spi)
<ssvb>
SPI is just a bus
<ssvb>
and it is used to transfer data, that's all about it
foodev has joined #linux-sunxi
<ssvb>
you can hook an SD card via SPI
<KotCzarny>
yes, that's why im wondering why sdcard-via-spi wouldnt work
<ssvb>
or you can hook a NOR flash via SPI
<ssvb>
but they are using entirely different higher level communication protocols and are not compatible with each other
<KotCzarny>
uhum
<ssvb>
it's the same as with TCP, you can have a web server (talking high level HTTP protocol over it), or you can have something like telnet (talking an entirely different high level protocol)
freemangordon has quit [Read error: Connection reset by peer]
freemangordon1 has joined #linux-sunxi
<ssvb>
KotCzarny: oh, wait, this partiucular SPI NOR flash chip is too small (only 512 Kbytes)
<ssvb>
still you can find something with 2.54 pin headers already soldered, but a few extra bucks
<ssvb>
apritzel: about "when I tried with the Pine64 and some SPI flash some month ago there were problems", what kind of problems were these?
<ssvb>
for example, in U-Boot I had to add an explicit delay between SPI transfers here - http://git.denx.de/?p=u-boot.git;a=blob;f=drivers/mtd/spi/sunxi_spi_spl.c;h=67c7edd92d0c6b1d7b20d072ab50085ce9bcc5b8;hb=29e0cfb4f77f7aa369136302cee14a91e22dca71#l221
<ssvb>
I mean, the udelay(1) for /* tSHSL time is up to 100 ns in various SPI flash datasheets */
<ssvb>
I think that the same fix is also needed for the sunxi-fel flasher, but I could not reproduce the same problem there (maybe because of the lower CPU clock speed, running without caches, etc.)
<calimero_82>
guys the development link of openelec doesn't work
freemangordon1 has quit [Quit: Leaving.]
freemangordon has joined #linux-sunxi
calimero_82 has quit [Quit: Leaving]
mzki has quit [Ping timeout: 268 seconds]
mzki has joined #linux-sunxi
<apritzel>
ssvb: wens: is there any reason we limit the SPL to 24KB for the sun9i-a80 in U-Boot?
<apritzel>
in the manual it looks like SRAM A1 is 40K, really
<apritzel>
or is this a BROM limitation? or is the SRAM flaky behind the first 24K?
<dgp>
Is it limited to 24KB of .text?
<dgp>
If so maybe saving space for stack etc?
<apritzel>
but the other SoC have 32KB in there
<apritzel>
and the stack is handled separately
<apritzel>
dgp: I wonder if this is just being cautious in the beginning and was never revisited
<dgp>
yeah could be
<ssvb>
apritzel: 24KB is the A10/A20 limitation
<ssvb>
apritzel: and whatever is used for A80 is up to the U-Boot A80 code maintainers :-)
<ssvb>
apritzel: what's the point inventing this CONFIG_SUNXI_HIGH_SRAM option when we can have just the SRAM base address and size defined for each SoC?
<apritzel>
because we have only two alternatives
<ssvb>
no, we have at least 3 alternatives
<ssvb>
24KiB at 0x0
<apritzel>
and I am getting really sick of this U-Boot habit of adding some stupid MACH_ symbol to sheer endless #if defined chains
<ssvb>
32Kib at 0x0
<ssvb>
and 32KiB at 0x10000
<ssvb>
and now you are suggesting that A80 might have it as 40KiB, which makes this a potential 4th configuration
<apritzel>
ssvb: no, I just want to have it 32KB
<apritzel>
because this aligns neatly with the other "high SRAM" SoCs
<ssvb>
it does not look very "neatly" to me
<apritzel>
frankly: this code is barely readable at the moment
<apritzel>
I think it's much clearer to have something like: if HIGH_SRAM: use that address; else use another; fi
libv_ has quit [Ping timeout: 244 seconds]
<apritzel>
ssvb: anyway: we can have this discussion on the list once I sent the series, I was just wondering about that A80 oddball here
<lennyraposo>
wish we could get sources for libhdmi from AW
<dgp>
Proper docs for everything would be better. Vendor code is usually complete crap
<dr1337>
Yeah
<dr1337>
if the vendor code works, it's good enough to ship.
<dr1337>
they don't even care if there's a bug in it
<wens>
dr1337: i botched up sunxi-next, missed a few branches when i updated it yesterday
<wens>
dr1337: and if you're referring to sun8i-emac not in sunxi-next, sunxi-next only contains patches that will be in torvalds' tree in the next release
<wens>
hence the name sunxi-next, instead of sunxi-devel :)
<miasma>
what's with the otg patch for h3? it's been around quite long and it's not in mainline yet
<wens>
miasma: nobody bothered to send them for inclusion
IgorPec has joined #linux-sunxi
Putti has quit [Ping timeout: 268 seconds]
<miasma>
what do you think about this page? https://linux-sunxi.org/Possible_setups_for_hacking_on_mainline -- someone asked to change the terminology from hacking to something else and also this page doesn't list usb/fel boot as an option. i was thinking of updating it to include that too
<wens>
s/hacking/developing/ ?
<miasma>
something like that
<wens>
and yes this page is really out of date
<miasma>
I'm dealing with fel & usb otg network gadgets and nbd/nfs/pxe with my boards so I thought I could update the documentation for those parts wrt uboot/kernel
<miasma>
i thought it might be kind to kind of ask for a permission from the veterans who originally created those pages :P
IgorPec2 has joined #linux-sunxi
<wens>
the original author isn't around
<wens>
and it's a wiki! by all means improve it :)
IgorPec has quit [Ping timeout: 256 seconds]
dr1337 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
Gerwin_J has joined #linux-sunxi
JohnDoe_71Rus has joined #linux-sunxi
f0xx has joined #linux-sunxi
[7] has quit [Ping timeout: 245 seconds]
TheSeven has joined #linux-sunxi
jernej has joined #linux-sunxi
reinforce has joined #linux-sunxi
libv has joined #linux-sunxi
Putti has joined #linux-sunxi
Putti has quit [Ping timeout: 256 seconds]
<dgp>
Emailed the orange pi guys and asked for a datasheet and actually got it. Really surprised
mihlit has joined #linux-sunxi
<wens>
r40 is sun8iw11p1
jernej has quit [Ping timeout: 260 seconds]
chomwitt has joined #linux-sunxi
lkcl has quit [Ping timeout: 265 seconds]
lkcl has joined #linux-sunxi
<freemangordon>
MoeIcenowy: I got my a33 booting to lxde with wifi working. the only problem I have is with TS. I followed firmware extraction instructions, got 10 .fw files, six of those are ~40k in size. I tried all six, none works (modprobe fails with either -6 or -5). TS chip ID is 0xB4820000. Any idea what I am doing wrong?
ErwinH has joined #linux-sunxi
<freemangordon>
MoeIcenowy: BTW there are 2 driver modules in android, I extracted firmware from only gslX680, but not from gslX680new as according to lsmod it is gslX680 that is loaded
<freemangordon>
trying to load gslX680new leads to "gslx680: no matched TP firmware(gsl1688_a70)!" ,so I guess GSL1688_A70_FW.fw is my FW file
<mripard>
MoeIcenowy: What was RFC about your USB OTG patches ?
lkcl has quit [Read error: Connection reset by peer]
lkcl has joined #linux-sunxi
_hp197 has joined #linux-sunxi
_hp197 has quit [Client Quit]
<muvlon>
dgp: you allowed to share that?
<dgp>
muvlon: not sure, it's nothing interesting really, electrical datasheet for the wifi chip on the pi zero
* dgp
is hoping they get the programming manual..
<MoeIcenowy>
mripard: I think Hans do not like this solution
<MoeIcenowy>
fremangordon: you will need script.bin now ...
<MoeIcenowy>
wens: I remember sun8iw10+ uses linux-3.10
f0xx has quit [Ping timeout: 258 seconds]
<mripard>
MoeIcenowy: as far as I know, hans didn't like the additional property you wanted to introduce
<mripard>
(usually, when you send an RFC, you also say what you want comment on ;) )
<MoeIcenowy>
but this time I didn't introduce any property :-)
florianH has joined #linux-sunxi
<MoeIcenowy>
and, mripard: do you think supports only musb mode now is proper?
<MoeIcenowy>
both host and peripheral mode is tested, and I have used the host mode of musb to attach a mt7601u wireless card for a long time
<freemangordon>
hmm, ok, got TS working, with a firmware found over the inet
<MoeIcenowy>
freemangordon: I suggest you use evtest to check your full TS
<freemangordon>
MoeIcenowy: well, it looks like working in LXDE
<MoeIcenowy>
or should I say "the whole TS"
muvlon has quit [Ping timeout: 260 seconds]
apritzel has joined #linux-sunxi
<freemangordon>
I got the idea, but what to test exactly?
<MoeIcenowy>
try to move your finger everywhere on the scren
<MoeIcenowy>
arete74: I even lost my kernel .config :-(
<KotCzarny>
MoeIcenowy: extract from prebuilt kernel?
<MoeIcenowy>
the kernel image is also lost :-(
<MoeIcenowy>
I have only a system.img now :-(
silviop has joined #linux-sunxi
<calimero_82>
i've resolved
ganbold has quit [Remote host closed the connection]
calimero_82 has quit [Ping timeout: 260 seconds]
ganbold has joined #linux-sunxi
calimero_82 has joined #linux-sunxi
<dizz74>
Somebody can give link with IMG file android 6.0, for study? I received it(6.0 img) from the developer, but file is strange. In sys_partition.fex file, section with "data" partition is commented by ";". But in the analog IMG file wth 4.4 - "data" section is ok. Why data is commented?
leviathanch has joined #linux-sunxi
<calimero_82>
guys heatsink is sufficient for the temperature?
<KotCzarny>
proper voltage/dvfs settings are sufficient for temperature
<calimero_82>
KotCzarny: the temperature is 67 70, isn't too hot?
<KotCzarny>
idle or full load?
<wens>
MoeIcenowy: getting android to work make for a good reason to get allwinner to ditch their vendor kernel? :p
<calimero_82>
at the start of openelec, i see info system KotCzarny
<calimero_82>
indeed on rasp2 is 44
<KotCzarny>
calimero_82: also, which soc/board
<calimero_82>
orangpione
<KotCzarny>
might be a bit too high in idle then
<calimero_82>
i've put on it 3 heatsink
<KotCzarny>
check armbian, if it also is hot after start, then something is wrong, if not, openelec needs to be fixed
<calimero_82>
KotCzarny: i use the development version on the forum
<calimero_82>
armbian is 56
<dizz74>
What about Allwinner T3? It is analog of what previous SoC? Somebody have information?
<KotCzarny>
then post on openelec forum your findings?
<tkaiser>
KotCzarny: There will be people dumb enough to buy this for sure. But that doesn't matter that much :)
<KotCzarny>
can't wait to read reviews ;)
<jelle>
what happened to that other kickstarter arm laptop
<tkaiser>
KotCzarny: Same kind of people that now buy OPi Zero with 256 MB to save $2 and to complain that there's not that much free memory available. Those wo don't get the meaning of 'you get what you pay for'
<wens>
jelle: ask lkcl
<tkaiser>
jelle: In case you mean the Azpen thingie also using A64 then shipping delayed
<KotCzarny>
actually it makes quite nice machine (that pinebook), all parts included (lcd, battery, kb)
<jelle>
oh yes azpen
<Wizzup>
freemangordon: cool! are you using the devuan image with a self-built kernel/dts and u-boo?
<Wizzup>
s/u-boo/u-boot/
<jelle>
tkaiser: A64 allwinner?
<apritzel>
KotCzarny: no Ethernet and no serial, and probably depending on blobs for basic operation (touchscreen, LCD)
<tkaiser>
KotCzarny: Glossy 1280x720 pixel, wrong colors, viewing angle 5¡ -- sounds like fun ;)
<KotCzarny>
ahm, missed that (no eth) :/
<KotCzarny>
lol
<KotCzarny>
seriously?
<KotCzarny>
where did you get the screen part number?>
<apritzel>
tkaiser: but the Pinebook mentions IPS
<tkaiser>
jelle: Allwinner A64, sure.
<muvlon>
just people putting out non-chrome os arm laptops is cool
<muvlon>
even if this particular one might not be very useful
<tkaiser>
apritzel: Maybe it's better than expected. Use only HiDPI displays any more so can't judge anyway.
<jelle>
I'm waiting on the olinuxo laptop, they probably build something better ;-)
<apritzel>
KotCzarny: you can just FEL boot it, put in the magic SD card and connect an A-A cable to one of the USB ports
<apritzel>
jelle: waiting is always good ;-)
<muvlon>
jelle: any indication they're making progress on that?
<KotCzarny>
apritzel: but it will eat usb port :/
<muvlon>
tbh I don't have my hopes up
<jelle>
muvlon: I don't know
<jelle>
the concept sounds cool though
<apritzel>
a notebook is even more "you get what you pay for"
<muvlon>
If we're waiting for vaporware, I want an EOMA-68 netbook
<apritzel>
given that a notebook user experience is dominated by things like screen, keyboard quality, hinges, battery ...
Putti has joined #linux-sunxi
<apritzel>
which you just can't do very cheaply without sacrificing quality
<tkaiser>
apritzel: And display stops working after 12 months since display cable is worn through
<apritzel>
tkaiser: so that gives you a year to setup the things as a headless server ;-)
whaf has quit [Ping timeout: 246 seconds]
<tkaiser>
apritzel: Sure, the enclosure can be opened by bending the plastic thingie over the knee to get to the PCB mounting the necessary heatsink ;)
<apritzel>
lol
<KotCzarny>
:>
<ErwinH>
Ok, megous kernel is a bit wierd, it will only boot if cpufreq_dt is blacklisted, but when I do a mod probe cpufreq_dt it works fine. Thumb test isn't fun...
<apritzel>
tkaiser: so to open the case you just drop it to the floor?
<KotCzarny>
they should mount pcb to the back of lcd, quite big heatsink ;)
<apritzel>
KotCzarny: don't give them funny ideas ...
<tkaiser>
ErwinH: sounds like a timing problem then :)
<ErwinH>
Jup. sun8i_ths should be loaded before cpufreq_dt
<ErwinH>
Next step, test if throttling works.
<KotCzarny>
apritzel, what about 21" screens? even bigger heatsink!
<wens>
ErwinH: sounds like a bug if it doesn't do deferred probing?
<silviop>
Direct firmware load for silead/gsl1680f.fw failed with error -2
<silviop>
but i put it in /lib/firmware and in /lib/firmware/silead
<KotCzarny>
silvioip, read backlog, around 3h ago freemangordon solved that
<silviop>
i dont need firmware files i nedd to understand why it not try to load my file
<silviop>
i have gsl1680.fw file extracted from my tablet
<silviop>
gsl1680f.fw
<tkaiser>
apritzel: Both 11" and 14" Pinebook share the same physical dimensions. Curious how the 11" variant then will look like. On their product page only pictures of the 14" can be seen.
<KotCzarny>
tkaiser: eee pc 7 and 9 ;)
<apritzel>
tkaiser: really? that would be a shame ...
<apritzel>
tkaiser: the whole point of a smaller screen would be smaller dimensions
<ErwinH>
Or bigger battery's...
<tkaiser>
apritzel: The average Pinebook customer prefers 11" over 14" since $10 less ;)
<apritzel>
tkaiser: true ;-)
<apritzel>
three inches for 10 $ doesn't sound right, that's overpriced !!1!!
<tkaiser>
Since it has that Windows key I bet they have to open an own forum over at pine64.org: 'No, Windows won't run on Pinebook'
<tkaiser>
Subforum: 'No, Windows IoT is something different'
<KotCzarny>
hey, hey, maybe they've dropped eth port because of the whole 'gigabit doesnt work 1!!1!1' ?
<KotCzarny>
but seriously, they've also missed opportunity to sell eth dongle that would simply connect to exposed header
<silviop>
firmware-name = "/lib/firmware/gsl1680f.fw"; drop path so i'dont know where firmware loader search for firmware
whaf has joined #linux-sunxi
<terra854>
Is there such thing as a "Pinebook"?
<terra854>
Cause google gave no results on it
<tkaiser>
terra854: Then it doesn't exist
<tkaiser>
terra854: Didn't read the announcement over in pine64.org forum?
<Wizzup>
MoeIcenowy: I (also) have a a33 tablet, seems to be similar to the inet on the wiki, although the fel boo mode is like the Q8. Ehm, any suggestions on how to go forward, and want me to back up some firmware?
<Wizzup>
I opened the thing up & will make pictures tonight
<tkaiser>
It's 1344 MHz the clueless Android users thinks since that's reported by cpufreq-info while budget cooling settings limit everything to 1152 MHz by default
<KotCzarny>
tkaiser, but thats on openair/heatsink, i wonder what would happen inside enclosure as tight as those two
<tkaiser>
So everyone is happy. Those having the slow laptop thinking it runs at high clockspeeds and those who sell this stuff
<tkaiser>
KotCzarny: The usual Android benchmark runs short enough so that's no problem either
<zoobab>
orangepi zero h2 received this morning!
<tkaiser>
KotCzarny: And by looking at those Geekbench results you clearly see that Allwinner and Pine Inc still ship with idiot settings:
<tkaiser>
When single and multi threaded score is identical then 3 CPU cores have already been killed by overheating.
<tkaiser>
When multithreaded is only ~2 as much then it's 2 CPU cores being killed. And they never come back :)
<tkaiser>
But no one cares since those shitty Android benchmarks also list '4 cores' and both users and Pine Inc don't give a sh*t
<Wizzup>
Is the pine64 considered 'community hw' ?
leviathanch has quit [Read error: Connection reset by peer]
fl_0 has quit [Ping timeout: 250 seconds]
<zoobab>
@Wizzup you want to know what I think about OSHW?
<Wizzup>
I am not sure.
<zoobab>
it should be renamed open schematics
<zoobab>
because the big black chip in the middle is barely documented
<Wizzup>
I still love olimex for doing what they are doing.
<silviop>
[ 3.604965] silead_ts: probe of 0-0040 failed with error -2
<silviop>
l@l:~$ ls /lib/firmware/silead/gsl1680f.fw
<silviop>
/lib/firmware/silead/gsl1680f.fw
<silviop>
A bug in firmware loader ?
<Wizzup>
silviop: if you read the backlog, someone fied your problem, look in more detail
fl_0 has joined #linux-sunxi
<silviop>
Semmes not the same problem
<silviop>
i have e file but loader not fount (-2) in filesystem
<silviop>
as you can see
<terra854>
tkaiser: Strange, there is nothing on the forums about the Pinebook
<Wizzup>
I don't think so, but I don' t have time to look in detail atm
gumblex has quit [Quit: Bye.]
<tkaiser>
terra854: Why strange? Do you have the impression Pine Inc is interested in the forum or the users there? They granted everyone writing more than 5 post moderator status and went away already ;)
<terra854>
Tbh, i have no idea about the current state of Pine Inc
<KotCzarny>
well, they are busy spewing more 'products'
<KotCzarny>
apparently
<KotCzarny>
pine64 was a success financially wise
apritzel has left #linux-sunxi [#linux-sunxi]
<terra854>
Well what a bummer, i was hoping that the eMMC is detachable (like the Odroid XU) so that I can replace it with one with a bigger storage capacity.
<terra854>
But, looks like it's not going to happen :(
<silviop>
I compiled silead in kernel and not as module , kernel_read_file_from_path seems fail , but is root mounted at this stage ?
reinforce has quit [Quit: Leaving.]
reinforce has joined #linux-sunxi
<silviop>
No root is not mounted, so there is a bug in kernel configuration that permit to compile in kernel drivers that needs firmware file when is not available.
<silviop>
ok is not a bug , it's a feature , you can add to kernel firmware blobs!!
lemonzest has quit [Quit: Leaving]
Pepe has joined #linux-sunxi
fkluknav has quit [Ping timeout: 240 seconds]
<silviop>
It works!! Nice CONFIG_EXTRA_FIRMWARE is the key
<ErwinH>
Throttling with the magus kernel is working nicely. Set the trip points to 35, 45, 55 and 75 and once it reaches 55 degrees it stays around that temperature.
<scelestic>
ErwinH: what board?
<ErwinH>
OrangePi One
<scelestic>
oh right, i wasnt able to get any readings on my orangepi plus 2
fkluknav has joined #linux-sunxi
nik has joined #linux-sunxi
<nik>
hey everyone, can i use the pi pc without a heatsink and the 4.9 armbian beta image?
tkaiser has joined #linux-sunxi
<ErwinH>
Modprobe cpufreq-dt is really hit or miss...
<KotCzarny>
add a usleep?
<KotCzarny>
(i know, bad idea)
<ErwinH>
I do a modprobe cpufreq-dt from the command line, after the boot is finished.
<ErwinH>
Sometimes it works, sometimes it won't.
<lkcl>
wens: jelle means the 11in A64-based $99 (or so) laptop that was on kickstarter, not the 15.6in EOMA68 laptop housing
<nik>
tkaiser: do i need a headsink on the pi pc with 4.9 beta armbian
JohnDoe_71Rus has quit [Quit: Miranda IM! Smaller, Faster, Easier. http://miranda-im.org]
<tkaiser>
nik: Unless you run heavy stuff I doubt it
<tkaiser>
miasma: BTW: Olimex uses 5.5/2.1 barrel with all their boards (since they target industrial customers not that much of a problem) and for those FriendlyARM H3 devices that should be powered through Micro USB maybe mention this: http://www.friendlyarm.com/index.php?route=product/product&product_id=89
<hramrach>
tkaiser: thanks for the link. While there is irclog it may or may not have said laptop mantioned in more detail and it might have been mentioned yesterday, last week, last moth, last year for all I know
<tkaiser>
miasma: Same 4/1.7 barrel as the other H3/H5 devices and cable included to connect to all their boards (4-pin header)
<hramrach>
btw are there some magnetic USB split cables good enough to use for devboards?
<KotCzarny>
if you intend to draw more than 1.8A, nope
<hramrach>
From what I saw there is a shitload of them and most are useless because they support only mini usb or only micro
<hramrach>
and yes, their electrical characteristics are usually not very well specified
<KotCzarny>
nah, microusb leads are tiny by the standard
<hramrach>
that may be why many boards use mini usb
<tkaiser>
hramrach: Which boards? I only know Olimex.
<KotCzarny>
tkaiser, doesnt matter if the jack is unchanged
<tkaiser>
KotCzarny: I know but problem N¡ 1 is undervoltage and not current limitation.
<KotCzarny>
if you connect anything to usb on device then current might be a problem too
DullTube has quit [Quit: Leaving]
<tkaiser>
KotCzarny: Still Ohm's law, as soon as the board needs high amounts of current with the average crappy 26AWG/28AWG cable still undervoltage is the problem. Since voltage then drops below 4.5V immediately. Avoiding crappy Micro USB jack if possible is the better way though.
mihlit has quit [Quit: Konversation terminated!]
<KotCzarny>
yes
<scelestic>
so i guess it's best to just get the barrel connector and make your own cable ?
cnxsoft has quit [Quit: cnxsoft]
<hramrach>
tkaiser: many tablets are micro usb and use sunxi chips. Most devboards that use usb for power use mini usb. With a magnetic splitcable solution that works with either you can just leave the tips in and use one cable
<hramrach>
unfortunately, I have not found any
<MoeIcenowy>
Wizzup: you just keep your NAND flash contents
<MoeIcenowy>
mainline currently can only be installed on SD
<hramrach>
nand still not solved? Looked pretty close last time I tried it
mzki has quit [Ping timeout: 245 seconds]
<hramrach>
hm, remix os
<MoeIcenowy>
hramrach: I tried to add NAND support for sunxi-spl
<MoeIcenowy>
but failed
mzki has joined #linux-sunxi
<hramrach>
I wonder when we get nightcore os
<MoeIcenowy>
sun4i, sun5i and sun7i seems to be ok
<MoeIcenowy>
but sun6/8i have a different DMA engine
<Wizzup>
MoeIcenowy: there's no DT for it, but I think most of the tablet dt is shared anyway, it seems
<Wizzup>
at least that is what I gathered by looking at the Q8 dts on kernel.org
<MoeIcenowy>
Wizzup: what board?
<MoeIcenowy>
not everything is shared, for example, touchscreen
rotcpy has quit [Quit: leaving]
<MoeIcenowy>
and the LCD settings must go to u-boot'
leviathanch has quit [Remote host closed the connection]
<MoeIcenowy>
Wizzup: I think this device can be considered as kind of q8_1024x600
leviathanch has joined #linux-sunxi
fkluknav has quit [Ping timeout: 250 seconds]
f0xx has quit [Ping timeout: 260 seconds]
rotcpy has joined #linux-sunxi
<MoeIcenowy>
Oh... Uniscom is making many A33 tablets with 1280x800 screen...
<tkaiser>
MoeIcenowy: With your A64 branch can I enable ehci/ohci on both USB nodes at the same time?
libv has joined #linux-sunxi
rotcpy has quit [Quit: leaving]
fkluknav has joined #linux-sunxi
libv_ has quit [Ping timeout: 265 seconds]
<MoeIcenowy>
tkaiser: currently, I use EHCI/OHCI for USB1, and MUSB for USB0 in host mode
chomwitt1 has joined #linux-sunxi
chomwitt has quit [Ping timeout: 252 seconds]
<tkaiser>
MoeIcenowy: Hmm... then I've to check again what's wrong with my setup. Anyway, blasting fast UAS performance on the lower USB port with Pine64.
reinforce has quit [Quit: Leaving.]
<MoeIcenowy>
tkaiser: check sunxi musb on the upper port
<MoeIcenowy>
(CONFIG_USB_MUSB_SUNXI
<tkaiser>
MoeIcenowy: # CONFIG_USB_MUSB_HDRC is not set ;)
<tkaiser>
MoeIcenowy: And no CONFIG_USB_MUSB_SUNXI too. Ok, next try...
<MoeIcenowy>
CONFIG_USB_MUSB_SUNXI depends on many options
<MoeIcenowy>
please ensure it and its dependencies (including CONFIG_USB_MUSB_HDRC, of course) are all met
deskwizard has joined #linux-sunxi
f0xx has joined #linux-sunxi
IgorPec2 has quit [Ping timeout: 256 seconds]
<Wizzup>
MoeIcenowy: cool
<Wizzup>
MoeIcenowy: the internals are almost exactly like the inet though (pics will follow soon)
<MoeIcenowy>
Wizzup: you should backup the vendor's gslX680 and gslX680new drivers.
<MoeIcenowy>
GPU is also supported in this tree, however, you need to manual enable display pipelines in device tree
<MoeIcenowy>
I have never a tablet with 1024x600 screen to test the pipeline
<MoeIcenowy>
my tablet is 800x480
<freemangordon>
mine is 1024x600, but I don;t really want to waste time now on that. 3d support will be needed as soon as we have hildon-desktop and all the satellites build, which will be soon(tm) hopefully
<MoeIcenowy>
you have still not hildon-desktop usable?
<freemangordon>
no
<freemangordon>
as I told you, I have one more(so far) maemo5 daemon to RE
<KotCzarny>
3d support is.. well, tricky
<freemangordon>
then we will start to gather pieces together, then, hopefully we'll have h-d with no application
<freemangordon>
*applications
libv_ has joined #linux-sunxi
<freemangordon>
KotCzarny: I guess h-d will work with SW rendering on 4 cores x 1,3GHz :)
<KotCzarny>
:)
<freemangordon>
with NEON and hrdfp
<KotCzarny>
at least you will be the reason to make 3d usable for anything
<freemangordon>
yet another typo day :(
<freemangordon>
lets first have clockd finished :)
<KotCzarny>
for now it's not used much so not much interest in mainlining
<freemangordon>
I see
<KotCzarny>
s/not much/at all/
libv has quit [Ping timeout: 260 seconds]
<KotCzarny>
hmm
<MoeIcenowy>
I have tried to run a gnome-shell on A33 with Mali :-)
<MoeIcenowy>
although with LLVMpipe a gnome-shell can also run :-)
<Wizzup>
llvmpipe should be fine for testing purposes
libv has joined #linux-sunxi
libv_ has quit [Ping timeout: 248 seconds]
<MoeIcenowy>
does all distros ship LLVMpipe on ARM now?
<MoeIcenowy>
I remembered in order to use LLVMpipe on non-x86 a special build option is needed for Mesa
<Wizzup>
I don't know, I think llvmpipe is usually always available
<Wizzup>
It's the default software renderer AFAIK
<Wizzup>
I have llvmpipe on arm anyway (gentoo) and didn't explicitly enable it
yann has quit [Ping timeout: 246 seconds]
Nacho__ has quit [Remote host closed the connection]
<MoeIcenowy>
"--enable-gallium-llvm" is needed for non-x86 architectures when building mesa
<Wizzup>
Just wondering, why do you want to find out?
<MoeIcenowy>
I think some distros may forgot to enable the option on non-x86
<Wizzup>
I guess users will find out, there's still the swrast, right?
<Wizzup>
I would think that most distros providing any kind of arm gui will enable llvmpipe
<MoeIcenowy>
yes
<MoeIcenowy>
but without it, the swrast will be softpipe
<MoeIcenowy>
softpipe is a disaster
Nacho_ has joined #linux-sunxi
<Wizzup>
right, users will find out :P
<Wizzup>
bbiab
deskwizard has quit [Ping timeout: 258 seconds]
gzamboni has joined #linux-sunxi
libv_ has joined #linux-sunxi
vagrantc has joined #linux-sunxi
libv has quit [Ping timeout: 245 seconds]
cptG_ has joined #linux-sunxi
cptG has quit [Ping timeout: 260 seconds]
vagrantc has quit [Quit: leaving]
apritzel1 has joined #linux-sunxi
netlynx has joined #linux-sunxi
lennyraposo has quit [Ping timeout: 260 seconds]
Ultrasauce has quit [Remote host closed the connection]
<tkaiser>
KotCzarny: In other words: Forget about A20 from now on ;)
<Wizzup>
unless you want SATA or fast IO
<KotCzarny>
yup. sata still wins
<tkaiser>
Wizzup: Wrong
<KotCzarny>
200/50 > 42/41
<tkaiser>
KotCzarny: Wrong
<tkaiser>
Care to read the link prior to commenting? ;)
<Wizzup>
I get speeds over SATA that you can never get over USB
<Wizzup>
s/USB/USB2/
<tkaiser>
Ok, platitudes mode ;)
<Wizzup>
170MB/s
<Wizzup>
try that on USB2
<tkaiser>
Wizzup: Yeah, platitudes mode reloaded
<Wizzup>
what...
<Wizzup>
care to explain yourself?
<TheLinuxBug>
tkaiser: you get an R40 board yet?
<Wizzup>
scrub status for a6c6dd71-c6ca-4f37-9be2-3b7a4ab6c816
<Wizzup>
total bytes scrubbed: 58.31GiB with 0 errors
<Wizzup>
scrub started at Wed Nov 16 12:40:43 2016 and finished after 341 seconds
<tkaiser>
Wizzup: No, why should I. Care to read through the link?
<KotCzarny>
in one regard iops are much higher
<Wizzup>
tkaiser: yes, I saw the link
<KotCzarny>
so for compiling etc uasp > sata apparently
<tkaiser>
TheLinuxBug: No, there is none around (only a Banana and I don't like Bananas that much)
<KotCzarny>
btw. would be nice to compare cpu usage in all situations
<tkaiser>
Wizzup: If your use case is 'reading data sequentially from disk' then A20 is for you. Otherwise you should start to check reality
<Wizzup>
with an SSD, does that matter?
<Wizzup>
I have a LIME2 (A20) with several disks on USB, and four disks on a SATA PMP, and I can very clearly tell the difference
<Wizzup>
none of those are SSDs (apart from the rootfs)
afaerber has joined #linux-sunxi
<jelle>
\o/ finally have my dtb compiled, Jean-Francios missed somehting =)
<tkaiser>
Wizzup: I will stop here since useless. One last sentence: Random IO is more important in most use cases. And here A20 simply sucks. I'm also talking not about USB 2.0 with A20 but with A64. And all of this can be read in the link(s) above so why should I repeat myself?
<Wizzup>
tkaiser: I have read the link before I commented
<TheLinuxBug>
tkaiser: Does banana not send out dev board for testing? I mean I am not a huge fan of all their stuff, but would be interesting to test out
<TheLinuxBug>
I just haven't been able to make an argument to spend $50 on it after ordering the EspressoBins ;z
<KotCzarny>
tkaiser, you might still do one real life test, same kernel tree, same config, time make -j2 or something on both
<KotCzarny>
and that will be REAL test
<Wizzup>
tkaiser: It seems pretty clear to me that SATA outperforms USB2, you may argue that the A64 may help here, but otherwise I don't understand it. Furthermore, what I also found very nice with SATA is that it seems ... quite a bit more stable. I can for example even unplug a disk and plug it back in, and it'll still work
<Wizzup>
with USB, I'll lose my rootfs
<Wizzup>
but I must admit that maybe this is not the case with UAS
<Wizzup>
furthermore, let me know when I can buy HDDs/SSDs with USB port instead of SATA port, it's also added complexity, and thus more things that can fail
<Wizzup>
I will agree with you that the speeds achieved in the post are quite decent, but I don't think that really paints a bad story for SATA on the A20 per se
pg12 has joined #linux-sunxi
<TheLinuxBug>
Wizzup: they actually now sell 'external' 2.5" drives with only USB 3.1 interface (as in its not an addon board, it is built into the drive) just in case you didn't know that :Z
Ntemis has joined #linux-sunxi
<Wizzup>
I didn't know that :)
ErwinH has quit [Ping timeout: 260 seconds]
<KotCzarny>
i dislike usb in general because its too susceptible to noise
<TheLinuxBug>
I remember a lot of people bithcing cause they were wanting to buy the 2TB 2.5" externals and remove enclosure to use in notebook or something
<TheLinuxBug>
but instead they open enclosure and find its ONLY USB
<KotCzarny>
TheLinuxBug: those drives often have testpoinst that expose sata interface ;)
<TheLinuxBug>
they exist but I imagine are not too popular
<TheLinuxBug>
ahh, well I never wanted to bother with that, but ill take your word for it :Z
alexvf has quit [Quit: Page closed]
<KotCzarny>
i was trying to figure out why one drive was failing and connected it via those points, but it was internal problem anyway so didnt help much
<TheLinuxBug>
So I was looking at that H5 board OPi PC2, but I don't really understand the benefits of it over some other boards, as I understand it I guess the cores can get up to 1.7Ghz with proper cooling but won't always run that way so they don't even advertise CPU speed in Ghz for them, also it looks like it doesn't have 4k and such... what is the actual thing that make it attractive to people?
<KotCzarny>
TheLinuxBug: a57 cores instead of a7, otherwise almost the same as h3
<TheLinuxBug>
okay but from usage standpoint what is there to gain from it?
<TheLinuxBug>
just A64?
pg12 has joined #linux-sunxi
<KotCzarny>
2.5 higher performance
<KotCzarny>
when optimized
<KotCzarny>
(for neon)
<TheLinuxBug>
hmm
<TheLinuxBug>
so more so a better android board maybe?
<KotCzarny>
but for regular software ist roughly the same
<TheLinuxBug>
I know android can take advantage of Neon for gfx and such
<TheLinuxBug>
but what other use cases will you be better off with that?
<TheLinuxBug>
I mean its 20$ and its almost intersting enough to add to my collection but I could never get my self to pull thr trigger
<TheLinuxBug>
just seems like I have a handful of other boards that can do about the same
<TheLinuxBug>
is there any good dev reason to have one? I mean I looked at even the H8 back in the day and it didn't even really hold enough of my interest and as I understand this is supposed to be similar?
avph has joined #linux-sunxi
<KotCzarny>
there is 2GB version in the works afair
<KotCzarny>
its almost the same as H3, so for your uses opipc/opi+2e would do the same
<TheLinuxBug>
From what I see still lacks USB 3.0, only 1Gb ram currently 8MB NOR flash I guess would be interesting though cause I am guessing uBoot can be stored there?
libv has joined #linux-sunxi
<TheLinuxBug>
has a composite video interface instead of only HDMI it seems
<jelle>
hmm seems my orange pi pc hangs on starting kernel
<tkaiser>
TheLinuxBug: I don't want anything from them even for free. Time has a value and those morons *never* documented their own hardware correctly.
<TheLinuxBug>
So really the only changes are NOR flash, composite video output, 64bit (assuming mainline provides for it) and a slightly more powerful GPU if you use Neon?
<tkaiser>
Wizzup: Performance wise it's easy, there's one are where A20 (with SATA) outperforms the more recent Allwinner SoCs: sequential read speed. So what?
<TheLinuxBug>
tkaiser: I understand... just thought maybe as the first R40 board it would be something you would have some interest in, but I agree, $50 is too much for testing an undocumented board -- do you know if any other vendors plan to start packaging it in the near future? Olimex, Cubieboard, etc?
libv_ has quit [Ping timeout: 244 seconds]
<KotCzarny>
TheLinuxBug: it has mali450 instead of mali400 too, yes
<tkaiser>
Wizzup: Nice to know that I can exceed 200 MB/s when reading from a fast disk. If it fits the use case also ok. But for stuff like databases or a small boring backup NAS I don't see any clear advantages for A20 any longer
<KotCzarny>
but neon could be used for number crunchy apps
<TheLinuxBug>
I see
<tkaiser>
Wizzup: And that's only about A20's SATA implementation, other SBC that feature SATA look way better here (see aforementioned link containing other links ;) -- anyway, habe to prepare food for 40 women and am off now :)
<TheLinuxBug>
KotCzarny: well I know there is a lot of use cases for Neon in Android as it can take advantage of the acceleration in a lot of apps, but from regular linux perspective I would be interested to know how you would really take advantage of that?
<KotCzarny>
one example, turbojpeg lib (libjpeg replacement) uses neon
<KotCzarny>
and any software which you compile to use neon for floating point i guess
<TheLinuxBug>
I see, is the improvement really that impressive?
<KotCzarny>
neon is like mmx/sse
<TheLinuxBug>
Do thay have an Xwindows driver that can take advnatgae of it as well?
<TheLinuxBug>
I see, so more like a computational flag
<TheLinuxBug>
so my understanding is a bit skewed
<TheLinuxBug>
I thought it was more closely related to the GPU
<KotCzarny>
long time ago cpus had fp, then mmx/sse came and software was 5-10x faster ;)
<KotCzarny>
s/fp/fp unit/
<TheLinuxBug>
yeah I remember that
<Wizzup>
KotCzarny: is sse not part of the cpu?
libv_ has joined #linux-sunxi
<TheLinuxBug>
I just don't think I connected the parts in my head correctly -- I always thought Neon was part of GPU accel not CPU, but what you said I think clears that up a bit.
<KotCzarny>
wizzup: in a way, no. because you can still have cpu without sse, so its more like extension
BenG83 has joined #linux-sunxi
<montjoie>
extention for x86, but mandatory to have for x86_64
<TheLinuxBug>
so as long as software is compiled to take advantage of Neon cpu extension you get additional boost in performance
<montjoie>
i mean optionnal
<montjoie>
for x86
libv has quit [Ping timeout: 246 seconds]
<KotCzarny>
ssvb, in his hacky wisdom, wrote fbturbo which utilizes neon for data shuffling
<KotCzarny>
that's why you think of it as a gpu component
<TheLinuxBug>
Well also you will find that a lot of video codecs in Android take advantage of Neon as well
<TheLinuxBug>
thats also where some of my thought process came from
<TheLinuxBug>
So whats the chances of a 64bit release for H5 actually coming to fruition? is it something even seriously being worked on?
<TheLinuxBug>
or is the goal to wait on mainline
<jelle>
TheLinuxBug: patches have been posted on linux-sunxi
<jelle>
but there is no u-boot support yet
<jelle>
but patches posted there too
<KotCzarny>
TheLinuxBug: forget about h5 for now
<TheLinuxBug>
intersting
<TheLinuxBug>
is the 8MB NOR actually being taken advnatage of?
<TheLinuxBug>
were they intending for uBoot to reside there?
<KotCzarny>
nor is going to be used to store uboot
<TheLinuxBug>
gotcha
<TheLinuxBug>
that will be nice
<jelle>
8MB NOR what's that?
<KotCzarny>
also, they actually put 16Mbit nor chips (2MB)
<jernej>
MoeIcenowy: Execute "hdmi_writel(0x10014, 0x42494E47);" and all registers will be like in the iMX6
<MoeIcenowy>
this allwinner is mysterious
<jernej>
I found that in R40 binary blob, but it works on H3 also
<MoeIcenowy>
or... maybe DWC just provides a obfuscation option
<MoeIcenowy>
iMX6 didn't enable it
<MoeIcenowy>
but AW enabled it
<MoeIcenowy>
:-)
<KotCzarny>
so, do we have working hdmi for mainline then?
<jernej>
I guess it works with all similar HDMI controllers
<jernej>
KotCzarny: Only partially, because mainline DWC driver is tightly coupled with DWC phy, which is not used here
<jernej>
only on A83T
<jernej>
AFAIK
<MoeIcenowy>
so first we need to decouple DWC driver with DWC phy?
<MoeIcenowy>
make it capable of using another phy?
<jernej>
I'm not really sure what is correct procedure
<jernej>
my way would be to add some kind of phy hooks
<jernej>
if they are defined, use them, otherwise use already present code
<MoeIcenowy>
who will modify dw-hdmi.c?
<MoeIcenowy>
you or moinejf?
<jernej>
but this may be too hackish
<MoeIcenowy>
This allwinner is always making interesting things :-)
<jernej>
I don't plan to and I'm not sure if moinejf considered this
<MoeIcenowy>
maybe we should check any new IP core which appears on sunxi SoCs with DWC ones :-)
<jernej>
jemk mentioned something about looking into reusing dwc driver
<MoeIcenowy>
if the controller is proven to be DW-HDMI
<MoeIcenowy>
mainline won't accept new driver
<MoeIcenowy>
just like the situation in sun8i-emac
<MoeIcenowy>
as it's discovered as a obfuscated dwmac
<ElBarto>
out of curiosity why new driver won't be accepted in mainline if it's DW-like ?
<jernej>
because kernel has policy that only one driver can exist for same device
<KotCzarny>
#ifdef to the hell and back then?
<KotCzarny>
;)
<MoeIcenowy>
or alike devices
<ElBarto>
what about i2c then ? it's marvell IP
<MoeIcenowy>
ElBarto: it's a clean marvell IP
<MoeIcenowy>
and just works with the marvell driver
<ElBarto>
register aren't at the same offset
<jernej>
yes, but logic is the same
<MoeIcenowy>
it's dealed in drivers/i2c/busses/i2c-mv64xxx.c
<jernej>
for that you don't need whole new driver
<ElBarto>
yeah sure
<ElBarto>
I understand why a whole new driver in unacceptable
<MoeIcenowy>
and it uses a struct named mv64xxx_i2c_regs
<MoeIcenowy>
it has two instances: mv64xxx_i2c_regs_mv64xxx and mv64xxx_i2c_regs_sun4i
<ElBarto>
yeah ok just word usage that I didn't get, "new" as "whole new"
<MoeIcenowy>
the first one is used in Marvell's SoCs, and the second one is used for Allwinner's SoCs
<MoeIcenowy>
kernel is always trying to reuse code.
<ElBarto>
yeah yeah I get it, same about FreeBSD, I've reuse the code of the marvell i2c for allwinner, as I said it was the word "new" that wasn't clear to me :)
yann-kaelig has quit [Quit: Leaving]
IgorPec has joined #linux-sunxi
<apritzel>
ElBarto: let me explain it once more to you ... ;-)
<dan0_0>
Re: Neon being 2.5x faster on the H5, I'm curious how much of a boost that will be for newer crypto like stream ciphers from DJB
<dan0_0>
I know he uses Neon heavily since performance is actually improving for it due to video decoding & video gaming, vs performance for crypto accelerators is in stasis on x86-64 or dropping by a few percent year over year
<dan0_0>
So H5 is a no buy right now?
<KotCzarny>
if you have pine64 board you can check by yourself too
<dan0_0>
if it was A15 or A57 I would be much more interested
<dan0_0>
at this point I'm only interested in the OrangePi PC 2 due to the gigabit and potentially better cjdns performance due to the better Neon performance
<dan0_0>
Neon makes Salsa20 go faaast
<apritzel>
dan0_0: ??? the H5 has the same cores as the A64
<dan0_0>
yeah
<dan0_0>
I know
<dan0_0>
hence my lack of interest
<apritzel>
so why are you interested in the OrangePi PC 2 then?
<dan0_0>
gigabit and better Neon performance
<apritzel>
???
<dan0_0>
for under $20
<KotCzarny>
pine64 has gigabit too
<apritzel>
and NEON performance should be exactly identical
<dan0_0>
yea
<apritzel>
(at the same frequency)
<dan0_0>
but aren't they shipping warped boards and stuff?
<KotCzarny>
:)
<apritzel>
how does this affect NEON performance? ;-)
<KotCzarny>
those were initial bubble envelopes shipments
dr1337 has joined #linux-sunxi
<KotCzarny>
later they switched to cardboard ;)
<apritzel>
seriously, Pine64 are packed into cardbox boxes these days
<apritzel>
so the Pine64 and the OPi PC 2 behave almost identical when it comes to NEON and Gigabit
<apritzel>
hence my question above ...
<dan0_0>
Idk, I've bought quite a few Opi PC's, trust Xunlong more. Will pay the $.98 premium for a smaller package
<apritzel>
sure, do whatever you like
<apritzel>
just wanted to state that there is no technical reason
<KotCzarny>
keep in mind H5 will take some time to boot/run with mainline
<dan0_0>
mmm
<apritzel>
I like the OPi PC 2 because it has SPI flash
<dan0_0>
That too
<apritzel>
KotCzarny: 4.11
<apritzel>
Maxime has nodded to my RFC patches already