#milkymist IRC log for Saturday, 2011-05-14

wolfspraulkristianpaul: I've just sold my last m100:12
kristianpaulwolfspraul: :D00:12
kristianpaulpanic mode now? ;)00:12
wolfspraulyeah, great!00:12
wolfspraulno, not at all00:12
wolfspraula little sad that I'm so slow with rc300:12
wolfspraulbut can only blame myself00:13
wolfspraulI also have no more units now to send to reviewers, which will be a problem I need to think about.00:13
wolfspraulchicken & egg problem...00:13
wolfspraulif I want to prepare for a launch, I need to send something to an interested journalist/reviewer00:13
kristianpaulI tought you owned a mm1?00:14
wolfspraulyes sure, we still have units here and there, of course00:16
wolfspraulbut I don't have one just to admire it00:16
wolfspraulanyway the last one for sale was sold, which is great00:17
kristianpaulyes :-)00:17
kristianpaulAre you aware of the stock from others distributors?00:17
wolfspraulnot really, just guessing00:18
wolfspraulthey are not a very talkative bunch00:18
wolfspraulmost of the time my emails to them get ignored unless I include some capitalized threats or so00:18
kristianpaulhe :-)00:19
wolfspraulI think tuxbrain and others must still have a few units00:19
wolfspraulbut not sure00:19
wolfspraulif they don't talk, I'm playing bad guy here and assume they are perfectly happy...00:20
wolfspraulhow is your gps-sdr project going?00:20
kristianpaulI finished the core for reading sige data stream, in simulation "seems" to work, now writing software part00:21
wolfspraulsounds good00:23
kristianpaulno more news, and as wpwrak pointed this gps receivre dint get a fixed point ;-) (YET)00:23
kristianpaulwell that compared with some weeks ago, yes :-)00:23
Action: kristianpaul too lazy00:23
wpwrakwolfspraul: i think the trick with tuxbrain is peer pressure. of you copy victor, the chance of a response seems to increase00:35
wpwrakwolfspraul: when i pinged him about smt, he replied that he'll try to find time next week to go through the production process. so nothing bad has happened yet :)00:40
wolfspraulhe needs to schedule a date with the smt shop00:45
wolfspraulI don't know that company but normally that will be 1-2 weeks out00:45
wpwrakyeah. i asked him for that date.00:46
wolfspraulwith such a small job, they may squeeze it into the hickup of a larger job, if tuxbrain is flexible enough to handle that00:46
wolfspraulsay a phone call at 8 PM "do you want to have your run tomorrow morning at 9AM?"00:46
wolfspraulfor example because a larger job ran into a problem, the customer has to do some analysis, and there is something like a half day or day off in between00:47
wolfspraul(before the larger job continues)00:47
wolfspraulthe factory will charge for that waiting time, but why not do a second little extra job in there as well :-)00:47
wolfspraulif he has bad luck, he will get a date 4 or even more weeks out, and no opportunities to squeeze it in earlier emerge00:49
wolfspraulbut shouldn't be...00:49
wolfsprauloh, we are in the wrong channel :-)00:49
wpwrakyeah. depend a lot on how busy they are.00:50
wolfspraulalso how aggressive in pursuing every last penny00:50
wolfspraulthis is Spain...00:50
wolfspraulthe Chinese will be squeezing and squeezing and squeezing00:50
wolfspraulyou always find a spot to quickly do your thing, even if it's Saturday night00:51
wpwraki wouldn't be surprised of the fab in spain has a low duty cycle00:53
wpwrakfrom their responses, they don't seem to see themselves just as a factory, but also do some basic consulting and such. basically what newbie customers like tuxbrain and me need. the hardened smt veterans can try their luck in china ;-)00:56
Action: kristianpaul just discovered the scripting capabillities of rtems by using the in-Memory File System02:19
lekernelazonenberg: because of lack of time to try other modes06:57
lekernelyeah it has TMDS and same problem06:57
azonenbergBtw, have any of you guys tried hand soldering 256-BGA at 1mm pitch? (like the xilinx fpga package)07:03
azonenbergHand meaning at home, not with an iron07:03
azonenberghot plate, toaster oven, etc07:03
azonenbergand if so, did you get it to work?07:03
lekernelno... btw, if you have it soldered professionally, the cost is usually in the $40 range for a prototype07:08
azonenbergFor a one-off unit?07:08
lekerneldecreases rapidly with volume07:08
azonenbergand is that for the entire board or a single large bga?07:08
lekernelsingle bga, but I don't think the entire board would cost a lot more (except if you use automated pick and place, then you'll have to add the machine programming time)07:09
azonenbergI'm doing a couple of FPGA designs right now that haven't left simulation but i'm going to have to move them to real hardware at some point07:09
azonenbergI see07:09
azonenbergI think i'll probably try my hand at homebrewing first, see if i can make it work07:09
azonenbergif it fails, $30 down the drain07:09
azonenbergand i go have it done right07:09
azonenberg($30 being the cost of the FPGA i'm gonna be using)07:09
azonenbergAlso, I have an interesting problem I'm working on, maybe you'll have some suggestions07:11
azonenbergI'm trying to do counting in arbitrary base07:11
azonenbergIOW i have a series of regs representing digits, and a value indicating the base of the arithmetic07:11
azonenbergeach cycle i add 1 to the rightmost one and if it exceeeds the base, wrap and bump the next one etc07:12
azonenbergDo you think carry lookahead is feasible in, say, base 94?07:12
azonenbergOr some other relatively large non-power-of-two base?07:12
azonenbergEach digit i just use a normal synthesized CLA to increment, but i need to handle carry between digits myself07:13
wolfspraulbtw we sold our last Milkymist One, at least I don't have any one anymore07:13
wolfspraulthat's a fantastic result, unbelievable that we ended up selling 33 of the 40 we made only in rc2, our second run07:13
wolfspraulsome of the distributors may still have some, but I don't07:13
wolfspraulnow I need to really get my act together with rc3 and produce them asap :-)07:13
azonenbergwolfspraul, lekernel: any ideas re the high-base counting?07:14
wolfspraulnot me, sorry07:14
lekerneluhm, I don't know07:15
lekernelwhat's the problem with that simple implementation you described?07:15
azonenbergYou mean ripple carry?07:16
azonenbergI think it'll be too slow07:16
azonenbergThis will be running at 200-300 MHz on a spartan-607:16
azonenbergand needs to generate a result every clock07:16
azonenbergback of the envelope guess for 16 base-94 digits says 30+ gate delay07:16
lekernel16 base-94 digits? what do you need that for?07:17
azonenbergwell 8 is more likely, 16 is an extreme07:17
azonenberg8-10 is a realistic limit07:17
azonenbergI'm generating candidate plaintexts07:17
azonenbergstarting at AAAAAA and going to ZZZZZ07:17
lekernelthen you can make your counter go slower when there is a carry07:18
azonenbergthe base depends on the character set chosen (lowercase letters, uppercase letters, alphanumeric, alnum with symbols, etc)07:18
azonenbergThe pipeline after this will expect a plaintext every cycle is the thing07:18
lekerneljust add a clock enable signal on it07:18
azonenbergThat slows it down07:18
lekernelno, it doesn't07:18
azonenbergNo, i mean it skips a cycle07:18
lekernels6 registers has dedicated CEs07:18
azonenbergwhich is a potentially significant slowdown07:19
azonenbergi am trying to avoid needing that stall07:19
azonenbergi'm trying to figure out if there's an efficient way to generate a plaintext every cycle, pipelining as deeply as necesssary in the input generation logic07:19
lekernelin base 94, there is a carry every 94 cycle07:19
azonenbergYes, but what about base 26?07:19
azonenbergor 10, for numeric only?07:19
azonenbergI'll certainly consider your suggestion if that turns out to be the best way, but i want to look at all options07:20
azonenbergIn my CPU implementation i used ripple carry but if there's a more efficient way of doing things in hardware i'm hoping to do that07:21
azonenbergI already figured out a block-RAM based method of converting from digits to ASCII text in a single cycle07:21
azonenbergvs the relatively long delay i had with multiple data-dependent memory transactions in the CPU code07:21
azonenbergi'll be using a dual ported block ROM to generate two characters, then cascade 8 of them side by side07:23
azonenbergas long as i can keep the routing delays low, since nothing else will happen that clock cycle, i should be good07:23
lekernelazonenberg: btw for the BGA you might ask Andrey from Elphel07:45
lekernelhe assembled a couple of camera PCBs by hand, I'm not sure how he did the BGA though07:45
azonenbergi see07:45
wolfspraulI would bet he went to a local shop in Salt Lake City07:45
wolfspraulbut who knows, ask is a good idea07:46
azonenbergThe one I'm looking at using is an XC6SLX16-FTG25607:46
azonenbergwhich is 16x16 full array, 1mm pitch07:46
azonenbergnot super fine07:46
azonenbergi have another part that's available in either 64-TQFP, 100-TQFP, or 11x11 0.8mm BGA07:47
azonenbergi'm not sure if i can pull off the 0.8mm without paying for a much more expensive process with tighter design rules07:47
lekernelhi xiangfu08:47
wolfspraulbtw, excellent 05/09 snapshot08:48
xiangfulekernel: hi08:48
wolfspraulI'm having a lot of fun - can click through the patches, F1 to see the title of the patch08:48
wolfspraulvery good!08:48
lekernelyeah, i'd like to cut release 0.4 and have that ready for flashing the rc3's08:49
lekernelI'm only waiting for xiangfu to clean up the keymap code ...08:49
wolfspraulI assume adding video streaming to m1 will be really hard and/or take some time08:51
lekernelyeah probably08:51
wolfspraulso maybe we look for some vga grabber to take high quality videos?08:52
wolfspraulI need to look into this soon (the vga grabber)08:52
lekernelI have a friend who has one, in Paris suburbs ...08:52
lekernelunfortunately that region has the crappiest transports of all western europe08:53
wolfspraulI will look into it too, I'm still seeing what happens about video streaming, but my feeling is it will take several months, at least.08:53
wolfspraulnot the highest priority now actually, just for making videos/clips it would be nice ;-)08:53
wolfspraulgotta run, bbl08:53
lekernelsame... demo time soon08:56
lekernelespecially given the time burden taken by the aforementioned transports08:57
kristianpaulFallenou: Is this a correct way for define a pointer in rtems, volatile unsigned int *point_buf = (volatile unsigned int *)0xb0000000;15:10
Fallenouit seems ok for me15:15
Fallenouyou can have a look at bios code15:16
Fallenoui'm roaming on my phone15:17
kristianpaulsave traffic :-)15:17
Fallenouwell i waited 30 min15:21
kristianpaulargh!, a mmu is kinda madness :S16:53
kristianpaulmay be i dont understand it well, yet ;)16:54
kristianpaulrejon: hi, how is LGM going?22:28
kristianpaulor was..22:29
rejongoing well22:32
rejonstill in progress22:32
rejonmy talk went over very well22:32
kristianpauland mm1?22:32
kristianpaulor your "freedom box", had people seening working?22:32
kristianpaulsomwhere.. between talks.. in a walll..22:33
rejonno one had heard of milkymist before22:34
Action: kristianpaul wonder how many milkymist one are in north america22:35
kristianpaulhola :-)22:36
kristianpaulFallenou: you there?22:52
kristianpaulI was about to find out the way to increse ramdisk size22:52
kristianpaulor add another ramdisk..22:52
kristianpaulBut i dont find a specif driver for mm1, so i guess this is something internal to rtems..22:53
kristianpaulThere is a way to directly modify ramdisk number of blocks in flickernoise?..22:54
kristianpaulI mean not going to rtems-milkymist sources..22:55
kristianpaulI foudn this, http://wiki.rtems.org/wiki/index.php/Using_the_RTEMS_DOS_File_System#RAM_Disk22:55
kristianpaulbut seems to create a another ramdisk, but i dont think i need to do that at all..22:55
Action: Fallenou is back23:26
kristianpaulWellcome back !23:29
FallenouI can see no "mount" for the /ramdisk in main.c23:37
Fallenouyes just that23:42
Fallenouso the /ramdisk is just normal IMFS directory ?23:42
Fallenouwould make sense23:42
kristianpaulah, thats cheating :-)23:45
kristianpaulso i should instead implement a proper ramdisk it seems23:45
Fallenouwell, what would be the difference ?23:45
FallenouIMFS is already a file system "in ram"23:45
Fallenou"in memory file system"23:45
Fallenouit sounds like a ramdisk to me23:45
Fallenoumaybe I'm getting something wrong23:46
kristianpaulnot i think you got the point23:46
kristianpaul(difference) well i need give around 64Mb size to /ramdisk23:48
kristianpauljust to ramdisk? well23:48
Fallenouso you may want to increase the IMFS size I guess23:48
kristianpaulI think, i do23:48
kristianpauldo want*23:48
Fallenoukristianpaul: well there is something called "ramdisk" in rtems23:49
Fallenoudifferent than the IMFS23:50
Fallenoubut I am not sure if the /ramdisk of flickernoise *is* a RTEMS ramdisk (ramdisk driver)23:50
Fallenousince I can only see a mkdir23:50
kristianpaulseems is not ramdisk23:50
kristianpaul"The In Memory File Systems is always the root file system and uses the standard C library heap for storage"23:54
--- Sun May 15 201100:00

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