<servili007>
ZaEarl: In the sense that China can do better, yeah. In the sense that it's a capacitive panel, screen, and basic a10 board, I've seeen worse
freakazoid0223 has joined #arm-netbook
freakazoid0223 has quit [Quit: Leaving]
<Alex1269>
mnemoc, I named driver gpio-sunxi, but it looks like driver irq part is soc specific. Is there any way for driver to get soc name at runtime ?
<Alex1269>
And I found that sys_config.c gpio_request_ex() function doesn't reserve gpios and can be called twice for same gpio. This is not good... :(
amoxoo has joined #arm-netbook
amoxoo has left #arm-netbook [#arm-netbook]
amoxoo has joined #arm-netbook
fragmint has quit [Read error: Operation timed out]
<TestModule>
servil, some locations it's still going on, but a week or two ago?
<TestModule>
Christmas clearance all Jan
fragmint has quit [Ping timeout: 264 seconds]
dyoung-away is now known as dyoung
hipboi has joined #arm-netbook
hark has quit [Ping timeout: 245 seconds]
hark has joined #arm-netbook
ZaEarl has joined #arm-netbook
dyoung is now known as dyoung-away
Jonathan_Eyre has quit [Ping timeout: 240 seconds]
Jonathan_Eyre has joined #arm-netbook
gimli has joined #arm-netbook
coder0x has quit [Remote host closed the connection]
discopig has quit [Ping timeout: 276 seconds]
eebrah has quit [Quit: ChatZilla 0.9.89 [Firefox 18.0.1/20130116073211]]
eebrah has joined #arm-netbook
Ershov has joined #arm-netbook
Quarx has joined #arm-netbook
<TestModule>
You know, the A10 isn't that slow if it's running at 1ghz the whole time. Is there a way to lower the voltage on the MK802 and / or overclock it?
<libv>
heh, spinning companion cube (with vertices and elements in buffers): binary driver: 176 fps. limare: 212 fps.
<libv>
and this with 1GHz cpu, 320Mhz gpu, and memory at the default 360Mhz
<libv>
limare is, above having less cpu overhead due to it being much more rudimentary, also apparently more aggressive at job scheduling
ZaEarl has quit [Ping timeout: 255 seconds]
<TestModule>
oh wow. Thanks rm
gimli has quit [Ping timeout: 248 seconds]
rellla has joined #arm-netbook
rellla has quit [Client Quit]
rsalveti has quit [Ping timeout: 248 seconds]
Quarx has quit []
pcat has joined #arm-netbook
rellla has joined #arm-netbook
cheng has joined #arm-netbook
gzamboni has quit [Read error: Connection reset by peer]
<amoxoo>
i was tring to open the EGL accelration using Mali armhf r3p0 bin from linux-sunxi mali page, but when i tried to run some demo like glmark2-es2 or even the test under Mali-libs, i got an symbol lookup error: /usr/lib/libEGL.so.1: undefined symbol: XextCreateExtension
<amoxoo>
how can I proceed?
<TestModule>
amoxoo , what hardware are you using?
<amoxoo>
i am running on CubieBoard
<amoxoo>
from hipboi
<amoxoo>
now running Ubuntu Precise armhf from linaro
<cheng>
is running armhf is faster than armel?
<rm>
yes
<cheng>
hmm...a quick find found this:http://www.memetic.org/raspbian-benchmarking-armel-vs-armhf/
<cheng>
armhf faster than armel by 20%
Kebianizao|work has joined #arm-netbook
<cheng>
amoxoo: can i know your BogoMIPS reading on /proc/cpuinfo?
<cheng>
i have this reading on my meleA1000
<cheng>
BogoMIPS : 59.63
<cheng>
while i have another cortex-a8, with 1GHz from TI reading this
<cheng>
apply the tweaks will increase the performance?
<cheng>
as I try to run some application, on TI is running fine, but is much slower on meleA1000
<TestModule>
cheng , yes.
<TestModule>
listen to rm , he is right
<cheng>
thanks, will give it a try
<amoxoo>
after tweaks , the BogoMIPS updated to 1001.88
<TestModule>
lol
<cheng>
no kidding...that's a BIG jump.
sspiff has joined #arm-netbook
<jelly-home>
BogoMIPS: still bogus as much as the day they were named on
<amoxoo>
and really enhanced alot on playing 1080p video, but still decoding by software...
<amoxoo>
from 2 fps => 10 fps
<cheng>
i bet changing my armel to armhf will further improve it
<cheng>
10fps is still not good enough though
<amoxoo>
my problem about EGL still exist
<amoxoo>
is GET_UMP_SECURE_ID_BUF1 return 0x01 and GET_UMP_SECURE_ID_BUF2 return 0x02 correct?
gzamboni has joined #arm-netbook
<ssvb>
amoxoo: usually yes
<ssvb>
amoxoo: have you already solved the XextCreateExtension problem?
<amoxoo>
not yet
<ssvb>
use the libUMP.so compiled from sources instead of the blob
<amoxoo>
nm -u libMali.so got a lot of undefined symbol which are start with X....
<amoxoo>
ok
<ssvb>
yes, as a workaround, these dependencies are pulled in from libUMP
eebrah has quit [Quit: ChatZilla 0.9.89 [Firefox 18.0.1/20130116073211]]
hansg has joined #arm-netbook
<amoxoo>
thanks ssvb~
<amoxoo>
succeded~
<amoxoo>
compile libUMP from source
<amoxoo>
thank you very much~
<amoxoo>
i got 70fps on glmark2-es2
<gzamboni>
hi, is this gpio documentation valid for all sunxi devices or its only for the A13? http://linux-sunxi.org/GPIO
<cheng>
70fps of HD decoding?
<amoxoo>
no~
<amoxoo>
glmark2-es2
<amoxoo>
EGL rending
amoxoo has left #arm-netbook [#arm-netbook]
<TestModule>
amoxoo , is X11 accel working for you as well?
<TestModule>
ugh
<TestModule>
r3p0 vs r3p1 = ?
<TestModule>
nm
pcat has quit [Ping timeout: 252 seconds]
<Alex1269>
gzamboni, this also works for A10
<TestModule>
Odd, for some reason my 2.4ghz mouse stopped working after i updated to the 3.4 kernel (using over basically a hub which is originally connected via the otg).
<gzamboni>
thanks Alex
<TestModule>
Which file actually ends up controlling the max / voltage and max mhz ?
<mnemoc>
Alex1269: including mach/foo.h should hide the sun4i vs sun5i difference
popolon has joined #arm-netbook
<gzamboni>
the default bsp board fex(s) dont have the gpio_para , they could have it with gpio_used = 0
<mnemoc>
the gpio driver was hacked in to allow allwinner developers test boards, it was never intended to be used in production devices
<mnemoc>
it's even called "ugly gpio driver"
<ganbold>
mnemoc: I forgot you were saying about wemac, is it PCI-E MAC/PHY?
<mnemoc>
it's mac, and native in the soc
<mnemoc>
no pci-e
<ganbold>
mnemoc: on the back of cubieboard there is RTL1802CP, what is the use of it? as it acts as phy?
<mnemoc>
it's the PHY, yes
<gzamboni>
yeah, i saw it: SUN4I_GPIO_UGLY
<gzamboni>
it isnt also at the board defconfs, i will recompile it with this enabled to test it
<mnemoc>
iirc that driver only works for outputs. Alex1269 is reimplementing it using gpiolib
<gzamboni>
humm, thats good new
<gzamboni>
news
<hansg>
TestModule, the otg usb port stinks, really really stinks, you should always use a real usb host port where available.
<TestModule>
:\
<hansg>
TestModule, my guess is that before you changed the kernel, you either did not have the hub in between, or that you've added some other extra usb-device in the mean time
<TestModule>
That is not the case
Quarx has joined #arm-netbook
<TestModule>
I had to remove all but one (the side OTG) for space reasons
<hansg>
Basically the otg can only address a limited number of usb endpoints, so 1 device tends to work, a hub + 1 device may work, a hub + 2 devices tends to not work
<TestModule>
hmm, so built in "usb" wifi + hub = max
<hansg>
All I can advise you is to re-think the way how you've hooked things up, as the otg stinks big time
<hansg>
No the A10 has 3 usb controllers 1 otg, and 2 real proper controllers, The usb takes 1 proper controller usually :|
<TestModule>
but one is being used by the wifi card, correct?
<hansg>
Right, the usb-wifi takes one of the 2 real usb controllers, The usb connector you've "removed" is the other real usb controller, leaving you with only the suck majorly otg one
<TestModule>
nm, i think i know what to do... lots of soldering, lol
<hansg>
Yeah that is best.
<TestModule>
Has anyone tried Link-Time-Optimization on the kernel?
<TestModule>
aliexpress listings get updated first
<TestModule>
some sell on both
<slapin>
TestModule: ah, cool
<slapin>
my amounts are quite low, about 300-400 boards per year, so I still was unable to interest any Chinese supplier :(
<TestModule>
So, buy consumer gear?
<TestModule>
hell, craig A10 tabs were 9.99 @ cvs a few weeks ago ;)
<TestModule>
Start a kickstarter, feed off your own kickstart ;)
<TestModule>
???
<TestModule>
profit
cheng has quit [Quit: Leaving]
Quarx has joined #arm-netbook
lerc has joined #arm-netbook
lerc_ has quit [Ping timeout: 248 seconds]
hipboi has quit [Remote host closed the connection]
pcat has joined #arm-netbook
hansg has quit [Quit: Leaving]
popolon has quit [Read error: Operation timed out]
<Alex1269>
mnemoc, look http://sprunge.us/SjTI?c I made a eint sources table with ifdefs. Is this correct ?
<Alex1269>
I found in documentation that 32 gpios can be used as interrupt sources by default and another 32 using pin mode=6. But there is no information what happens with default eint0 pin after switching another eint0 source to mode=6
<mnemoc>
but... do we really need sunxi_gpio_to_irq() to export predeclared input and output pins on /sys/class/gpio ?
<Alex1269>
this function adds "edge" attribut into gpio sysfs directory. Then you can echo "rising" into it and poll on value attribut.
<mnemoc>
ok
<mnemoc>
fair enough :)
<Alex1269>
I think sun4i and sun3i eint enabled gpios differs
<mnemoc>
ignore sun3i for now
<mnemoc>
but sun5i is critical
<mnemoc>
(sun3i code is too incomplete to be usable, and no one here has hardware to test)
<mnemoc>
sun5i is more complicated as there are 2+ SoCs implementing it
<mnemoc>
maybe sunxi_gpio_to_irq() itself should be sun4i-only for now
<Alex1269>
so I added a table(s) :) but not sure about muxed eint gpios... It looks like they work as interrupt enabled inputs in mode=6 (http://linux-sunxi.org/A10/PIO) and defaults became non-interrupt enabled.
<Alex1269>
:) For non sun4i platforms currently I have "static structgpio_eint_data gpio_eint_list[] = { {-1,-1,-1} }" so gpio_to_irq will allways return -EINVAL for non sun4i
L84Supper2 has joined #arm-netbook
L84Supper2 has quit [Changing host]
L84Supper2 has joined #arm-netbook
vinifm has joined #arm-netbook
<Alex1269>
I downloaded this manual allready... :-)
<slapin>
how do you generally handle hundered thousand mails from maillists? I tried to convert my mbox to maildir, but nothing changed, everything is still very slow :(
<mnemoc>
for maildir you need a fs efficient with tons of small files
<slapin>
mnemoc: I don't see performance increase with conversion, everyone adviced me to convert to maildir :(
<specing>
slapin: you unsubscribe from LKML, lol
<slapin>
specing: I need it, and lotso of other subsystem mail lists to check various patches as work process :(
<specing>
ReiserFS?
popolon has joined #arm-netbook
<slapin>
specing: NO, no filesystem changes on this VPS
<mnemoc>
slapin: then enjoy an awfully slow mail server :<
<specing>
:D!
L84Supper2 has quit [Ping timeout: 246 seconds]
<slapin>
mnemoc: it is not slow, only mail reading is slow
<mnemoc>
:)
<mnemoc>
which just happens to be the most important feature of it :p
L84Supper2 has joined #arm-netbook
L84Supper2 has quit [Changing host]
L84Supper2 has joined #arm-netbook
<specing>
slapin: what client are you using?
<mnemoc>
slapin seems to be the kind of using "claws"
<mnemoc>
(rm too)
<specing>
I use claws
<mnemoc>
:D
<specing>
Ofcourse not on LKML, Im not that crazy
<specing>
Actually I was subscribed to LKML once
<specing>
when I was still using mutt
<rm>
I delete mail from/to mailing lists that's older than 1 year
<aholler>
so you would hold around 500k messages for lkml ;)
<slapin>
mnemoc: mutt
<slapin>
specing: mutt
<mnemoc>
hramrach: the cubieboard_defconfig is getting removed
<slapin>
can mutt cache anything except imap headers, which I don't use?
<slapin>
my 100K messages Maildir opens 30 seconds, my 160K messages mbox opens 48 seconds
<slapin>
they all tell me mutt caches mail, but i simply can't see where it does it, bith regular mutt and mutt-patched from Debian.
<aholler>
you have to enable it, e.g. with set header_cache=~/.mutt/tmp/.mutt
<aholler>
upps, should be ~/.mutt only
<aholler>
there is also a message_cachedir, just see man muttrc
<aholler>
a good solution is to use a local imap-mail-server. Thats what I'm using an arm-box for.
<slapin>
set header_cache is here from beginning, the directory is empty
<aholler>
maybe because it doesn't make much sense with mbox or maildir
<aholler>
as you already have all headers local
<slapin>
aholler: indexing?
<aholler>
I don't know how mutt works internally, but when using imap the headers will be cached.
<aholler>
(if enabled)
<slapin>
aholler: imap headers are really cached
<slapin>
aholler: or gmail use would be impossible for me (8M messages)
<mnemoc>
gmail messes with the reference headers :<
<mnemoc>
replicating in imap the same threading as in their web interface
<slapin>
well, Maildir is quite usable, as no mutt "Sorting" thing, so I can really use mail without throwing tantrums
hansg has joined #arm-netbook
<slapin>
it seems it is mutt's design pita
<slapin>
and mutt is not loading cpu to 100% any more
datagutt has joined #arm-netbook
<slapin>
any conversion tools for mbox to maildir?
<mnemoc>
slapin: several, google knows them :)
<slapin>
and what sets MAIL env var in Debian? lazy grep in etc have not found anything
<slapin>
but I thout the whole thing will be faster, I wonder...
cheng has joined #arm-netbook
* slapin
wants console mail client which would handle mail delivery/playing in separate thread, allowing me to read it in UI
<andoma>
isn't console mail clients written by old timers that hate threads and only use signals, setlongjmp(), etc?
<slapin>
andoma: and what if you hate setlongjmp() too?
gzamboni_ has joined #arm-netbook
<andoma>
well, then it's tough ..
<slapin>
andoma: I don't think so, remember mcabber, etc? the use case is to run this stuff on grey legal area VPS, and connect it from anywhere using ssh, and have the same setup all the time
<slapin>
andoma: have your open source email and work environment separate from personal mail and environment is cool, too.
<slapin>
andoma: I use gmail as personal mail, but my own mail server on VPS for open source work
<andoma>
i've got both dayjob and personal mail in gmail
<andoma>
not on the same account ofcourse
<slapin>
andoma: and different demands, different setup
<aholler>
if you use your own mailserver, why do you fiddle with mbox/maildir?
<slapin>
andoma: what do you propose?
<andoma>
but i'd like to use mutt or similar for my personal mail (still hosted on gmail), but i've never really bothered to set it up
<andoma>
i don't really propose anything
<andoma>
my setup it lame
<slapin>
andoma: I prefer mbox, but it and mutt can't handle 200K mails.
<slapin>
andoma: I run mutt on my mail server
<andoma>
i used to run my own mail server too but i always forgot how to configure stuff so eventually i got tired of it
<slapin>
not that Maildir is any better, but mutt at least allows me read mail, not always sorting it.
vinifm has quit [Quit: Ex-Chat]
<andoma>
now i just run a relay service for my domain in order to snatch off some mails that are supposed to be delivered to automatic systems since gmail can't really do that kind of forwarding AFAIK
<slapin>
andoma: just configure it once
Jonathan_Eyre has quit [Quit: Leaving.]
<andoma>
at least not free gmail
<slapin>
andoma: non-free too.
<slapin>
btw, is any imap server except zimbra support decent web mail?
<andoma>
no idea
<slapin>
zimbra is cool, but have too much demands...
<aholler>
sure, e.g. squirrelmail
<aholler>
there are several other imap-webmail-clients
<slapin>
squirrelmail is too primitive
<aholler>
so you have to define your requirements better
rsalveti has joined #arm-netbook
<aholler>
anyway, this is likely the wrong channel for such questions
<slapin>
aholler: not worse than zimbra, fast with firefox, can read stupid outlook and thunderbird users' mail, etc.
<slapin>
aholler: yeah
<aholler>
ask e.g. in #debian or try google first
<slapin>
my employer migrated from zimbra to ows, but I can't go this path
<slapin>
mutt eats 0.3% CPU and 100% memory, spamassassin eats 100% CPU and 0.2% memory - ho do they even live?
<aholler>
write something better
<aholler>
and don't use perl or python
<slapin>
this is also good use case for A10 devices, as infrastructure bricks, it is interesting how cost-effective this might be, having your mail server of 100 sticks
<andoma>
aholler: :)
<aholler>
(or even ava)
<aholler>
+h
<aholler>
hmpf, +j ;)
<slapin>
as I understand Maildir works perfectly via NFS, so the thing can be built just for appropriate load
<specing>
slapin> mutt eats 0.3% CPU and 100% memory, spamassassin eats 100% CPU and 0.2% memory - ho do they even live?
<specing>
lmao
<slapin>
specing: and they both on top
<slapin>
irssi runs there, too and very non-interactive
<specing>
hehe
<br->
specing: i'm presently working on exact gmail and reader clones, low enough spec to run on a little arm chip
<br->
more precisely, currently working on reader, but gmail is on the list
<br->
reckon you could have a 70% useful workalike with very little effort. currently integrating the crap i have with xapian, since it supports incremental indexing well
Jonathan_Eyre has joined #arm-netbook
<slapin>
br-: sounds interesting
<br->
it annoys me that people try and innovate on mail/reader by throwing out the winning UX and redesigning their own crap.. in the late 90s abiword was just plain and simple cloning ms office :) that worked
<Jef91>
imx devices historically have good Linux support, I was reading in another thread x11 drivers for opengl already exist for this imx6 chipset
<mnemoc>
RaYmAn has one
<mnemoc>
there is a 16GB model too
<Jef91>
Was thinking of picking one up
<Jef91>
I just want a "stick" PC that runs Linux well.
<mnemoc>
a good thing with the hi802 (aka gk802) is that uart pins are properly marked and it has a button to enable usb boot
<mnemoc>
err, not uart pins, uart holes :) ... you'll need to solder the pins
L84Supper2 has quit [Ping timeout: 248 seconds]
<slapin>
mnemoc: you mean haipad overpriced one?
ohniceidea has joined #arm-netbook
<slapin>
$100 is quite expensive one...
<mnemoc>
sure it will fall
piezo has quit [Ping timeout: 264 seconds]
<slapin>
mx6q tablets are not more expensive than allwinner or rockchip ones, but sticks...
<specing>
When are we going to see a cheapass MX6 board with SATA?
<mnemoc>
that's the "wand"
<specing>
"wand"?
<mnemoc>
wandboard
<slapin>
price on aliexpress is the same
<mnemoc>
i don't understand why they neglect to populate the sata header
<mnemoc>
it can't make such a difference in cost
<specing>
I don't care if they givem the the connector in the package
<specing>
also where is the quad version?
<mnemoc>
that was my first question when I saw it months ago too
sspiff has quit [Quit: Leaving]
<mnemoc>
"it would make the board too expensive to be attractive" is what I heard
<specing>
So packing 4 cores makes the board * 2 expensive?
<specing>
Also wtf is up with them and a single-core cortex-a9?
<slapin>
"it would make the board too expensive to be attractive" = "we don't want to do more than allwinner/rockchip guys as people buy it as is"
<slapin>
slapin@build:~$ cat /proc/17346/status
<slapin>
Name:mutt
<slapin>
State:D (disk sleep)
<slapin>
damnit
<slapin>
already a hour in such state, no activity.
<slapin>
it seems that Maildir+mutt is nono
piezo has joined #arm-netbook
<specing>
they won't get much purchases on an a9 solo board, I think...
<Jef91>
I just want a sub 100$ stick
<Jef91>
that runs Linux well
<Jef91>
and doesn't need Windows to flash images to it
<Jef91>
(like the RK3066 devices do)
<Jonathan_Eyre>
I've used repo sync to download the aosp, and I later removed one of the folders. Now repo sync fires an error on that folder, but it does not update it. Does anyone here know how to get repo to update a modified/missing folder?
<mnemoc>
Jef91: for allwinner sticks there is a livesuit for linux (x86_64) and free tools in progress ;-)
<Jef91>
I know mnemoc I like the a10 based devices
<Jef91>
they are just a bit slow
<buZz>
Jef91: i was looking at iMX6 devices, they look pretty sexy
<mnemoc>
i don't find the $120 of the hi802/16 *that* ugly
<mnemoc>
kind of on the limit
<Jef91>
link mnemoc
discopig has joined #arm-netbook
discopig has quit [Client Quit]
<traeak>
149 for the exynos 5 board...seems a little steep though...
<br->
is there a button you need to press on the wireless fly mouse to get the soft keys on the bottom of the screen to appear?
<TestModule>
mnemoc , 1GB ram?
<mnemoc>
TestModule: yes...
<Jef91>
So same as the gk802 but more internal storage mnemoc?
<mnemoc>
Jef91: gk802 is just a rebranding, so yes
SouL_ has joined #arm-netbook
hansg has quit [Quit: Leaving]
<slapin>
mutt eats all 8GB of RAM+swap and dies on 300MB maildir, wtf...
SouL_ has left #arm-netbook [#arm-netbook]
<slapin>
it seems there's no usable mail clients over there :(
fragmint has joined #arm-netbook
gimli has joined #arm-netbook
<Turl>
lol
<Turl>
slapin: my ImapMail is 1.2G and thunderbird behaves just fine if you can live with a graphical client
<slapin>
Turl: whatever client, I need to work with patches, and thunderbird makes such things very painful
<slapin>
Turl: alpine works well, but is not suitable for patches, too
<ohniceidea>
Hello everyone. I'm a newbie here. I found massive resources about hacking allwinner A1x chips, I wonder if Rockchip-based tablet is as easy as Allwinner for hacking? If so, Could anyone point me some related resources?
<slapin>
it is fascinating why mutt eats so large amounts of RAM when used with maildir, on mbox it seems normal, but slow
<Turl>
slapin: can't you right click-save as and then am that?
<hramrach>
yes, I tried mutt too
<hramrach>
long time ago
<hramrach>
it did not work
<hramrach>
obviously they did not fix it
<hramrach>
but ram is cheap so if it eats ram and works otherwise
Jonathan_Eyre has quit [Ping timeout: 245 seconds]
<slapin>
what is this nanocopter thing?
<mnemoc>
probably another device that will never reach the market
<mnemoc>
always innovating has a history on that
<slapin>
wow, what CPU is on this thing?
<slapin>
mnemoc: I have their touchbook, which is died for unknown reason (CPU doesn't start)
<mnemoc>
ouch
<Turl>
looks like a mini ardrone
popolon has quit [Quit: Quitte]
<slapin>
I found how I really can use mutt, but use dovecot's /usr/lib/dovecot/imap running from mutt, then it starts slower, but no catastrophic memory consumption
<slapin>
damn how they did it
Jonathan_Eyre has joined #arm-netbook
<br->
slapin, maybe check out mairix
<br->
i'm not sure how much ram it uses during indexing, but it only feeds mutt a mbox of the search results
Jonathan_Eyre has left #arm-netbook [#arm-netbook]
* br-
finds it impossible to go back to mutt after using gmail's search. hates gmail, can't live without it
<slapin>
and now it eats only 400MB RAM on 3GB maildir - not so bad.
<slapin>
so, mutt's Maildir support is sick
<slapin>
br-: gmail is not that good to work with patches
Kebianizao|work has quit [Remote host closed the connection]
<slapin>
I don't search mail too much with this account
<valhalla>
br-: I used to use gmail, now I use notmuch for email search (and mutt for everything else)
<br->
got lost in ruby hell last i tried to install notmuch, maybe worth a retry
<herdingcat>
what's the meaning of "l" in armv6l ?
<valhalla>
br-: I've just done an apt-get install notmuch :)
<br->
seems there is a macports package for it now. yes i know burning in hell already
<br->
waaiit.. when did notmuch get rewritten in C?
<br->
last i tried this it was a 50 ruby packages type nightmare
<ssvb>
where they seem to have all empty directories
<techn_>
oh
<mnemoc>
they don't do precise anymore
<mnemoc>
12.11 was the last
<mnemoc>
ssvb: what about moving your package to linux-sunxi's github? :)
<techn_>
glxgears fives 10fps fullscreen :(
<techn_>
240fps in small default screen
<ssvb>
techn_: with what kernel?
<techn_>
ssvb: 3.0.. but I have some clock modifications which could effect this
<ssvb>
techn_: and does it report any problems in /var/log/Xorg.0.log ?
<techn_>
ssvb: the normal ones [ 91952.204] (EE) AIGLX: reverting to software rendering
<ssvb>
techn_: defining DEBUG in sunxi_mali_ump_dri2.c should make it rather verbose
<ssvb>
anything like [175357.239] (II) SUNXIFB(0): enabled display controller hardware overlays for DRI2 ?
<ssvb>
or "GET_UMP_SECURE_ID_SUNXI_FB ioctl failed, overlays can't be used"?
<techn_>
yes.. I have it there.. but now I remember.. I have scaler layer enabled in my script bin
<techn_>
which causes that it wont get layer with scaling ability?
<ssvb>
no, scaler layer should just cause some funky effects like the layers at wrong positions or misplaced cursor
<ssvb>
but it should not affect performance
<techn_>
lol
<techn_>
I got jumping screen when I scaled es_gears window :D
<ssvb>
what is the screen resolution?
<techn_>
1080p
<ssvb>
it seems to be having troubles sending data over HDMI at this resolution, when something is heavily using memory at the same time
<ssvb>
reducing the framerate from 60Hz to 50Hz helps, but does not eliminate it completely
<ssvb>
techn_: just to clarify, did you get "GET_UMP_SECURE_ID_SUNXI_FB ioctl failed" message?
<techn_>
no.. I'll get 0x3 id when I rescale window
<techn_>
everytime
<ssvb>
it is normal
<ssvb>
how much FPS did you get with xf86-video-mali?
<techn_>
ssvb: I send you email
<techn_>
ssvb: Not much more ~30fps in fullscreen :/
<techn_>
or even less
<techn_>
but in smaller screen I got much less than with yours
<techn_>
I can rebenchmark
<techn_>
another observation.. when I move window top of gles window.. fps drop ie. from 190fps to 35fps
<ssvb>
techn_: I'm more worried about the fullscreen performance, seems like it just does not want to enable disp layer for it. The debugging output with DEBUG define in sunxi_mali_ump_dri2.c can help to identify the root cause
<techn_>
ok.. I'll enable that
<ssvb>
techn_: for the another observation, it is normal, the driver just disables overlay and fallbacks to doing memory copy if something is on top of the window
ibrah has quit [Quit: ChatZilla 0.9.89 [Firefox 18.0.1/20130116073211]]
<ssvb>
techn_: and because fbdev driver uses shadow framebuffer (for faster 2D), it is actually doing more memory copies than xf86-video-mali in the case of overlapped window
<techn_>
hmm.. what if you enable alphablending for that hwlayer??
<techn_>
and then just move that layer back.. or how it works :p
<techn_>
otherway around
<ssvb>
techn_: it's not so easy, but can be tried later
<ssvb>
in any case, I would guess that the users normally don't want to have some other windows on top
<ssvb>
and if the other windows are on top, the users are likely to be interacting with these other windows, so the FPS drop should not bother them too much
<techn_>
thats true.. trying to learn how things work :)
<mnemoc>
wiki wiki
<rellla>
mnemoc: no issues with 3.0 so far :P
<techn_>
rellla: scaler enabled in script.bin?
<rellla>
hm. dunno.
Alex1269 has quit [Quit: Page closed]
<rellla>
will play with script.bin tomorrow.
vinifm has joined #arm-netbook
<rellla>
and want to go on with mx xbmc analysis. mpeg2-ts is streamed very fine, but i have a few video files with pink artefacts, stuttering ...
<rellla>
s/mx/my/
<ibot>
rellla meant: and want to go on with my xbmc analysis. mpeg2-ts is streamed very fine, but i have a few video files with pink artefacts, stuttering ...
<rellla>
ready for investigation ... tomorrow ;)
<rellla>
techn_: should scaler be enabled? whats the matter for your question?
<techn_>
rellla: no.. I'm just checkin since I have it enabled
<techn_>
and ssvb doesnt have it enabled
<rellla>
i was turning it on/off/on/off and can actually not remember, how it is set *:)
<techn_>
screen jumps even on small screen.. now it's full screen.. not just window
tinti has quit [Quit: Leaving]
<techn_>
it must be lack of vsync
<techn_>
and some misalignment
<ssvb>
techn_: no, it looks like A10 just can't handle 1080p reliably
<ssvb>
techn_: I suspect that display controller can't fetch data from memory in time and substitutes it with some kind of padding, which causes this screen shaking effect
<ssvb>
techn_: it could also be that the display controller is misconfigured and has lower priority than mali with the regards to memory access
<ssvb>
techn_: going down to 720p fixes this issue completely
discopig has joined #arm-netbook
Jonathan_Eyre has joined #arm-netbook
<Jonathan_Eyre>
Is there a way to modify the kernel on a ROM?
<Jonathan_Eyre>
I've extracted the ROM to fex files using the a10 flash kitchen.
<ssvb>
techn_: the fullscreen jumping is "normal" in 1080p mode, it happens for me and rz2k also has it
<techn_>
ssvb: framebuffer xbmc didn't have it
<techn_>
but fps is there ~27
<ssvb>
techn_: then it is not taxing memory bandwidth too much
<Jonathan_Eyre>
I extracted the root.fex file from A10 flash kitchen, moved it to my virtual ubuntu, ran split_bootimg.pl on it to get the ramdisk + kernel, put the zImage and the ramdisk from split_bootimg.pl through mkbootimg, and then packed it back up with PhoenixCard. It was a fail.
<ssvb>
techn_: or has a bit different memory access pattern which does not trigger the problems
<Jonathan_Eyre>
Wow, the FirstStep wiki is REALLY different from the livesuit tutorials I've read. Does this work?
<techn_>
Jonathan_Eyre: firststeps are for sd card
<Jonathan_Eyre>
I do not have cubieboard, but I have very similar hardware.
<Jonathan_Eyre>
I've been struggling all week to get something to run on it, using livesuit tutorials and a combination of allwinner packs manually copied into 4.0.3 aosp.
<Jonathan_Eyre>
Painful.
<Jonathan_Eyre>
I'm not certain I have the fortitude to start over from scratch.
<Jonathan_Eyre>
It would be worth it, if it worked on my A10 hardware.
voronaam has quit [Quit: Leaving.]
<Jonathan_Eyre>
Hmm, I'm thinking I could go through the FirstSteps tutorial and get a working boot image.
voronaam has joined #arm-netbook
<Jonathan_Eyre>
Then I could use the botched system image to create a full SD Card, and at least see the boot try to load the incompatible system.
<Jonathan_Eyre>
Wouldn't that prove that it at least boots my kernel?
lerc_ has quit [Ping timeout: 264 seconds]
gimli has quit [Ping timeout: 264 seconds]
tinti has joined #arm-netbook
<rz2k>
ssvb: i had tearing all over the screen when scaler_layer was off
<rz2k>
if I remeber correctly
<rz2k>
it was when that Weng guy did the VLC port
hg_5 has joined #arm-netbook
Jonathan_Eyre has quit [Quit: Leaving.]
<ssvb>
rz2k: it happens both with and without scaler_layer, I partially solved the problem by reducing refresh rate to 50Hz for 1080p mode (screen shaking still happens sometimes, but is very difficult to trigger)
<ssvb>
rz2k: with scaler_layer enabled, the display controller may be just buffering a bit more pixels when fetching them from memory
<rz2k>
probably
<rz2k>
anyway, time for a nap :)
<rz2k>
good night.
rz2k has quit []
tinti has quit [Quit: Leaving]
popolon has quit [Quit: Quitte]
xymox has quit [Ping timeout: 260 seconds]
Jonathan_Eyre has joined #arm-netbook
xymox has joined #arm-netbook
specing has quit [Remote host closed the connection]
lerc has joined #arm-netbook
ganbold_ has quit [Remote host closed the connection]