#milkymist IRC log for Thursday, 2011-02-17

rohbrrr...00:01
roh'make it stop! -  burn it with fire!! - nuke it from space - ...'00:02
Fallenou:)00:05
kristianpaulFallenou: low troughtput no matter if you still able to transfer the whole thing01:21
kristianpauladamw_: hola :-)01:59
kristianpaulHow did you tests last time? (jtag-serial pod)01:59
adamw_http://en.qi-hardware.com/wiki/JTAG/Serial_Cable_run_1_for_Milkymist_One#Tests_and_Results02:00
adamw_some things wrong?02:00
kristianpaulno no, well let me check first :-)02:01
adamw_ok...i didn't use it to flash NOR via high speed though.:-)02:02
kristianpaullooks good in any case :-)02:03
adamw_must tell me if we missed important tests.:)02:04
kristianpaulhmm flash need a script, wiki is ok but you may need something faster to type, but for one board is okay02:05
adamw_u know me that I quite don't know the s/w details.02:05
adamw_but just for those automatics program used so i can know the 'h/w connection' is ok on jtag/serial pod.02:06
adamw_hmm ..yes but last time i did really know the high speed bug. man!02:06
adamw_didn't know.02:07
kristianpaulwhat? 8.2Mb flickernoise binary, i hope can still boot...02:12
kristianpauladamw_: will be nice compare high speed jta-serial pod with xilinx cable02:15
adamw_hmm...surely02:16
adamw_but i would focus on fixing boot bug first.:-)02:17
kristianpaulyes yes02:17
kristianpaulok, no i cant belive from 1.4Mb with pdf support to get pdf suppport i got now a 8.2Mb bin !!02:21
kristianpaullekernel: ^02:21
kristianpaulE: Invalid flash boot image length02:22
kristianpaulhmm02:22
kristianpauldammit, i cant boot...02:30
kristianpaulah weird, is just flicernoise??.. the httpd network demo is oka..02:36
kristianpaulFallenou: Where are you using MAXIMUM_FRAME_SIZE?? and how are you _handling_ frames larger than that in order to reject those02:51
kristianpaulFallenou: i'm just reading page 48 of netwotking.pdf about Stress Tests02:53
kristianpaulETHERNET_FRAME_LENGTH  = MAXIMUM_FRAME_SIZE?02:55
kristianpaul // TODO: Use a softc or global variables, but not both!!!...02:58
kristianpaulare you using softc right now?02:58
kristianpaulah yes02:59
kristianpaulnow i ask my self how well this driver follow rtems guidelines...03:08
TechnicusHello . . . perhaps a little off topic, but are there any instructions avaliable for connecting a Cannon video camera via usb to a Debian system as a video capture device?05:21
adamw_lekernel, while starting a Program_B(configuration) got LOW, shall RP# be activated as LOW from fpga h/w itself or your codes too?07:30
Fallenoukristianpaul: this driver surely is not a model , it is by several aspects wrong09:53
Fallenouanyway we should not receive any frame larger than the usual MTU09:54
FallenouI guess no network card would send bogus packets in normal configuration09:55
lekernelkristianpaul: yeah, the binary is now 8MB with fonts etc., and yes, it still works10:44
lekernelwe have 128MB RAM anyway...10:44
lekerneladamw_: the FPGA does not reset the flash when it is reset10:44
lekerneli.e. pulsing PROGRAM_B low does not make it assert RP# low10:45
lekernelbut normally people don't access PROGRAM_B, so I don't think we should add circuitry to handle this case10:46
adamw_so I'd like the time FPGA asserts in reset(configuration started) syncronized with flash in parallel.11:07
lekernelwell yeah, you could simply add a diode between PROGRAM_B and RP# and put a reset IC on PROGRAM_B11:08
adamw_yes, but seems that needed to add two diodes:11:10
adamw_one is located between PROGRAM_B and reset ic's out11:11
adamw_um...no11:11
adamw_i described wrong. sorry11:11
adamw_should be:11:11
adamw_i nyour last email replied me on list , you mentioned two diodes?11:14
adamw_but now it seems that we can only just add 1 diode. seems it's enough.11:15
adamw_right?11:15
lekernelyeah, it's enough11:26
lekernelthe one diode solution is better11:27
adamw_yeah~seems it's ok.11:33
lekernelhttp://www.stanford.edu/class/ee183/handouts_win2003/VerilogQuickRef.pdf what a mess :)11:34
adamw_btw, i checked the p.36 of http://www.numonyx.com/Documents/Datasheets/319942_J3-65nm_256-Mbit_MLC%20DS.pdf11:35
adamw_the P3 needs 300us MIN.11:36
adamw_this seems that my reset ic's delay time at least 300us is a must?11:37
lekernelyeah, but I read that the FPGA needed 5ms, so it's a bit surprising we ran into this issue11:37
lekernelanyway11:38
lekernelyeah, your reset IC should pulse low for at least 300usec11:38
lekernelbut i'd put more (like 10ms) to be sure11:38
adamw_yeah...i think so11:39
lekernelor maybe 50ms11:39
lekernelso everything comes out of reset when the power supply is well stabilized11:39
adamw_right. currently i used 20us delay time only. It's not enough.11:39
adamw_right..11:39
lekernelyou could even put 1s delay there :) it's only for the first power up...11:39
lekernelno need to optimize this one11:40
lekernelso let's take very large margins and be safe11:40
adamw_correct. end user won't feel it's a long time.11:40
adamw_but should solve our boot tiny period (power cycling) problem.11:41
adamw_ok...so I'll order a sample with more than 300us and try later.11:41
wolfspraullekernel: I told you we did pretty thorough/extreme testing.11:41
lekerneladamw_: take a few dozen milliseconds11:41
adamw_hope we are all right on analizing at the beginning.11:41
wolfspraulbut now that we see this and find a way to make it more robust - great.11:42
adamw_lekernel, I'll..tks.11:42
lekernelaeris: http://golang.org/doc/gccgo_install.html12:36
lekerneltry that :)12:36
lekernelhttp://www.rtems.com/wiki/index.php/GCCGoRTEMS12:37
aerisAutant faire du Java...12:39
cocoadaemonbwerk12:40
cocoadaemon;)12:40
cocoadaemonola lekernel aeris12:41
cocoadaemondis donc, ça se peuple ici ;)12:41
cocoadaemongood job lekernel12:41
lekernelc'est quoi le problème avec google go?12:41
lekernely0 cocoadaemon12:41
cocoadaemonj'ai rien contre go, je disais bwerk pour "autant faire du java" ;)12:42
aerisça semble mignon, comme language, mais t'en qu'a prendre un language évoluer, autant prendre Java, il est connu et pleins de lib12:42
cocoadaemonmais, aeris, c'est un Projet® Google© !12:43
cocoadaemon</troll>12:43
lekerneljava c'est lent + besoin de JVM12:43
lekernelgo c'est compilé nativement et moins lent que java12:44
aerisça se compile le Java12:44
lekernelc'est pas fait pour...12:44
lekernelgo works nicely with RTEMS too... java does not12:48
aerisJe pense que ce n'est pas une bonne idée de partir sur des techno inconue...12:49
lekernelmoi je pense que si :)12:50
lekernelgo semble bien plus propre que c++12:50
lekernelet il n'y a pas de raison qu'on n'arrive pas à le faire fonctionner sur milkymist12:51
aerisMais personne ne pourras bosser dessus12:51
lekernelhttp://en.wikipedia.org/wiki/Google_Go#Examples12:51
lekernelça semble assez clair comme syntaxe12:51
Fallenouruby seems really nice and is really becoming a good language12:53
Fallenouwith a lot of usefull and well done librairies (gem)12:54
lekernelruby is totally slow12:54
lekerneland I don't know if a sizable proportions of those libs would work at all, and if they do, without taking one second to draw one pixel12:55
lekernelgo seems a good compromise12:56
lekernelfast, high level, and shouldn't be too much hassle to get it to work12:56
Fallenouok :)12:56
lekernelhttp://www.engineyard.com/blog/2009/ready-set-go/13:01
kristianpaulgo go..16:08
kristianpaulgo + mtk ;-)16:08
lekernelfunny: xst does not use carry chains for adders with operands of strictly less than 6 bits18:16
lekernelhttp://lwn.net/SubscriberLink/428100/9d3311d20cd0ea12/21:02
lekernelhrm21:02
Fallenouhum sounds like a strange technical choice21:10
sh4rm4why was rtems chosen over linux ?21:19
sh4rm4btw: congrats to MTK, lekernel. i've never seen such skillful code removal before21:21
sh4rm4it looks as if you ripped 50% of the code off21:21
lekernelbecause linux was unstable, and is slow to boot, complex and it takes 10 times more times to write a linux device driver than a rtems one21:22
sh4rm4hmm...so RTEMS is POSIX compliant ? i wonder how hard it would be to port stuff there if it wasnt21:23
lekernelyes it is21:24
sh4rm4oic, nice21:24
lekernelmaybe we'll switch to linux, when/if it does not introduce regressions compared to rtems21:25
sh4rm4i heard some rumours about openwrt a while ago21:25
sh4rm4(here)21:25
lekernelthe initial plan was to make it run linux and openwrt21:25
lekernelbut it turned out to be a bloodbath21:26
lekernelgetting that mess to work far exceeded my limited patience21:26
sh4rm4hehe21:27
lekernel(and time)21:27
sh4rm4well, if rtems runs nice, why would you considering switching back to linux ?21:27
lekernelmore features21:27
lekernellike bluetooth, wifi, etc.21:27
sh4rm4i c21:27
lekerneland geeks like it better21:28
sh4rm4how much work would it be to adopt MTK for linux usage ? (i.e. on fbdev)21:28
lekernelit's trivial21:28
lekernelall the mtk code is platform independent21:29
lekernelit writes to a framebuffer, and takes events in21:29
lekerneland calls application callback functions21:29
lekernelthat's it21:29
sh4rm4nice21:29
mwalleis {git,www}.qemu.org working for anyone?21:48
sh4rm4http://www.downforeveryoneorjustme.com/git.qemu.org21:49
larscwell, i can't reach it either21:50
lekernelworks for me21:53
lekernellol, it never ceases to amaze me that xilinx fpga editor uses sunrpc (don't know what for) and takes ages to start up with RPC set-up errors if portmap isn't running on your machine21:54
lekernelah, maybe they use that for IPC with PlanAhead21:54
lekernelbut it's still a silly use of RPC21:54
mwallelekernel: could you check if the the HEAD is at e14da0af640e4255b15d81907a93a2637e14e478 (Fix vmport segfault (v2)) ?21:56
lekernelyup, it is21:57
mwallethx21:57
--- Fri Feb 18 201100:00

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