mnemoc changed the topic of #arm-netbook to: EOMA: Embedded Open Modular Architecture - Don't ask to ask. Just ask! - http://elinux.org/Embedded_Open_Modular_Architecture/EOMA-68 - ML arm-netbook@lists.phcomp.co.uk - Logs http://ibot.rikers.org/%23arm-netbook or http://irclog.whitequark.org/arm-netbook/ - http://rhombus-tech.net/
dyoung-away is now known as dyoung
dyoung is now known as dyoung-away
freakazoid0223 has quit [Quit: Leaving]
hg_5 has quit [Ping timeout: 248 seconds]
Ershov has joined #arm-netbook
stefanro1 has joined #arm-netbook
stefanro has quit [Ping timeout: 248 seconds]
voronaam has quit [Quit: Leaving.]
bfree_ has joined #arm-netbook
bfree has quit [Ping timeout: 252 seconds]
WarheadsSE has quit [Quit: WeeChat 0.4.0]
cnxsoft has joined #arm-netbook
bfree__ has joined #arm-netbook
bfree_ has quit [Ping timeout: 252 seconds]
Aex1269 has joined #arm-netbook
freakazoid0223 has joined #arm-netbook
valhalla has quit [Read error: Connection reset by peer]
herdingcat has joined #arm-netbook
muts has quit [Ping timeout: 252 seconds]
muts has joined #arm-netbook
Jef91 has joined #arm-netbook
tinti_ has quit [Quit: Leaving]
WarheadsSE has joined #arm-netbook
br-_ has joined #arm-netbook
aholler has quit [*.net *.split]
br- has quit [*.net *.split]
acfrazier has quit [*.net *.split]
acfrazier_ has joined #arm-netbook
rsalveti has quit [Ping timeout: 248 seconds]
rsalveti has joined #arm-netbook
eebrah has quit [Remote host closed the connection]
acfrazier_ is now known as acfrazier
Jef91 has quit [Remote host closed the connection]
aholler has joined #arm-netbook
freakazoid0223 has quit [Ping timeout: 240 seconds]
aholler_ has joined #arm-netbook
aholler has quit [Ping timeout: 248 seconds]
KoH_ has joined #arm-netbook
KoH__ has quit [Ping timeout: 248 seconds]
Gumboot has quit [Ping timeout: 246 seconds]
rz2k has joined #arm-netbook
gimli has joined #arm-netbook
Gumboot has joined #arm-netbook
L84Supper has joined #arm-netbook
dyoung-away is now known as dyoung
Ershov has left #arm-netbook [#arm-netbook]
dyoung is now known as dyoung-away
gimli has quit [Ping timeout: 240 seconds]
gsilvis_ has joined #arm-netbook
DEAT has quit [Read error: Connection reset by peer]
DEAT has joined #arm-netbook
gsilvis has quit [Ping timeout: 256 seconds]
aholler_ is now known as aholler
Quarx has joined #arm-netbook
pcat has joined #arm-netbook
rellla has joined #arm-netbook
doneill has joined #arm-netbook
<doneill> hey guys. anybody played with a q88 with a zet6221 capacitive touch controller on it? is there module source kicking around somewhere?
valhalla has joined #arm-netbook
<Aex1269> :(
<Aex1269> Can't get A10 gpio interupts working... :((((
<Aex1269> Its just not working. No int28 at all... :((((
<Aex1269> Is this correct interrupt no ?
hansg has joined #arm-netbook
rellla2 has joined #arm-netbook
sspiff has joined #arm-netbook
rellla has quit [Ping timeout: 276 seconds]
cobalt60 has quit [Ping timeout: 240 seconds]
cobalt60 has joined #arm-netbook
niftylettuce has quit [Ping timeout: 252 seconds]
Avernos_ has joined #arm-netbook
Avernos has quit [Ping timeout: 248 seconds]
herdingcat has quit [Ping timeout: 256 seconds]
herdingcat has joined #arm-netbook
simosx has joined #arm-netbook
simosx has quit [Changing host]
simosx has joined #arm-netbook
aesok has joined #arm-netbook
torindel has joined #arm-netbook
pcat has quit [Ping timeout: 255 seconds]
<Turl> Aex1269: most GPIOs don't have interrupts on A10
<Turl> so you need to poll them
swivelgames has quit [Read error: Connection reset by peer]
swivelgames has joined #arm-netbook
Thomas42 has quit [Ping timeout: 264 seconds]
Thomas42 has joined #arm-netbook
Alex1269 has joined #arm-netbook
<Alex1269> Turl, I known about non-int gpios. I tried pg8-pg9, ph14-ph15 (mux6 mode) but no success....
<Alex1269> I'm working with cubieboard. Look code http://sprunge.us/QaaM?c
<Alex1269> I connected pg8 with pg9 through 1K resistor and then switched pg9 to output mode and pg8 to input.
<Alex1269> when I echo 1 > value in pg9 I got value=1 in pg8.
popolon has joined #arm-netbook
<Alex1269> So then I tried to check interrupt code and got nothing :( /proc/interrupts show me 28: 0 sw_vic sunxi-gpio
<Turl> Alex1269: where did you read those pins can interrupt?
<Turl> try with PH0-21, PI10-19 on mux 6
niftylettuce has joined #arm-netbook
<Alex1269> I tried ph14-ph15 in mux6 mode...
<Alex1269> Pin in this mode allways return value=1 and doesn't throw interrupt :(
<Turl> Alex1269: what mode did you configure btw?
simosx has quit [Quit: Αποχώρησε]
<Alex1269> I passed 6 to gpio_set_one_pin_status sys_config.c function
<Turl> what does line 322 on your code print?
<Alex1269> mul_sel member
<Turl> I mean what do you see as output of "322 pr_info("TYPE written %08lX for irq%d (offset %d) cfg=%X\n", (unsigned long)reg_val, d->irq, offset, int_cfg_addr[reg_num]);"
<Alex1269> 5 sec...
<Alex1269> TYPE written 00400001 for irq109 (offset 13) cfg=204
<Turl> so the register reads
<Turl> 0000 0100 0000 0000 0000 0000 0001
<Turl> double edge for pin 9
<Turl> or that's what I'd expect though
<Turl> (INT9)
<Alex1269> hm...
<Turl> INT9 would be PH9 on mux 6
<Alex1269> Ooops... This was ph14 sysfs node. It looks like I made a bug somethere
<Alex1269> Turl, I checked this twice. How did you get int9 ?
<Turl> Alex1269: register 0x204 handles INT8-INT15
<Turl> so you see INT9 is 0100b2 (0x4)
<Turl> unless I'm reading the register upside down
<Alex1269> Yes. But low bit is on the right side 0x00400001
<Alex1269> So this is ph13
<Turl> indeed ph13
<Turl> I missed a 0000 too
<Turl> so it'd have been ph10
<Alex1269> anyway - I tried 0x04040404 for all conf registers and 0xffffffff for gpio int control and .... nothing irq28 zero shots....
<Alex1269> sorry.. 0x44444444 for conf...
<Turl> and pins correctly muxed to 6?
<specing> lol GPIOs
<specing> took me a while to blink a led on a cortex-m4f
<Turl> Alex1269: if you print PIO_INT_STAT_OFFSET do you see all 0?
<Alex1269> Not sure about mux... :) I'll check this one more time. 5 min...
<Alex1269> Adding status read/print...
<Turl> also I suppose you unmask your interrupt somewhere?
<Alex1269> Turl, there is *unmask function in code. Is this not enough ?
<Turl> Alex1269: I don't see anything calling them
<Alex1269> it called indirectly by gpiolib code when I echo "rising" > /sys/class/gpio/gpio5_ph14/edge
<Alex1269> gpiolib call gpio_to_irq and receive virtual irq number handled by virtual chip in my driver. Then it call mask/settype/umask
<Turl> maybe you need to clear pending irq before unmask? hm
<Alex1269> [ 97.360000] TYPE written 44444444 for irq109 (offset 13) cfg=204
<Alex1269> [ 97.360000] UMSK written FFFFFFFF for irq109 (offset 13)
<Alex1269> [ 97.360000] status: 00000000
<libv> techn is gone... but here is the answer anyway
<libv> my prototype driver runs 2% faster with 3% less cpu usage
<Alex1269> but something happens - it hungs after I pulled up second gpio connected to this one...
<Turl> Alex1269: and now you go low->high on pin and irq doesn't trigger?
<Turl> Alex1269: :p
datagutt has joined #arm-netbook
<libv> i am scheduling slightly more aggressively (and get a bit worse per frame latency in some cases, and more context switches)
<Turl> 2% faster, neat
<libv> and i do not do any checking in my graphics driver
<libv> so this proto driver needs to be faster to prove that i know everything that there is to know
<Alex1269> So i need to connect serial cable and look for messages... It looks like interrupt shot with bad handler....
<libv> i have optimised things to be as fast as the original driver now
<libv> and all the changes are generic
<libv> not q3a related at all
<Turl> libv: functionality wise are you on par with the proprietary driver?
<libv> all the other demos have had 2-5x increase in performance due to those things
<libv> Turl: for the limited functionality that i support, yes
<libv> Turl: but it is not GL, and applications need to be specifically ported
<Turl> let me rephrase my question then
<libv> with mali, there is no software trickery
Alex1269 has quit [Quit: Page closed]
<libv> the buildup of the command stream, and the way the hardware works, with its crazy vertex shader...
<libv> that is where all the optimization has already been done.
<Turl> with the knowledge you have now and infinite time and resources allocation, are you in a position to write a new driver in feature parity?
<libv> those norwegian demosceners optimized in the right places!
<libv> Turl: yes, and with the same performance.
<Turl> awesome :)
<libv> mali is awesome :)
<libv> i still haven't crashed it
<Turl> heh
<Turl> I have
<Turl> not by sending crap commands though
<libv> i have gotten thousands of bad renders, sometimes got it to complain to me about SIGSEGVs
<libv> but no kernel or hw lockups (that weren't recovered immediately)
<libv> Turl: overclocking will do it
<libv> but nothing else will
<Turl> libv: yeah, that's how I hung mine a couple of times
<libv> which is absolutely unique
<Turl> too high a clock
<Turl> goes unstable
<Turl> after a bit of load on the 400s iirq
<libv> Turl: which is very very logical
<Turl> yeah, I don't blame it
<libv> Turl: my allwinner CPUs are unable to handle 1.2GHz for instance
<Turl> mine do, but not stable at all
<libv> my mele only does q3a reliably if i clock it at 1.152, my tablet doesn't even do that
<Turl> 1.1x works most of the time
<Turl> most other arm hardware usually can clock 1.5x the factory clock
<libv> yeah, i am thinking about getting an ODROID-X2
<libv> up to 2GHz quad A9s, 600MHz Mali400-MP4
<libv> i want to figure out how to use multiple fragment shaders
<oliv3r> what soc is that?
<libv> exynos
<oliv3r> nice
<oliv3r> when will the video be up? :)
<oliv3r> i wanna see what happend at fosdem
<libv> and then i should get about 4x the performance on q3a and see whether my scheduling needs adjusting or not
<libv> i will put a run of the demo up soon
<libv> recorded on sunday during the only 2h that i actually went to talks
aesok has quit [Remote host closed the connection]
<Aex1269> :( Ok... It just hung without any messages... Console useless.
<Turl> Aex1269: :/
<Turl> libv: U2 is cheaper btw
<Turl> and U even less, if you can live with the lower clock and half the ram
<SPG> vivien master 024c73c rhombus allwinner_a10/pcb/discussion.mdwn * http://git.hands.com/?p=rhombus.git;a=commitdiff;h=024c73c
<libv> Turl: i cannot, an odroid will be ordered to balance performance even better :)
<libv> i now want to see what the mali is capable of
<Turl> 45000 sdcard connectors heh
<Turl> libv: U2 and X2 should be the same hw wise, just more compact and less IO
<libv> Kusma, the norwegian who is doing tegra-re, who worked at phalanx and arm until 4 years ago, and who was at fosdem...
<xenoxaos> the u2 and x2 are slightly different
<libv> he did not know that on the exynos good-yield chips, the mali-400 clocks to 600MHz
<libv> it put a rather big grin on his face
<xenoxaos> the x2 has one HSIC bus with all usb on that
<libv> and then he immediately said "wow, i hope the memory can keep up"
<xenoxaos> the u2 has 2 hsic busses, one is used for ethernet, the other for all other usb stuff
dyoung-away is now known as dyoung
<Turl> xenoxaos: sure it's not the other way round?
ganbold_ has joined #arm-netbook
<xenoxaos> yup
<Turl> so it's more compact and better designed, I would have expected it to be the opposite :)
<xenoxaos> except no lvds out
* xenoxaos heads off to work
hansg has quit [Quit: Leaving]
Jef91 has joined #arm-netbook
<ssvb> libv: do you know that there is #odroid irc channel here?
<Aex1269> found it.... IRQF_ONESHOT needed. Now it works.
<Aex1269> Turl, as I understand there is no input mode eint capable gpios - only mux6
<Aex1269> Is there any information about this mux mode ? I just checked - it allways returned 1 as pin value
<libv> ssvb: ah, i didn't but i am also not in the olimex channel :)
<ssvb> well, there is a guy who is maintaining the kernel for odroid devices, so it is a useful place to be lurking in :)
rz2k has quit [Read error: Connection reset by peer]
<libv> ah, when that doesn't work like expected, i will go there and complain then ;p
<Turl> Aex1269: I don't know really
<Turl> Aex1269: but if interrupt works you could connect your input on parallel on both and when you get interrupt, pull a value from the other one :p
<andoma> anyone using r3p0 drivers in framebuffer mode? I upgraded kernel and switched from r2p4 to r3p0 and now my code fails because it cannot create a GLContext
<libv> andoma: yes, i am
<andoma> just wanted to check if i've messed something up with my kernel or and/if there are some problems with the gldrivers or so
<Aex1269> :))
<andoma> libv: ah ok
<libv> andoma: try the test in mali-libs
<andoma> will do. i got to the point where i failed very late last night so i kinda gave up and went to bed
<andoma> gonna try more today once i get home
Aex1269 has quit [Quit: Page closed]
<andoma> thanks
<libv> andoma: do you permission to access to /dev/ump, /dev/mali and /dev/fb0
<libv> +have
<andoma> not sure tbh as i'm not with my device right now, but i'll check that too
<libv> which reminds me that i should still hack this test to support X11
<buZz> libv: made any cool announcements at fosdem? :D
<libv> will be a 15 minute job, but it will take a lot longer to set things up so that X11 works with mali
<libv> buZz: q3a
<libv> buZz: 2% faster than binary driver
<buZz> not posted on your site yet?
<buZz> ah coolness :D
<libv> i woke up 1h ago
<libv> and fosdem is social madness
<libv> very little sleep, bad for the liver
<libv> but fantastic fun
<ssvb> libv: btw, thanks for these mali related talks at fosdem, they provided a lot of useful information
<libv> so i did not have time to write up the relevant blog entry during the event
<libv> ssvb: were you there?
<libv> we had hansg out with us on friday, and zaEarl and his wife ran into us at exactly the right time for supper on sunday
<ssvb> libv: yes, sorry that I did not introduce myself, but you seemed to be quite busy talking with the other people and with the fosdem staff
<mnemoc> :)
<buZz> :)
<libv> ssvb: with all the devroom madness, saturday morning i am indeed unapproachable
herdingcat has quit [Quit: Leaving]
<libv> ssvb: but after that...
herdingcat has joined #arm-netbook
<andoma> shame that cwabbott's presentation got screwed up though :(
<libv> every evening we get a crowd together
<libv> he rescued quite well
<libv> he covered what he needed to cover
<libv> but the slides were a bit bonkers
<andoma> sure
<andoma> ineed :)
<andoma> indeed..
<libv> he made good sense though
<libv> but it all was a bit much in too short a space of time
<andoma> yup
vinifm has joined #arm-netbook
<Turl> hno: ping
<buZz> libv: G2D is not part of Mali, right?
<Turl> libv: it's an allwinner thing
<buZz> Turl: that was aimed at me i bet ;)
<Turl> yeah
* Turl is failing a lot today
<buZz> is there any project to build an accelerated driver actually using G2D for xorg or similar?
<Turl> buZz: talk with ssvb
<buZz> yeah i saw his driver, but afaik it was just re-implementing all that already existed?
<buZz> > xf86-video-sunxifb - video driver for the devices based on Allwinner A10/A13 SoC
<ssvb> buZz: I have a lot more code that is not public yet
<ssvb> buZz: and the inital xf86-video-sunxifb release is deliberately dumbed down, because I first want to make sure that it is stable
<buZz> :D heh cool
<buZz> so 2D acceleration 'might' be coming? :)
<buZz> with FOSS drivers?
<buZz> and i bet libv's work will them end up as a Gallium driver or similar .. hmm nice
<mnemoc> ssvb bet xf86-video-mali and libv bet libmali.so, nice job ARM :p
<buZz> :P
<buZz> also noticed my phone has an Adreno gpu ..
<buZz> maybe i should try to debian it ;)
<mnemoc> s/bet/beat/g
<ibot> mnemoc meant: ssvb beat xf86-video-mali and libv beat libmali.so, nice job ARM :p
<libv> mnemoc: no, i did not beat the phalanx driver
<libv> the mali is just too good
<libv> i equalled the phalanx driver
<libv> but i am beating ARM in a completely different way :)
<mnemoc> :D
Jef91 has quit [Remote host closed the connection]
<ssvb> libv: is GP in mali really kind of like a general purpose DSP as advertised in http://web.archive.org/web/20060515063211/http://www.falanx.no/download/Mali%20Geometry%20Product%20Spec%20USL.pdf ?
<ssvb> libv: if so, then it might be useful for 2D acceleration
<libv> it is a scalar processor that can very very fast if everything is timed right
<libv> no
<libv> gp should not be used for 2d accel
<libv> pp should, and we are not far off from removing the gp/plbu from that equation and giving us the ability to use the pp with limited overhead
<Turl> fuu watchdog, why is it that 5 seconds feel like 10-15
vinifm has quit [Quit: Ex-Chat]
<libv> ssvb: how would one use the gp for 2d anyway?
<libv> ssvb: gp transforms vertices and relations
<libv> it does not deal with textures
<libv> it's a vertex shader
<ssvb> libv: any general purpose processor can do 2D, it is more important to have fast access to memory (which is also seemed to be advertised in that pdf)
<ssvb> libv: NEON unit in ARM core is already several times faster than necessary for simple 2D graphics
<ssvb> libv: but if PP is better for this task, then so be it :)
<libv> ssvb: flatmush and cwabbott both believe(d) that the t6xx shader would be a 32bit variant of the mali pp
<libv> exas compose is what, combining textures?
<libv> then pp is the place to be
<libv> all the other operations seem like something g2d should manage for us
<libv> for the compose operations that g2d can do for us, g2d should also be used
<libv> it is all a bit of hassle, i know
<libv> but it makes much more sense than trying to balance assembly all day
<ssvb> the problem is, g2d can't really do much
<libv> i never got why people prefer playing with assembly over just implementing the blocks of silicon so those can be used like they were intended to
<libv> ssvb: you told me that before
<aholler> libv: everything can be fixed by sw
<libv> ssvb: g2d is there, requires little setup, it should be used as much as possible
<aholler> thats the credo of hw-peopel
<libv> aholler: nope
<libv> aholler: cfr the mali and the phalanx guys
<libv> this is why i like the mali so much now that i have q3a running this fast
<libv> it proves that those hw people who say that are absolute idiots
<aholler> exactly
<aholler> but you can't tell them
<libv> you can now.
<aholler> and managment doesn't care
<libv> and that is what my q3a announcing blog entry will finish off with
<libv> i am the one who brought initial structure in modesetting
<libv> and i had to hear all the time how modesetting was easy while 3d was hard
<libv> such a big lie.
<aholler> many hw has become unbelievable broken.
<libv> 3d chip designers were just idiots.
<libv> the mali guys proved that
<aholler> not just 3d chip designers.
<Turl> aah, no wonder it felt like a lot of time, the timeout is not adjusted :<
<libv> and this is why mali was the king in 2011 and 2012
<aholler> hw-people have to release something every few month, and thus there seems to be no time to find and fix even the worst bugs. it's easier to write an errata and let the sw-guys fix it all
<libv> aholler: and then you end up with powervr :)
<libv> which is exactly that, baked into the gpu from the getgo
pcat has joined #arm-netbook
br-_ is now known as br-
<aholler> it's especially teasing if the sw-guys don't belong to the hw-company (=> no costs)
<Turl> samsung? :p
<Turl> mripard: ping
<hramrach> how fast is the single shader mali when it works?
<hramrach> compares to TNT2? radeon 7000? 9000?
<Turl> heh, I have a RIVA TNT :)
<hramrach> I have a TNT2 lying around somewhere I think
<Turl> those days GPU had small coolers :)
<hramrach> were good cheapo cards with decent VGA DACs
<hramrach> they also melted a lot
<Turl> my TNT never melted (luckily?)
<hramrach> luckily.
<hramrach> or has 20% larger cooler which was still comparably small but good enough
<Turl> I had an nvidia 4Kish burn in some way and output purple stuff at times though, to end up dying.
<hramrach> some just cut corners too much
<Turl> MSI
hg_5 has joined #arm-netbook
<hramrach> seems that the brand with small green brick-like cooler melted a lot but the one with black 1" pins survived
<hramrach> basically what you see on chipsets today, heh
aholler has quit [Quit: 3.7.6]
rellla2 is now known as rellla
aholler has joined #arm-netbook
adibek has joined #arm-netbook
hg_5 has quit [Ping timeout: 264 seconds]
wingrime has joined #arm-netbook
dyoung is now known as dyoung-away
sv has joined #arm-netbook
discopig has quit [Ping timeout: 276 seconds]
herdingcat has quit [Remote host closed the connection]
Marex has quit [Ping timeout: 264 seconds]
Marex has joined #arm-netbook
cnxsoft has quit [Remote host closed the connection]
revident has joined #arm-netbook
ganbold__ has joined #arm-netbook
sv has quit [Ping timeout: 276 seconds]
gimli has joined #arm-netbook
discopig has joined #arm-netbook
ganbold_ has quit [Ping timeout: 256 seconds]
doneill has quit [Ping timeout: 276 seconds]
orly_owl has left #arm-netbook [#arm-netbook]
aesok has joined #arm-netbook
gsilvis has joined #arm-netbook
gsilvis_ has quit [Read error: Connection reset by peer]
ibot_ has joined #arm-netbook
ibot has quit [Ping timeout: 252 seconds]
ibot_ is now known as ibot
ibot has quit [Remote host closed the connection]
ibot has joined #arm-netbook
hg_5 has joined #arm-netbook
Lofde_ has joined #arm-netbook
rellla has quit [Remote host closed the connection]
<Lofde_> hello i am looking for a evb.bin file that is set for the mk802 to 1080p60hz .. does anyone have one i can download please?
<rm> also you can always simply make your own
<rm> the proper name for this file is "script.bin", only older bootloaders/images expect it to be called "evb.bin"
<rm> there's a tool called bin2fex, which decodes it to human readable format
<rm> then you can edit it
<rm> then fex2bin
<Lofde_> yeah i was reading about it.. and looked at the git repo at the files, but wasnt so sure about compiling it for my ubuntu 12.04 box
dyoung-away is now known as dyoung
<Lofde_> thankfully the file i found online with a 24hz refresh rate worked.. thanks for your help rm
<Lofde_> now if i can just find the way to adjust for the overscan hehe
<focus> An A10 CPU based prototyping / breakout PCB in KiCAD is finished and uploaded http://www.gplsquared.com/SoM2/SoM2.html - have fun! :)
<Lofde_> nice found a menu option on the tv to fix it :D..
<focus> goto go - catch up later
<Lofde_> is this the main channel to discuss issues or features with mk802 and linaro release or should i go to a different channel
dyoung is now known as dyoung-away
fragmint has joined #arm-netbook
pcat has quit [Read error: Connection reset by peer]
techn_ has joined #arm-netbook
<rm> Lofde_, afaik there isn't a better one
valhalla has quit [Read error: No route to host]
<rm> there is also #cubieboard for http://cubieboard.org/ which is mostly the same hardware as the MK802 and runs same software
valhalla has joined #arm-netbook
pcat has joined #arm-netbook
<adibek> Hi everyone I created fex file from my MK802+ (cs102) script.bin from /dev/block/nanda , Can you tell me how to use this to making bootable SDcard with debian ???? Fex file : http://pastebin.com/AzYsm39k
arokux has quit [Ping timeout: 244 seconds]
<specing> adibek: wiki
arokux has joined #arm-netbook
bfree__ is now known as bfree
Quarx has quit []
arokux has quit [Remote host closed the connection]
arokux has joined #arm-netbook
ln2 has joined #arm-netbook
jirnenkaya has joined #arm-netbook
<mripard> Turl: pong
rellla has joined #arm-netbook
<libv> hramrach: 1 pixel per clock per pp
tinti has quit [Quit: Leaving]
<libv> hramrach: 35MTris for our mali 400 at 320MHz
<libv> so 320MPix and 35MTris
<libv> for the odroid-x2, at 600MHz, we get 65MTris and 2.4GPix
<traeak> what's the odroid x2 soc ?
<libv> exynos 4
<libv> so, a quad core a9, at max 2GHz
<hramrach> hmmm, that would be somewhere in the range r128 - radeon 9500 wrt fillrate
<hramrach> Tris are not listed for radeons in the tabel I am looking at so can't really compare
<hramrach> good for a somall system I guess
<hramrach> small
<libv> beating geforce2 iirc
<libv> heh, 3.
<hramrach> I guess for fillrate the memory interface is primary which is about the same as cards in that range
adibek has quit [Quit: Page closed]
pcat has quit [Ping timeout: 264 seconds]
<libv> and the lower end of r200 apparently
<techn_> libv: how difficult it would be to implement opencl driver for mali.. or is it even possiple?
<hramrach> HD 6310 has a bit lower fillrate than the odroid which is slowish but usable
<hramrach> and that is determining factor for low performance of that card
hg_5 has quit [Ping timeout: 264 seconds]
hg_5 has joined #arm-netbook
<hramrach> the geometry numbers are something I cannot compare and not sure to which extent it is relevant
<techn_> .. seems that ARM has implemented OpenCL only for newer gpus
<hramrach> because the geometry engines of these low-end cards sit unused in anything but artifical benchmarks due to low memory bandwidth
<hramrach> also the mali numbers will most likely change as the driver matures
tinti has joined #arm-netbook
pcat has joined #arm-netbook
<hramrach> techn_: OpenCL is probably doable when you know how the card works
<hramrach> it's the compute stuff, right?
<techn_> open computing language. yes
freakazoid0223 has joined #arm-netbook
<hramrach> you need cmpiler to the respecitve GPU assembly and some interface for uploading the programs and downlaoding the result
<hramrach> unless the assemly is really limited compiler is possible. just some(or lots) work
dyoung-away is now known as dyoung
<hramrach> GPU crypto ;-)
<libv> actually, 3x...
<libv> 195MVertices
<libv> techn_: about opencl, i do not know opencl
<libv> and you would have to ask connor about that
<hramrach> libv: I guess the fillrate number is interesting but the other number which is interesting but missing is MTex
<hramrach> it may differ a lot
<libv> i guess it is just below 1flop per clock
<libv> err, no, 1 flop per 3 clocks.
<libv> i would have to ask connor
<libv> although, 1 vertex is also 3 floating point numbers, so yes just under 1 flop per clock
* libv should get his head straight
<hramrach> 1 flop 3clocks would give like 750Mflops on Odroid
<libv> hramrach: 600MHz 1 core
<libv> GP is the thing pushing out vertices
<hramrach> with 195Mvertices this would be on par with some Radeon 9x again
<hramrach> the newer cards have way higher compute rate even in the low end
<libv> sure, but the newer cards have unified shaders
<hramrach> Mali has not? I have no idea about the arch,really
<libv> ~1 flop per clock for the GP, ~3 halffloat ops for the PP per clock
<hramrach> ah
<hramrach> so the MTex is really interesting number to collect then
<hramrach> and benchmarks aside it also remains to be seen how it fits together
<hramrach> lots of bugs to work around on the radeons slowing things down
adibek has joined #arm-netbook
<adibek> I use wiki and my fex http://pastebin.com/AzYsm39k to buld linux image for Mk802+ but still no effect , Three kinds of SDcards and no efect :(
<hramrach> by the perf numbers an Odroid would be about on par with netbook onboard graphics like HD 6250 - HD 6320
<hramrach> adibek: no effect in waht way?
libv has quit [Ping timeout: 264 seconds]
<adibek> techn_: Thx i will check out this forum
ganbold__ has quit [Ping timeout: 276 seconds]
<hramrach> also if you said what the problem is maybe somebody would know how to solve it ;-)
SouL_ has joined #arm-netbook
SouL_ has left #arm-netbook [#arm-netbook]
jirnenkaya has left #arm-netbook [#arm-netbook]
libv has joined #arm-netbook
<rm> I finally got myself a Raspberry Pi
<rm> two in fact
<rm> now what the heck do I do with them
<rm> all my "low power ARM computer" needs are already covered by the MK802's and the Mele :)
<hramrach> you lend one to specing so he can buy the mpeg playback lic and then go complain to mpeg-la that ther eis no sw patents in EU so he wants the money back :>
<RITRedbeard_> Poop on it and leave it on Eben Upton's door step... err yacht bow in a flaming brown bag.
adibek has quit [Quit: Page closed]
<hramrach> I have no mk802 :o
<hramrach> but then maybe I have fewer low-power ARM computer needs
<hramrach> I mostly need more powerful computers than A10
<fragmint> theres always the ug802 or the freescale quad core stick thingies
<hramrach> the power I need is better interconnect
<hramrach> eg. PCIe or more SATA ports
<fragmint> true. =*(
<hramrach> the minimum NAS has like 1 GBit ehternet and a few SATA ports
<hramrach> and that would be very nice application for low-power ARM computer
rellla has quit [Remote host closed the connection]
<fragmint> I know the pogoplug has gbit and one of them has an internal sata port
lejonet has quit [Ping timeout: 255 seconds]
<hramrach> my current board has 2 SATA
<specing> hramrach: Which one?
<hramrach> it's a x86 board
<specing> ah
<hramrach> somewhere on my todo list is to take the machine apart and put in one with more ports
lejonet has joined #arm-netbook
<rm> I just use a full-blown amd64 server with 8 HDDs
<WarheadsSE> Depends on how much processing power you end up needing.
<rm> and actually one of the MK802s is an "X-terminal" to it
<WarheadsSE> A lot of these with gigglebit don't quite actually hit that speed
<hramrach> not much
<fragmint> rm, are you running raid5?
<hramrach> but the PCIe /SATA/GBit ethernet does not come without
<WarheadsSE> If you think you are going to actually saturate the spec, you're in for a surprise
<rm> fragmint, RAID6
<fragmint> software or hardware?
<rm> software (mdadm)
<fragmint> any clue what the speeds are like?
<hramrach> no, the GBit card I have is cheap realtek that doess 400MBit
<hramrach> but that's better than 100MBit and about as much I need given disk/USB/other bus limits
<hramrach> I could replace it with an Intel card that can +/- saturate the spec but no need
<hramrach> and yes, you probably need the Intel card when you don't have much computing power because the Realtek would kill the CPU with IRQ storms or other overhead
<WarheadsSE> lol
<WarheadsSE> There are several marvell SoCs that support sata w/ multiplexers
<WarheadsSE> the GbE on them is alright
<hramrach> haven't seen a reasonably priced board with linux support that uses such SoC
<hramrach> also SATA multiplexers aren't exactly reliable technology in my experience
<hramrach> tried an external enclosure which claim multiplexing but does not work with two different sata contollers
<WarheadsSE> did those sata controllers support multiplexing?
<WarheadsSE> not all do, after all
<hramrach> reportedly yes
rellla has joined #arm-netbook
datagutt has quit [Quit: kthxbai]
ChanServ has quit [*.net *.split]
ChanServ has joined #arm-netbook
ln2 has quit [Remote host closed the connection]
ln2 has joined #arm-netbook
popolon has quit [Quit: Quitte]
ChanServ has quit [*.net *.split]
ChanServ has joined #arm-netbook
ln2 has quit [Remote host closed the connection]
ln2 has joined #arm-netbook
ChanServ has quit [*.net *.split]
ChanServ has joined #arm-netbook
ln2 has quit [Remote host closed the connection]
ln2 has joined #arm-netbook
eFfeM has joined #arm-netbook
voronaam has joined #arm-netbook
<gzamboni> video.fosdem.org/2013/ is on its way
<buZz> w000t
dyoung is now known as dyoung-away
freakazoid0223 has quit [Quit: Leaving]
<andoma> libv: ping .. still can't get mali in fb to work :(
<andoma> Error: eglCreateContext failed: 0x00003005
piezo has quit [Quit: Quit]
rz2k has joined #arm-netbook
ln2 has quit [Remote host closed the connection]
<Turl> mripard: I got hno's watchdog driver for sunxi working with dt :)
<mripard> great :)
<Turl> rm: install xbmc on them, that's one thing they can do well
sv has joined #arm-netbook
sv has joined #arm-netbook
sv has quit [Changing host]
<Turl> mripard: http://goo.gl/Xw5aw
discopig has quit [Ping timeout: 276 seconds]
<Turl> mripard: it works, but it may have extra stuff since it was non-dt-based back then
eFfeM has quit [Quit: Leaving.]
<mripard> Turl: could you send it by mail so that I can comment on it properly?
ChanServ has quit [*.net *.split]
<Turl> mripard: sure, sec.
ChanServ has joined #arm-netbook
<mripard> thanks
<Turl> mripard: on its way
tinti has quit [Quit: Leaving]
<mripard> yep, received it
rellla has quit [Remote host closed the connection]
revident has quit [Quit: Cache Invalidation... it's always a problem.]
<mripard> Turl: I'll try to reply to it this week, but I'm doing a training, so not exactly sure about when it will be.
<Turl> mripard: ok, no hurry :)
Jef91 has joined #arm-netbook
gimli has quit [Ping timeout: 252 seconds]
<Lofde_> does anyone have a solution or a way to make this work, i would of liked to used hamachi but they dont have an armhf build... anyone know if neorouter will work the same?
<RITRedbeard_> What problem?
RITRedbeard_ is now known as RITRedbeard
discopig has joined #arm-netbook
discopig has joined #arm-netbook
<Turl> Lofde_: neorouter doesn't support a arm build from what I can see
<Turl> Lofde_: you can probably achieve what you need with a properly configured openwrt though
<hramrach> Lofde_: also you can run armel binaries on armhf system
sv has quit [Ping timeout: 276 seconds]
<hramrach> Debian/Ubuntu have multiarch support so will allow you to install both armhf and armel libs
wingrime has quit [Ping timeout: 260 seconds]
ganbold__ has joined #arm-netbook
ZaEarl has joined #arm-netbook
<Lofde_> i tried to do a force install.. but they don't give me the binaries just the .deb's
<Lofde_> i was hearing i could achieve the same thing using openvpn and maybe vyatta .. but i am wondering if i want to go through all that to set this up
<Turl> sorry, I'm fail today
<Turl> not openwrt, OpenVPN
ganbold__ has quit [Remote host closed the connection]
<Lofde_> i am using the mk802 for an "advertising" business its going to be running ad's on a HDTV in different retail locations so i've got a lot of work to do with this thing
<hramrach> yeah, looking ata hamachi they will likely not have armel build either
<hramrach> and the software is junk anyway
<hramrach> Lofde_: what do you want it to do, really?
<Lofde_> since all these places they will be installed will behind NAT's etc.. i need the box to "reach out" and connect to the rest of the world so i can SSH into the box or VNC, etc
<hramrach> you can set it up to SSH into your server using a pre-configured key and do some port forwarding if you have only one-way access
<hramrach> or use that earlier mentioned OpenVPN
<Lofde_> i would rather not put pinholes in each locations router..
<hramrach> SSH does not need holes
<hramrach> OpenVPN probably neither
<hramrach> OpenVPN will allow you to see the boxes as if they were in your network
<hramrach> OpenSSH will only forward the ports you tell it to
<hramrach> eihter way you have to set up each box(stick) with unique IP (OpenVPN) or port number to forward (ssh)
<hramrach> the key is that you configure the stick to connect to your server on boot and reconnect when the connection breaks
Jef91 has quit [Remote host closed the connection]
<Lofde_> yeah thats the plan.. going to have to set up shellscripts to ping 8.8.8.8 every so often to make sure box is online and if not reset the wifi card and try again..
<Lofde_> going to try to also set up nagios to monitor the boxes in the field
<aesok> .
danlench1 has joined #arm-netbook
<danlench1> hi all, i need to talk to someone about hardware development for an FOSS project
<danlench1> we will be launching a kickstarter project soon. the software is ready but i would like to offer a hardware solution as well
<traeak> ouya? :-p
<danlench1> no
<danlench1> traeak: not a gaming platform
<traeak> you have to at least mention what the project is about
<danlench1> did you see my forum post over there today
<danlench1> think about an opensource free time capsule
<danlench1> with dropbox capabilities
<danlench1> server and desktop clients will be free
<leming> netatalk + owncloud? ;)
<danlench1> mobile clients will be about $5
<danlench1> no they are not free
<danlench1> nofolder.com
<danlench1> software has not been mabe public yet
<danlench1> we are still alpha
<danlench1> so...
<danlench1> hardware...
<danlench1> anyone care to chat
<danlench1> is this not http://rhombus-tech.net/
<danlench1> can i get a phone number of someone
<danlench1> all of a sudden everyone left?
<danlench1> leming: not just apple, win and linux clients as well
<danlench1> the server runs native on any amp stack
<danlench1> so windows can be a server
<libv> pasting blog entry to livejournal right now
<danlench1> different question,if i want to buy something who do i speak with
<aholler> the vendor?
<traeak> from here there's not much to buy
<danlench1> who is Rhombus Tech
<traeak> umm...lkcl is the one doing hardware here
<danlench1> so if i want 20k units i talk to lkcl
<specing> danlench1: Don't be so impatient. This is IRC and not some random tech support