#milkymist IRC log for Sunday, 2012-01-29

Fallenou25 minutes to generate official Milkymist SoC bitstream using a debian virtualized in virtualbox on my mac book pro Core 2 Duo 2.4 GHz00:35
Fallenoulekernel_: 25 min to generate bitstream, does it sound a lot to you ?00:35
wpwraksounds like you have a fast machine :)00:36
wpwrakit also takes ~25 min on my Q6600, native00:37
Fallenouok so i'm not in such a bad position then00:39
wpwraknaw, you're suffering just like the rest of us :)00:40
FallenouI can't blame my computer if mmy design does not go as fast as I expect :)00:40
wpwrakhehe :)00:40
FallenouI confirm timings are met using official milkymist git branch00:41
FallenouI must have done something wrong00:41
Fallenouwill have a look00:41
Fallenouwpwrak: do you know if lekernel_ is using xst or synplify to generate nowadays Milkymist SoC bitstreams ?00:47
Fallenouit seems Xst is the default ATM in the makefile00:47
wpwraki would think he uses the makefile, yes. but i don't know for sure00:48
FallenouI've heard about something like Xst generating errors in lm32 caches that got solved using synplify but don't know if it's still the case with recent ISE milestones00:49
FallenouOK latest Milkymist SoC seems to meet timings with ISE 13.3 according to http://www.milkymist.org/wiki/index.php?title=Recommended_ISE_versions00:50
FallenouI have 13.200:50
Fallenougn8 !01:11
wolfspraulI downloaded 13.4 recently, should give it a try...01:14
Fallenouwill try with it tomorrow, I need to backup 13.2 install (13 GB) before01:16
Fallenouin case I have to go back to it01:16
Fallenoudon't have enough space to have both installed on my virtual machine :)01:16
wolfspraulI think lekernel_ is on 13.3 (that's what the wiki page says)01:23
wolfspraulthe latest now is 13.401:23
Action: kristianpaul too01:23
wolfspraulif you do that you may be the first one with 13.401:24
wolfsprauljus tfyi01:24
rohwolfspraul: do you have contact to the people doing this: http://www.elinux.org/Embedded_Open_Modular_Architecture ?01:44
kristianpaulroh: the guys that will beat the 39usd pi compuiter?01:48
kristianpaulirc channerl is decert..01:48
rohand since it seems to be a sincere grassroots attemt i rather check out that than the pi01:49
kristianpaulbut the guy already wrote to mail list01:49
kristianpaulyou rememnber mail in #qi-hardware ML about A10 stuff and fped01:49
rohr-pi means no specs, no schems, no datasheets, and only partial opensource drivers from what we know now01:49
rohkristianpaul: thats why i am asking.. if there is talking happening ;)01:50
kristianpaulwell, by talking the guy is very quiet nevr reply even the mail thread..01:50
kristianpaulguess to busy hacking its coming boards to reply mails :)01:50
kristianpaulalso i and considering it is milkymist channel, i will care more of having a working ethernet driver for uclinux milkymist port01:56
kristianpauland a _VERY_ rought starting point for a freedom box with that :)01:56
kristianpaulfreedom hacker box ;)01:57
wpwraklekernel_: "regf" (in various comments in compiler.h) == "register file" ?05:35
stekernwpwrak: (u-boot) why would you need locking in a program that is singel threaded and (hardly) doesn't use interrupts?08:20
stekernI kind of agree with the swiss army argument though ;)08:21
stekernI've devoted some time into the openrisc u-boot port, mostly out of PR reasons08:22
stekernit's a well known project, and I believe in wolfspraul's comment about "making people feel safe"08:22
lekernel_wpwrak: yes09:28
Fallenougood news, timings are met with 13.4 but I did not test to flash and boot the generated bitstream09:47
Fallenoudamn, git head of urjtag does not seemt to compile (under debian squeeze)11:23
Fallenoumisses libftd2xx.h11:25
wolfspraulI just recently built it, let me see what I installed11:27
Fallenouwill try this11:28
wolfspraulapt-get install dh-autoreconf autoconf libtool pkg-config bison flex python2.6-dev libftdi-dev libusb-1.0-0-dev11:29
wolfspraulalso apt-get build-dep urjtag11:29
wolfspraultry those11:29
Fallenouok thanks11:30
wolfspraulrun ./autogen.sh --with-libusb --with-libftdi11:34
Fallenouwo it seems ok now11:35
Fallenouoh no, same problem11:36
larscconfig.log says libftdi was found?11:37
Fallenouyes libraries are found (libusb yes, libftdi yes (no async mode), libftd2xx yes)11:39
Fallenoubut the header is missing anyway11:39
Fallenouwill copy it manually from ftdi website11:40
wolfspraulyou can search it with apt-file search libftd2xx.h11:40
wolfspraulbtw I think you don't need this 2xx stuff11:40
wolfspraulapt-file cannot find libftd2xx.h11:42
wolfspraulI'm running my ./autoconf.sh again, and libftd2xx is :no11:42
Fallenoueven with the header from official website it fails11:44
Fallenouoh ok I will disable it then11:44
stekernyou need to build explicitly without the ftd2xx stuff11:44
Fallenouyep I'm trying11:45
larscwolfspraul: libftd2xx is the thing from the FTDI themself11:45
stekernI had problems to when those got pulled in11:45
Fallenouunrecognized option "--without-libftd2xx"11:45
Fallenouok just --without-ftd2xx11:47
Action: Fallenou needs to sleep more11:47
Fallenouwill update the wiki to add --without-ftd2xx11:48
Fallenoubut after lunch11:49
wpwrakstekern: it tends to gets multithreaded and interrupt-driven as soon as you add USB ;-)12:23
stekernwpwrak: oh, you are speaking about the usb-implementation... I feel your pain ;)12:28
stekernbut as long as your driver doesn't do interrupts, it should still be pretty much just polling12:30
wpwrakyeah, as long as you don't use u-boot for much, it's pretty okay. like the plastic swiss army knife. as long as you don't open it, it'll be in good shape forever ;-)12:35
stekernyou shouldn't be backtalking the swiss army knives though, my 20 year old trusty victorinox still has its corkscrew in excellent shape! (and it has been opened on several occassions)12:41
wpwraki'm talking more of the malaysia-made ones ;-)13:02
stekernheh, ok13:04
stekernanyways, in my openrisc-m1soc merging adventure I've taken the path of porting BIOS to openrisc rather than porting the m1soc to u-boot13:09
wpwrakgood :)13:10
stekernI deemed that to be easier, and I'm lazy ;)13:11
wpwraklaziness, the engineer's #1 virtue13:11
stekernI agree13:12
Fallenoudo I need svf or bsdl support in my urjtag ?15:38
Fallenoucause none of both compiles15:38
kristianpaulsvf is not just for cpld?16:00
Fallenouok got urjtag to work and flashed the bitstream successfully16:15
wolfspraulFallenou: cool! it seems you are making massive progress!16:35
Fallenouyep !16:37
FallenouI'm getting good to go :)16:37
Action: Fallenou heating up16:37
Fallenouwiki updated17:02
Fallenouftdi chip is hotter than fpga17:31
lekernelFallenou: you shouldn't need to flash your bitstream, you can load it directly with pld load xxx.bit17:47
Fallenouby flashing I meant loading it17:48
FallenouI just did 'make load-bitstream'17:48
Fallenoummu branch meets timing with ISE 13.4 :o good17:48
Fallenounow need to write FSM for TLB management (check/miss/update/flush)17:49
lekernelok, just make sure that all hits take only 1 cycle17:50
FallenouI hope tlb miss won't happen too often17:52
Fallenoubecause it would generate exception, kernel would have to lookup pfn using memory loads and then update tlb line and then return17:54
Fallenoulekernel: would an update to a ITLB line update the DTLB line as well ?18:16
Fallenouor would we just want to be able to update independently each TLB18:16
FallenouI think we could gain having two different contents in those two TLBs18:17
Fallenouso updating ITLB and DTLB independently18:17
lekernelboth I and D caches and TLBs operate independently18:17
Fallenoujust wanted to be sure :)18:17
kristianpaulah m1nor need to rewritten18:22
kristianpaulto m1load18:22
kristianpaullekernel: sorry for mistype words but can you please stop writing private mails everytime you get angry by reading others mistakes!18:35
kristianpaulsorry OT18:36
Fallenouwell, maybe it's better keeping it private rather than sending corrections each time on the public list, isn't it ?18:38
kristianpaulsorry all18:41
Action: kristianpaul checks how to use spell checker inside mutt18:45
Fallenouyep spell check is helpfull sometime :)18:46
Fallenou-l ;-)18:46
Action: larsc needs a spell checker for irssi18:49
kristianpaullarsc: ha mee too ;)18:50
FallenouIt will be fun to test if MMU is working properly19:27
Fallenoueither modifying RTEMS, or Linux .... or a custom bios19:28
Fallenoumaybe a custom bios19:28
Fallenouis it possible to reboot into bios ?19:29
Fallenouholding a button or something19:29
Fallenouinstead of booting to rtems shell/FNoise19:29
lekernelyes, press esc19:29
Fallenouoh, I need a keyboard :19:32
larscpress all three buttons19:32
Fallenouit reboots into rtems19:33
lekernelit works with the serial console, too19:34
lekernelwpwrak: by the way, how do you plan to implement "hard" reset if there's only one button?19:36
kristianpaultime pressed?19:36
larschavn't you seen his statemachine for the button behaviour?19:37
kristianpaultrying to find it..19:37
Fallenouif I boot pressing escape in the uart it does not boot19:37
kristianpaullaunch flterm first19:38
lekernelFallenou: which is what you want, no?19:38
kristianpaulah, you mean boot from serial?19:38
larscpressing enter does the same thing btw19:38
Fallenoulekernel: I want to have the bios prompt19:38
Fallenouit does not boot the bios either19:38
Fallenouit just does nothing19:38
lekernelof course you should press ESC19:39
lekernel*while the board is rebooting*19:39
wpwraklekernel: something along these lines: http://downloads.qi-hardware.com/people/werner/tmp/onebutton-fsm.pdf19:39
lekerneli see19:41
lekerneland I guess we map that to the middle button of the existing boards? and keep the current behaviour for the other two buttons?19:42
Fallenouhum strange, I tried unplugging power supply, keeping ESC pressed (in uart, not a real USB keyboard plugged to M1), then plug power supply again, it prints a few letters and then nothing more19:42
wpwraksounds good to me19:42
lekernelFallenou: you can use reset instead of unplugging the power supply, and what are those letters?19:43
lekernelyou can also use Q instead of ESC19:43
kristianpaulreset or make boot :)19:43
lekernel(capital Q)19:43
wpwraklarsc: btw, thanks for reminding me about my FSM - i had already forgotten about it ;-))19:45
wpwrakFallenou: the Esc is not very timing-critical. it's sufficient to press if when you see the first sign of life from the BIOS19:46
Fallenouthis is with holding capital Q when booting19:46
wpwrakpressing something while booting may uncover unexplored dangers :)19:46
wpwrakhmm, just one press should do19:47
Fallenouoh damn19:47
Fallenouit's just minicom crap19:47
lekernelFallenou: use flterm19:47
Fallenouusing flterm I get the bios19:47
lekernelyes, that's also why I wrote flterm... iron out all this serial port mess19:47
FallenouI think the bios will be helpful to debug mmu19:48
wpwrakah, you wrote flterm. interesting.19:48
Fallenousimple program, easy to check program control flow19:48
lekernelor you can find out what set of minicom options happens to work, if you have some time to waste19:48
wpwrakhow about making it easier to pick the gdb pass-through device ? not sure what would be the best approach. perhaps setting a symlink specified by the user or such19:49
Fallenouooooh I get all the bios prelude now when booting ! flterm rocks19:49
kristianpaulFallenou: ;)19:52
kristianpaulflterm plus jtag-boot19:52
Fallenoujtag-boot ? what is it ?19:56
Fallenouserialboot ?19:56
kristianpaulFallenou: http://projects.qi-hardware.com/index.php/p/wernermisc/source/tree/master/m1/jtag-boot19:59
kristianpaul make boot; make standby; life esier !19:59
Fallenouwhat does it do exactly ?20:01
kristianpaulavoid you push buttons20:01
Fallenouwell make load-bistream loads a new bitstream and reboots the system without pushing any button20:02
wpwrakyeah, jtag-boot is for when you have the bitstream in NOR. also, if you have the new standby bitstream that doesn't wait for a button, you can just do a  pld reconfigure20:14
Fallenouhum ok20:14
Action: Fallenou just understood that he needs to modify lm32-gnu-as to add support for MMU CSR20:15
wpwrak(pld reconfigure) instead of jtag-boot. not instead of pld load.20:16
Fallenouare you using gcc 4.5.2 or 4.5.3 ?20:18
kristianpaulmay be too early to ask, but this MMU will be optional in the future? :)20:20
Fallenouimo it will easily be optional20:37
Fallenoujust like ICACHE and DCACHE20:37
larsclekernel: any better idea how to find out which actors may qualify for timesharing than comparing each actor which every other actor?21:15
larscfinding a valid schedule fast is the next problem21:17
kristianpaullekernel: had you ever used TIG in constraint file? ;)21:32
Action: kristianpaul wonder if lattice have such extensive documentation as xilinx does21:35
lekernellarsc: no, not really...21:38
lekernelkristianpaul: yes, see tdc core21:38
kristianpaulah intereting21:40
--- Mon Jan 30 201200:00

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