#qi-hardware IRC log for Friday, 2010-12-03

kyakviric: (benchmark) don't know.. maybe need to ask in mailing lists, someone should remember02:47
virickyak: do you know if openwrt brings the SIMD mplayer?03:12
kyakdon't see it in packages03:22
qi-bot[commit] kyak: config.debug: the minimal .config with debugging capabilities http://qi-hw.com/p/openwrt-xburst/1c5d30903:32
viriclarsc: does oprofile work fine on the nanonote?03:37
virickyak: ok. I wonder if someone is building it nowadays.03:37
viricI even don't know what parts there were 'accelerated'03:38
kyakare you talking about SIMD player?03:40
kyaki don't know what SIMD is :)03:40
wolfspraulkyak: config.debug could also include logging daemons set to 'debug' logging03:40
kyakmplayer is ported, however03:40
kyakwolfspraul: good idea! however, now we have those daemons disabled in /etc/init.d/boot :)03:41
kyakwolfspraul: it's just some things depend on things other than .config.. maybe their own config files03:42
kyakso maybe let the user who is into debugging make this decisions by his own?03:42
virickyak: Single Instruction Multiple Data. MXU instructions, on xburst, I think.03:55
kyakviric: have no idea.. openwrt mplayer is not patched, so it might not take advantage of xburst specifics03:58
kyaki've seen patched maplyer from ingenic03:58
virickyak:03:58
kyakbut it's based on some older version03:58
viricYes, I talk about that03:58
viricSo there is noone in the nanonote using a 'fast' mplayer? Also jlime uses only the upstream?03:59
kyaki'm not suffering of some lack of performance, to tell the truth :)03:59
kyakdon't know about jlime04:00
kyakand mplayer in openwrt is just a standart mplayer04:00
viriccan you play theora videos?04:00
kyakas it goes from mplayer developers04:00
kyaknot tried playing theora videos04:01
kyakthereis a weird bug though04:01
kyakboth in openwrt and jlime04:01
kyakwhen you fast forward several times in a row, thereis a sqeeqy sound from speaker04:02
kyakand it won't stop04:02
kyakuntil you reboot04:02
kyakperhaps you will find it yourself04:02
viricoh04:03
viricIn fact, the mplayer I built had a very bad performance.04:03
kyakit depends on used codecs a lot04:04
kyakwhich do you use?04:04
kyakand the video resolution, of course04:04
kyaksaying simplier, what;s your command line?04:05
kyakto encode and to play videos04:05
viricMaybe I chose it to use floating point, and I did not notice.04:05
kyakhttp://projects.qi-hardware.com/index.php/p/openwrt-packages/source/tree/master/mplayer/README04:05
kyakthis approch works pretty well here04:06
viricah, I took an h264 video at 640x480 :)04:06
virickyak: I'll check. thank you"04:06
kyakoh, good for you then04:06
virichehe04:06
kyakeven my laptop is not capable playing such videos :)04:06
viricwhat codecs you advise?04:06
kyakdoes your kernel has floating point emulation?04:07
viriccan it do xvid?04:07
viricEven if it has, I don't want to use it04:07
viricI don't remember if I disabled it finally04:07
kyaki think xvid should be fine04:07
kyaki convert it to mpeg to watch on ben04:08
viricmpeg2?04:08
kyakvcodec=mpeg1video04:08
kyakjust have a look at commnd line i shown you04:08
kyakmplayer has some floating point computations04:09
kyaki've seen it fixed in ingenic's patch04:09
virickernel emulation would be very bad...04:09
viricMy toolchain builds always for soft-float, so it should never end up using the kernel emulation04:09
kyakhowever, i don't think this causes some real overhead04:09
kyakbtw, you can try playing your video with -nosound04:10
kyakand then with -vo null04:11
viricand for sound, what do you use?04:11
kyakthen you can see if it is video or audio that's causing this performance droppage04:11
viricfor the audio track04:11
kyakviric: haven't you had a look at the link i gave you? ;)04:11
viriclink?04:12
viricAH04:12
kyak12:03:54 < kyak> http://projects.qi-hardware.com/index.php/p/openwrt-packages/source/tree/master/mplayer/README04:12
viricwill read.04:12
viric:)04:12
kyakhowever, i then switched to that:04:13
viricmp2 audio, mpeg1 video? now I understand why you don't experience performance problems with the upstream mplayer :)04:14
viricbtw, that mplayer command line is to me like a 'blob' is to free software ;)04:14
kyakcan't find the command line, switched to liblame04:15
kyakviric: i don't really care if it's "free" or "not free" when it comes to my own convenience04:16
virichehe04:16
kyakbtw, that's why mplayer depends on BUILD_PATENTED04:16
kyakit's in my todo list to split it to free and patented :)04:17
virichave you tried the vp8?04:17
viricThey have specific assembly for arm, at least04:17
kyakno, what's that?04:17
viricI don't know if for mips04:17
viricthe video codec patent free for webm04:17
viricbsd licensed04:17
kyakis it supported by maplyer?04:18
kyakviric: sorry, got to go now.. thanks for you hint. might want to have a look later...04:19
virickyak: vlc plays it fine :) and ffmpeg can use it fine04:20
virickyak: I don't remember playing it in mplayer though04:20
viricmaybe vlc would work better than mplayer04:26
bartbeshmm08:00
bartbesis the entirety of gtk/glib running on the ben?08:00
kristianpaulhmm08:00
kristianpaulin wich os?08:01
bartbesowrt08:01
kristianpauli think you can compile owrt with gtk support too08:01
kristianpaulit was a blog entry from mirko i think08:01
kristianpaullong time ago08:01
bartbesI know there's gtk support08:02
bartbesone of the default apps uses gtk08:02
bartbesstardict iirc08:02
bartbesyeah, looks gtk-y08:02
bartbesI guess I could just try08:06
bartbesbut it'd be crappy if it fails08:06
bartbes(as I'm going to write a program)08:06
kristianpaulin gtk?08:11
bartbesyes08:23
kristianpaulwht not SDL? :)08:24
bartbesbecause.. I'm going for a GUI, not anything needing real graphics08:27
kristianpaulah ok08:28
kyakwow.. a minimal system with debug and without stripping is > 700 Mb12:26
kyakto the hell with that debug idea12:27
kyakoh, maybe this is when qemu is coming handy12:27
wolfspraulkyak: 700 mb image?12:52
wolfspraulmaybe still makes sense via the ImageBuilder, so people don't have to rebuild it all, just add a package they want to debug...12:52
kyakmy bad, this image has stardict.. pretty many dependencies13:12
virickyak: 700mb image? it's not that bad14:06
kyaktrying to do some debuggin inside the qemu..14:08
kyakgdb says utils.c:904: internal-error: virtual memory exhausted: can't allocate 13263733 bytes.14:09
wpwrak_that's things like libsomething-dbg ?14:09
virickyak: give memory to qemu14:09
kyakof course there is enough memory14:09
viricah :)14:09
kyakso it's strange..14:09
viricdoesn't anyone here have a more powerful mips than the nanonote?14:10
viricI think I could run the nanonote programs in a chroot in a Fuloong MiniPC14:10
kyakyou? :)14:10
viric:)14:10
viricspeed is not the problem :)14:10
kyaki'll try to debug something else14:11
kyakmaybe stardict is too damn tricky14:11
virickyak: try gdbserver, and debug on the PC14:11
kyaka very good idea14:11
kyakshould i run gdbserver on my PC?14:11
viricno14:11
viricgdb in your PC, gdbserver in the nanonote14:12
viricor qemu14:12
kyakah, ok.. will have to read about that14:12
viricyou need a gdb that can understand mips though14:12
virica 'cross-gdb'14:12
kyakyes, such is avaialble14:12
kyakin a toolchain14:12
viricyou can use the network14:12
kyakStarting program: /bin/ls14:13
kyakvfork: Cannot allocate memory.14:13
viricyou start a gdbserver in one side... and in the cross-gdb in the pc you run "target remote xx.xx.xx.xx:1234" whatever ip:port there.14:13
kyakit's really-really strange14:13
viricmaybe that gdb is not very well built14:13
kyakgdb is not able to debug even simpliest apps14:13
viricI'll try with mine...14:13
kyakyes, maybe14:13
kyakwill try with gdbserver14:14
rafaviric: I have ran gdb in nanonote without problems14:17
viricrafa: good to know! :)14:55
lunavoraxHi everyone !15:50
lunavoraxI wanted to ask a software question about the Ben Nanonote15:50
bartbeslunavorax: ask away15:51
bartbeswe won't know if anyone can answer it before it has been posed15:51
lunavoraxWhat display server (or windows manager, I have troubles figuring out the good word) is used on the Ben Nanonote instead of X.org (or is it X11 ? can't figure out)15:51
bartbesnone15:51
bartbesthe openwrt kernel uses direct framebuffer access15:51
lunavoraxSo how does the display work ?15:51
lunavoraxOk15:51
bartbesjlime has an X server15:52
lunavoraxopenwrt kernel = regular gnu/linux kernel right ?15:52
bartbesthe one that comes preinstalled and the official qi-hardware one, yes15:52
bartbeswhy do you ask, if I may?15:52
lunavoraxWell I was just wondering how it was working15:53
lunavoraxI started wondering with something unrealated in fact15:53
lunavoraxThe Wayland project15:53
bartbeswell, good thing a lot of toolkits have fb support15:53
lunavoraxYeah15:53
bartbesoh heh, I doubt we'll see the needed hardware acceleration15:54
lunavoraxAlso I saw a guy made one here "because X11 sucks" on github15:54
lunavoraxNo no I wasn't talking about that15:54
lunavoraxIt's just that it started my question about how all this stuff was working15:54
lunavoraxAlso I already knew that X11 was slow and I was "at last a descent replacment"15:55
bartbeswell display server architecture is an interesting topic15:56
bartbesand I love it how they said they weren't going to care about network transparency so much, and then specifically the shitstorm of scared users saying "but I need networked displays!"15:56
bartbesnot realizing that the wayland guys still want that possibility15:57
lunavorax(also I was wondering when I saw that some libraries like Allegro could make softwares that doesn't need X to run15:57
bartbesah well15:57
bartbeswell, afaik there are only a few real options for rendering15:58
lunavoraxYeah I actually didn't understood what network transparency (in fact all the network stuff on a local display server concept)15:58
bartbesframebuffers are quite direct15:58
bartbesthere's X15:58
bartbesthere's OpenGL15:58
bartbesand then now there's wayland, but afaik it runs on top of gl15:58
lunavoraxYeah15:59
lunavoraxYou think it's a default ?15:59
bartbeswell, you see, X' network transparency is kind of weird15:59
bartbesyou see15:59
bartbesinstead of using direct calls15:59
bartbesor even unix sockets15:59
bartbesor named pipes15:59
bartbesyou use full-blown network sockets15:59
bartbes*always*16:00
bartbeswhich is inefficient16:00
lunavoraxOh wait OpenGL16:00
lunavoraxDoes that means that when I use Compiz, it doesn't use X anymore ?16:00
wpwrak_bartbes: (always) unless you use SHM16:00
bartbesshm?16:01
bartbeslunavorax: it does use X16:01
bartbes;)16:01
lunavoraxHaha ok16:01
bartbesbut iirc the X server needs GL extensions16:01
viricbartbes: X uses unix sockets locally16:02
bartbesoh right16:02
bartbesbut still, sockets16:02
wpwrak_bartbes: shared memory. x has a number of ways for using more efficient paths when available.16:02
bartbesI mean, I didn't spend too much time looking into this stuff16:02
viricbartbes: you have a strong opinion though ;)16:02
bartbesI just read tiny bits of info spread over time, then combined that16:02
wpwrak_viric: the less you know, the easier it is to hold a strong belief ;-)16:03
bartbeswpwrak_: that is not true16:03
bartbesI mean, I said I read tiny bits16:03
bartbesbut I read a lot of them16:03
bartbes:P16:04
bartbesI mean, in all this time I have absorbed quite a bit of information16:04
bartbesand it helps me not be a sheep16:04
viric:)16:04
bartbesit's one of the reasons I'm not shouting "I need  network transparency, X rules, wayland sucks, and canonical sucks"16:04
viricIt feels comfortable that, most software we have today is built on X instead of something else that may not be easily used through a network16:05
viricMaybe it could be better, and new software is going to show that... but luckily we don't have something worse :)16:06
bartbeshmm16:06
bartbesthe thing is16:06
bartbesit makes no diff in the programs themselves16:06
bartbesit's probably a small diff in the toolkit (for the networking, of course X support is quite a bit more complicated)16:06
wpwrak_bartbes: yeah, my comment was a little nasty, i admit :)16:06
bartbesand that is the goal of the transparency16:06
bartbesa program doesn't need to have a clue16:07
viricwell, a program for X does not have to have a lot of clue.16:07
viricIt could be better, sure.16:07
bartbestbh there are lots of programs that don't have a clue16:08
bartbessame goes for programmers16:08
bartbesbut that is something entirely unrelated :P16:08
viric:)16:08
viricI never wrote a program dealing with the X libs directly...16:08
viricThe kind of abstraction you may want is that given by Qt or similar. Isn't it?16:08
wpwrak_yup. or Gtk, SDL, ...16:11
wpwrak_they may draw some concepts from X, e.g., color allocation, but they wrap them into their own style and mechanisms16:11
bartbesSDL is very nice16:12
bartbesit can do fb, X, svgalib and more16:13
bartbestransparent to the program16:13
bartbesanyway, I did some xlib16:14
bartbesnot to put something on the screen, but I needed the Xlib features16:14
viricisn't xcb any better?16:14
bartbesno idea16:14
bartbesI just know it had the direct access I needed16:14
rafaI have been tempted to write a new GUI for freerunner using just C + fb and sdl, because after some tests it is the fatest thing I have seen :).. but I dream with a lot of things I will be doing.. And then, at the end of the year, I realize that I just did two or three things of that 200 items list I often dream ;-))16:15
viricX is very bad at caching application buffers though16:15
wpwrak_rafa: (3/200) a common problem ;)16:17
bartbesviric: caching?16:18
bartbesyou mean what wayland is trying to solve (amongst other problems) by letting the application tell the server what to redraw when?16:19
viricbartbes: so any redraw needed in the server goes to the client.16:19
rafawpwrak_: it gets worst because the 200 items number increases a lot every year.. and the items done does not so fast than total items todo :D16:19
viricbartbes: I'm not following wayland16:19
wpwrak_rafa: ah, exponential growth. now that is bad :)16:19
bartbeswpwrak_: you should clear the list sometime ;)16:20
wpwrak_viric: are redraws really that big a problem ? and X11 should still have server-side backing store. http://www.ess.co.at/FAQ/faq001.html16:21
viricserver-side backing store? I never saw it16:22
viricohh nice option16:22
viricwpwrak_: but that does not work as cache, right? I mean that the option will change what the user sees, but not the network usage.16:23
wpwrak_as far as i understand it, it does work as a sever-side cache16:26
wpwrak_http://www.x.org/archive/X11R6.8.1/doc/ati5.html section 5.1516:28
viricwhy isn't "backingstore" a default?16:35
viricit may have a counterpart other than 'it takes more memory'16:35
wpwrak_if you're local and the content isn't particularly hard to recreate, redraws may be good enough. in fact, i think redraws may be relatively infrequent anyway.16:39
viricthey happen on window resizing, or a window getting in front of another16:40
wpwrak_i think the tendency nowadays is a "redraw from model" pipe anyway16:40
wpwrak_yes, and how often do you resize/restack windows, compared to other operations ? :)16:41
wpwrak_restacking could be a problem if don't have enough screen space, though.16:41
wpwrak_ah, and when you resize, you probably want to redraw anyway16:42
viricwpwrak_: I resize often, using a tiling window manager :)16:43
wpwrak_particularly if you follow the "redraw from model" paradigm16:43
wpwrak_urgh :)16:43
viricwpwrak_: also causes a redraw the switching of 'desktops'16:57
viricor however you call it.16:58
viricFull screens of organized windows.16:58
wpwrak_yeah, switching desktops is something i do a lot, too17:00
viricdo you have that option switched on, for caching?17:01
wpwrak_ah, the other one is save-under. e.g., when you pop up a window on top of another, the occluded window gets saved17:01
wpwrak_no, neither of them17:02
wpwrak_i'm not sure they would accelerate what i'm doing much. most things are fast anyway ;-)17:03
viric:)17:13
viricyou helped me a lot letting me know the options are there.17:13
viricWhen I'll experience X11 network speed troubles, I may remember them17:13
virickyak: gdb works here, I think17:54
kristianpaul(the less you know, the easier it is to hold a strong belief) That works almost everywhere17:57
kristianpaulgit can just tell me commit history for my current folder?18:55
kristianpaulsie code generator?..18:56
kristianpaulwill be nice if carlos camargo once at week join irc and tell us what he is doing18:56
kristianpaulor at least ping he here18:57
kristianpaulthats the top !!18:58
kristianpaulWhat on earth is doing a  deb package there  http://projects.qi-hardware.com/index.php/p/nn-usb-fpga/source/tree/master/Examples/UNFURO/avrdude_5.10-1_i386.deb18:59
kristianpaul:S18:59
kristianpaulwhat a mess :18:59
wpwrak_;-)19:03
kristianpauloh something is wrong with my local git, i'm tryin to commit to qi and is pushing to github? O_o19:09
kristianpaulgit is self hosted config isnt?19:14
kristianpaulcause i reinstalled debian some weeks ago and dint backup nothing about git config (in general)19:15
kristianpaulfatal: Path 'tools/adc/sofware/jz47xx_gpio.c' is in submodule 'tools/adc/sofware'19:18
kristianpaulwhat that means "speaking git" ?19:18
qi-bot[commit] Cristian Paul PeƱaranda Rojas: created notes folder http://qi-hw.com/p/ben-gps-sdr/1c2e0de19:22
wpwrak_kristianpaul: there's a ~/.gitconfig, e.g., with your name and email19:29
kristianpauldammit19:29
wpwrak_not sure about the other things19:29
wpwrak_(gitconfig) it's simple: [user]\n\tname = Your Name\n\temail = you@mail.foo\n19:30
wpwrak_that's it :)19:30
kristianpaulyes19:30
kristianpaulno i said dammit cause i still getting "fatal: Path..."19:30
kristianpauland i just cloned the repo again and copied files in side19:31
wpwrak_ah ... that sounds like more fun :) google ?19:31
kristianpaulyes19:31
qi-bot[commit] Cristian Paul PeƱaranda Rojas: missing simple(stupid) code to initialize and read adc buffer http://qi-hw.com/p/ben-gps-sdr/351e1d519:35
kristianpaulsolved19:35
wpwrak_what was it ?19:38
kristianpauldont know i think related to cache19:41
kristianpauli just copied and pasted some commands19:41
wpwrak_;-))19:42
kristianpaulbtw about my problem with 8Mhz signaling i solved it, well was for sie initially but can be re-used in the ben19:42
kristianpaula SPI  to Parallel converter19:42
kristianpaulis easy cause the signal is just 4 bit per sync19:43
kristianpaulson ben actually can handle  2Mh IO19:44
kristianpaulonce it work on SIE i may try fit it on CPLD or just wire sie to ben19:44
kristianpauli was thinking to implemented a 2k buffer on FPGA, as the original adc in SIE (wich is SPI too) works19:45
kristianpaulWe need  a map with people timezones :)19:51
kristianpaulwpwrak_: you remenber sdram speed of ben?20:21
kristianpaultroughput**20:21
wpwrak_hmm no, don't remember. it ought to be somewhere in the archive ;-)21:10
wpwrak_(spi-par converter) congratulations !21:11
kristianpaulnah not yet21:31
--- Sat Dec 4 201000:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!