#milkymist IRC log for Monday, 2011-12-19

cladamwwpwrak, seems (2011-07-13) on your M1pre-rc-4 works well. And mine ( usb switch )works well from latest images.01:16
wpwrakcladamw: yeah. need to try more recent images. if it works with the july version, it most work with newer versions, too :)01:30
wpwraks/most/must/01:30
cladamwwpwrak, okay.01:31
wpwrakso all seems fine so far. also the NOR bug hasn't reared its head01:32
wpwraki now have ~11500 cycles. last check was at 10000. not one bit went missing :)01:33
kristianpaulnice01:38
wpwraklekernel: yes, the syntax looks a lot better01:43
wpwraklekernel: you're still visibly limited by python, though. e.g., the structure of what goes on inside "If" is a bit odd01:46
wpwraklekernel: oh, and do you need to initialize things like self.tx in __init__ ? if you did it in get_fragment(), you wouldn't have to use all those "self"01:47
wpwrak(stuff inside If) at least i assume that foo.eq(bar) means "foo == bar". but it's not obvious why you sometimes use .eq and sometimes ==01:49
wpwrakyou could gain access to those operators by building everything with your own classes and then overriding the corresponding functions. but it would be a bit of a mess.01:51
wpwraki explored the concept of using python as a framework for making my own language (with a pythonish syntax but my semanstics) quite a bit in the tmc package (the stuff i use to decode usb). you can get pretty far, but you also have to tweak things very deep inside python if you really want control.01:53
kristianpaul:) * WARNING * WARNING * WARNING * WARNING * WARNING * WARNING * WARNING *01:57
kristianpauljust one instr class?, lets read more..01:59
kristianpaulIntersting, python have some sort of flexibillity to achive this own language stuff, even more interesting when links to real hardware02:01
kristianpaulm1.send_midi_val1(value)02:04
kristianpaulfn.set_variable(wave_mode=6)02:05
kristianpauland the live coding start :)02:05
kristianpaulor something like that, tought i never had used python before i guess i did lot of typos02:05
wpwrakyou want to make the syntax more complex ? :)02:06
kristianpaullol no i tought in your aprouch with the multipler, thinking in milymist patch editor and previous comment of people wanted to do live coding on m102:07
kristianpaulbut i guess sintax could be simpler02:07
kristianpaulfn.wave_mode(6) ?02:07
kristianpaul:)02:08
wpwraki doubt "live coding" makes a lot of sense. not that the idea per se would be flawed. but you don't have the resolution for showing much code. so you'd have to scroll around all the time. rather inconvenient.02:09
kristianpaulhmm, yeah i forgot that part02:11
kristianpaulof course asuming support for on the fly patch compilation :) too02:16
wpwrakanother issue is that recompiling the whole file is slow. and you have to do this after every keystroke. while rendering. so you may have to add preemption and partial recompilation. nothing infeasible, but messy.02:16
kristianpaulyeah..02:16
whitequarkwpwrak: it's Ruby which is called a "DSL creation language", not python. And it has a reason :)04:11
whitequarkwpwrak: (not one bit) do you check the whole NOR for errors each time ?04:12
wpwrakwhitequark: i check the lock bits, the entire standby partition, and the crcs of the other partitions. about each day.05:03
wpwrakwhitequark: by far most corruptions would hit the standby partition05:04
wpwrakso there are some bits i don't check. but they're far out. also a bit could fail back and forth, but that's only likely in the case of lock bits, and that would almost certainly be accompanied by massive corruption of data (unless it conveniently corrupts to locked soon after i start, corrupts to unlocked shortly before i stop, then corrupts to locked again once i restart, etc.)05:06
whitequarkhuh, "conveniently corrupts"05:08
whitequarksounds like a description of how government works.05:08
natarajHi06:32
natarajIs this an arm arch device?06:32
xiangfunataraj, Hi this channel is talk about Milkymist SOC. :)07:02
lekernelwpwrak: self.tx/self.rx are meant to be accessed from the outside10:30
lekernel.eq means assign10:30
lekernelI couldn't find a better word... "set" clashes with set(), "assign" is too long, "ass" would be the target of easy jokes10:32
wpwrakah, so the first argument of If is the condition, the rest the body. i see. good old basic's "let" ?10:44
wpwrakor, better english, "be" ;-)10:45
lekernel"be" looks like a good candidate11:24
wpwrakoh, and don't you want to get rid of that silly size on constants ? does it actually ever make sense to have that ?11:39
lekernelin many places, no11:40
lekernelconstant sizes are appropriate, however, when concatenating bit vectors together11:41
wpwrakoh, totally11:41
wpwrakbut that's maybe 1-5% of all uses :)11:42
xiangfuthe ftpd bug maybe relate to the yaffs filesystem. error on create file.11:49
wpwrakxiangfu: did you check the file name it's trying to use ? maybe there is something invalid in there11:56
xiangfuwpwrak, no. it not happen on particular file.11:57
wpwrakyou mean it happens on all files ?11:58
xiangfuwpwrak, different files. when upload files 'parallel'  the error will happen.12:00
wpwrak"parallel" = you have more than one file open at a time ?12:08
xiangfuI found it failed when do 'yaffs_create_obj'12:11
xiangfuwpwrak, 'parallel' mean. copy 4 ~ 10 files in GUI ftp application. then it will open more ftp connection upload parallel.12:11
xiangfucopy one file is rarely give error.12:11
xiangfuwpwrak, another thing: the '/ssd' is 644 not 755 :(12:12
wpwrakaah, i see12:12
xiangfuwpwrak, you can check you m1. the '/ssd' is 644. I try to add one line after mount /ssd it will give 75512:12
xiangfuhave to add chmod after 'mount("/dev/flash5", "/ssd", "yaffs",.........'12:13
lekernelmkdir("/ssd", 0777);12:14
lekernelno, you should NOT add chmod12:14
lekernelbut fix the bug in the first place12:14
lekernelthe RTEMS filesystem architecture is already an horrid, steaming pile of shit, please don't make it even worse12:14
lekernelas in "let's duplicate the clumsy and buggy string manipulation code in every filesystem" and things like that12:16
xiangfulekernel, yes. fix the root cause.12:17
lekernelbtw, there's a hard limit of 4 (iirc) parallel FTP connections12:18
lekernelif it refuses connections after that, it's expected12:18
xiangfu312:18
xiangfuthe error is give is:  Access failed: 550 Error creating file. (Berries_by_Orb9220.jpg)12:19
lekernelbut I wouldn't be surprised if there are races, the RTEMS filesystem architecture is actually designed to encourage race conditions (there are even some cases where it's impossible to fix them)12:19
xiangfu0777, I think it's mount make it change to 644.12:21
xiangfuhttp://pastebin.com/EQSTuzZ512:21
lekernelthat filesystem arch was probably designed by a 10-yo or someone like that... and still I know 10-yos who could do better12:21
xiangfuI add three mkdir.. only /ssd give 644.12:22
lekernelI see12:22
lekernelwell, have a look at the implementation of stat() in the YAFFS port12:22
xiangfuhmm.. no far didn't see the message  'ycb_statvfs  printf("%s\n", __func__); '12:31
lekernelstat, not statvfs12:33
wpwraklekernel: celebration of seasonal diseases ? :)14:17
lekernelhttp://www.fiber-space.nl/call2012/14:34
azonenberglekernel: Just out of curiosity (not that i'd do it, or have heard of anyone doing it)14:39
azonenbergBut is it possible to GPL a patent?14:39
azonenbergiow, you invent some technology14:39
azonenbergYou patent it14:39
azonenbergAnd then you issue an open license to all GPL'd projects14:40
azonenbergbut a non-GPL'd project can't legally use it without licensing14:40
azonenbergbasically a copyleft patent14:40
kristianpaulif it copyleft why you patent it?14:41
azonenbergThe intetn would be to prevent proprietary software from implementing your idea14:41
azonenbergso only free software is compatible14:41
azonenbergI am not recommending it, just curious from a legal perspective if you guys think its even possible14:42
kristianpauli mean do you want recognition of your work trought the patent? or wand a fee and absulte itelectual pooperty control?14:42
kristianpaulah isee14:43
azonenbergBasically ,the same way GPL means only GPL'd projects can use your *code*14:43
azonenbergcan you do the same thing with an idea14:43
azonenbergnobody can implement your algorithm unless they GPL the code14:43
azonenbergAgain, i'm not recommending the practice14:43
azonenbergJust wondering if it could even theoretically happen14:43
lekernelthe original GPL maybe not (it's based on copyright), but a similar system could be possible imo14:43
lars_sure, if you own the patent you can define the conditions under which the patent is used14:43
azonenberglars_: thats what i was thinking14:44
wpwrakafaik, there would always be the risk that you lose ownership of your patent14:44
azonenbergso basically you patent it, then grant an open license to any open source project14:44
wpwrakand the new owner wouldn't necessarily be bound by your licenses14:44
azonenbergwpwrak: How would that work?14:44
kristianpaulpantented but fee less i guess lars_ ?14:44
wpwrake.g., if you go bankrupt14:45
lars_azonenberg: you sell it14:45
lekernelpatenting anything costs $10k and incredible amounts of time wastage due to paperwork, legalese and lawyers14:45
azonenbergagain, i dont think it's likely to happen14:45
azonenbergwpwrak: easy, create a shell company or foundation that exists solely to oversee the patent14:45
kristianpauli hope dont :)14:45
azonenbergand has no assets or liabilities14:45
lars_there are opensource patent pools14:45
kristianpaulyeah, creative commons tried..14:45
azonenbergPersonally i dont like copyleft14:46
azonenbergi like permissive licenses14:46
lars_which are itendent to be used when a opensource software is attacked with a patent to start a counter patent claim14:46
kristianpaulazonenberg: cc0 :)14:46
azonenbergso rather than patenting something i'd just publish it :p14:46
wpwraklars_: if you sell it, that would normally include contractual obligations. but if it's taken away from you, that strips many obligations14:46
lekernelCC0 is just public domain no? why do they call it 'creative commons'?14:46
kristianpaulazonenberg: publis it certainly is a better way to avoid soemboy patent it, but of course you still need to make lot of noise14:47
wpwraklars_: basically the extra obligations stay with the previous owner, and the new owner just gets the valuable item, to try to turn it into a profit14:47
kristianpaullekernel: because public domain in each country is a mess14:47
wpwrakazonenberg: a company without assets wuold be even easier to destroy ;-)14:47
azonenbergwpwrak: how?14:47
kristianpaullekernel: so creative commons try to stablish a common way to public domain something valid/reviewed in each country14:47
kristianpaulsame as the other cc licenses14:48
wpwrakazonenberg: create any upset you want ;) poof, it runs out of money. remember, you can sue anyone at any time over even the most ludicrous things14:48
kristianpaulthats why you usually use the reviewed cc lincese for your place14:48
kristianpauleven when ie, eur uses cc by 3.0, colombia may just use upto 2.5 version of same license14:49
azonenbergwpwrak: maybe14:49
kristianpaulah, lawyers..14:49
azonenbergIn any case i wasnt recommending the practice, or considering it feasible14:50
wpwraklekernel: PD isn't always free. e.g., some countries have what's called "public domain payant". anything that enters the public domain then requires royalties to some typically cultural organization14:50
azonenbergmore curious if it had been tried14:50
kristianpaulazonenberg: still a patent so same dislike a copyrighted license like gps or other has, a patent well.. follow the principles..14:50
wpwraki would also consider publishing the safest approach. that way, you avoid creating "intellectual property" in the first place14:51
azonenbergwpwrak: The only conceivable reason to do something like that14:51
azonenbergwoudl be if you want to, say, implement a file format with some novel feature in your open source project14:51
azonenbergand prevent m$ products from ever being compatible14:51
azonenbergwhile allowing open code to interoperate14:51
azonenbergIt sounds like something stallman would try if he had half the chance14:52
kristianpaulof course if you like the idea of making something extremely free to use14:52
wpwrakbesides, with every patent you get, you add one to the statistics that show how successful the patent system is14:52
azonenbergIt's an offensive technique, not defensive like PDing14:52
kristianpaulwell i dont see why you concern about some one implement it in propietary way14:52
wpwrakplus you hand some amount of money to the enemy14:52
kristianpaulfree well :)14:52
azonenbergkristianpaul: Personally i'd just BSD the code and to heck with anyone else14:53
azonenbergif they implement it, i dont care14:53
azonenbergbut knowing stallman and his offensive strategies, it seems like he'd try that14:53
azonenbergeven the GPL is an offensive license14:53
azonenbergnot a defensive one14:53
wpwrakwhy on earth would you want prevent MS from implementing it ? if it's something you want to see widely used, the more sources the merrier14:53
azonenbergwpwrak: The intention would be to become the dominant application in that field14:54
azonenbergand force m$ out of the market14:54
azonenbergbecause they cannot interoperate14:54
wpwrakwould you rather have MS and their buddies force some competing format into the market, heavily guarded by their patents ?14:54
azonenbergThe hypothetical application would be an office format14:54
wpwrakdream on ;-)14:54
azonenbergthat somehow became dominant14:54
kristianpaulwpwrak: indeed14:54
azonenbergbut couldn't be implemented by proprietary applications14:54
azonenbergAgain it seems unrealistic14:55
azonenbergbut knowing the FSF nutjobs i suspect if the opportunity ever arose they'd try something like that14:55
kristianpaulazonenberg: for you initial question gpl is not like it is, interionally i think, i mean you need to be offensive if want to defend your position14:56
wpwrakso each time, say, a state entity would ask fir bids, all the proprietary guys would shout bloody murder because it's a discriminatory standard. very good :)14:56
kristianpaulazonenberg: actually if that patent you think about exist, will be somethin or even more complicated that reading the glpv3 or such :)14:56
azonenbergkristianpaul: its offensive because it was designed by the FSF14:56
azonenbergand stallman doesnt want to compete with proprietary software14:57
wpwrakhmm, do you really know the FSF legal guys ? :)14:57
azonenberghe wants to destroy it14:57
azonenbergi know stallman well enough14:57
azonenbergi've had a few f2f discussions with him about his offensive tactics14:57
kristianpaulwpwrak: (legal)even monglen? guess i dont either14:57
azonenbergcausing splits in the community (vs *bsd etc)14:57
wpwrakwho won ? ;-)14:57
wpwrakkristianpaul: yeah, for example14:57
azonenbergWe called a halt at his stop lol14:58
wpwrak;-))14:58
azonenberg(we were driving somewhere)14:58
kristianpaullol14:58
azonenbergno clear victor14:58
azonenbergneither was able to convince the other of anything14:58
azonenbergNot that i was surprised at atll14:58
kristianpaulhe, you had a bad experience with rms, that used to happen ;)14:59
kristianpaulso better ignore him and keep doing you nice work :)14:59
kristianpaulreally14:59
kristianpaulmix social behavior with licenses is not a good thing14:59
azonenbergI know, just saying i got the impression pretty quickly he wants to destroy proprietary software by any means possible14:59
azonenbergwhile i'd rather compete with it in a free market and win because i have a superior product15:00
azonenbergor lose, because mine is inferior, in which case they deserve to win15:00
azonenbergin other news i think i am finally figuring out how i am going to handle the interconnect in that multicore SoC i've been designing15:00
rohthe moment where you think more about licenses than good technical designs you have already list.15:00
azonenbergi wanted something with support for DMA and out-of-order request handling which wishbone didnt seem to do15:01
rohirgh. s/list/lost/g15:01
azonenbergmy new design has a lot in common with infiniband15:01
wpwrakxiangfu: hehe, that media wall somehow reminds me of the series "The Prisoner" ;-)  1960-70 style mystery tech15:01
azonenbergswitched fabric supporting multiple simultaneous connections15:01
rohlicenses do not save you from using common sense and rethink what it is you want to archieve by using some specific one.15:01
azonenbergand out of order request handling15:01
azonenbergTCP-esque retransmission if a packet is dropped due to congestion15:02
azonenbergmy system will tentatively have 16 ports on the switch between the three CPU cores, RAM, and various high-speed peripherals like the cameras15:03
rohan yes.. ignore rms... he is kind of a troll when it comes to hw.15:03
azonenbergSound like a reasonable design?15:03
rohsounds scalable15:03
azonenbergroh: Good, because while my current design is 3 cores i want to do 5 in the future :p15:04
azonenbergon an xc6slx7515:04
azonenbergeach packet is 32 bits address and 64 of data plus a read/write flag (in read mode, data is ignored)15:04
azonenbergtimes 200 MHz is 12.5 Gbps per port15:04
azonenbergWhich conveniently is exactly the bandwidth of 16 bit DDR2 80015:04
rohi dont know about details.. i am not that deep into hdl design myself. but i know archs and what you propose doesnt sound unreasonable15:05
azonenberghmm, i wonder if thats a coincidence :p15:05
azonenbergSince the RAM has a whole switch port dedicated to itself15:05
azonenbergBut yeah, i basically want a switched fabric network rather than a bus15:05
azonenbergand something that supports out of order delivery of data15:05
rohisnt pci-e similar to that? or is that really still a bus?15:05
azonenbergso for example if core 0 is busy talking to the camera and core 1 wants data from RAM, an outstnading core0 request from RAM should not block the core1 request15:06
azonenbergroh: Similar, but pcie is meant for off-chip use15:06
azonenbergmultiple high-speed serial lanes15:06
azonenbergthis is an on-chip bus15:06
azonenbergalthough i suppose i could potentially add future support for a serialized version that does off-chip communication15:06
rohsure15:06
azonenbergI dont know if pcie is switched or still a bus though15:07
azonenbergi know infiniband is15:07
azonenbergand this bus has comparable bandwidth in this implemenataion15:07
azonenbergDDR 4x infiniband is 16 Gbps and my bus is 12.5 (one way)15:08
rohmy impression was that all 'bus' type interfaces got serialized to save pins, differential for data integtrity and peer based (only 2 communicating partys per 'bus) nowadays15:08
azonenbergthe difference is that i'm parallel on-chip and infiniband is serial off-chip15:08
kristianpauloh you have already 3 cores working with tool chain and libc supported included in your soc?15:08
azonenbergkristianpaul: I havent implemented the cores fully yet, most of the design is done15:08
kristianpaulis this work published some where? :)15:08
azonenbergI wanted to design the architecture fully first15:08
kristianpauli see15:08
kristianpaulloooks nice15:08
azonenbergit's heterogeneous15:08
azonenbergsomehwat like the Cell15:09
azonenbergone mips-like core for non-parallel code15:09
azonenbergfast and pipelined but no FPU15:09
azonenbergand then two highly parallel DSPs15:09
azonenberg2-way superscalar, 16-way barrel15:09
azonenbergi.e. 16 hardware threads and i issue two instructions from one thread every clock15:10
azonenbergper core15:10
azonenbergThe DSPs are a completely custom archiecture, i havent even begun the ISA yet15:11
azonenbergbeen working on microarchitecture15:11
azonenbergbut it will be 64 bit VLIW with two 32-bit instruction fields15:11
azonenbergonly one of every execution unit but you can issue instructions to any two every clock15:12
azonenbergi have the register file implemented already, it's 4 reads and two writes every clock15:12
azonenbergall reads, and all writes, must be within the same two 16 bit windows15:12
azonenberg32 register*15:12
azonenbergread and write windows need not be the same15:12
azonenbergin fact they will usually be shifted by a bit due to pipelining15:12
azonenbergRight now the current focus is the interconnect fabric15:13
azonenbergonce i get that designed i'm going to work on the CPUs more15:14
azonenbergi was thinknig a shared bus and even maybe using wishbone for a while but it didnt seem like it'd play nicely with multiple cores and DMA15:14
azonenbergand specifically request reordering15:14
azonenbergWhich is going to be a must if i want the system to scale15:15
Action: kristianpaul remenebers a xmos video15:21
wpwrakreordering of requests from the same source ? or just global reordering ?15:22
wpwrakand what will that beast do when it's finished ?15:23
azonenbergwpwrak: Flight control for a UAV my school's electronics club is building15:23
lekernelroh: you could have stopped at "troll" :)15:23
azonenbergincluding onboard machine vision and path planning15:23
azonenbergi'm shooting for peak performance of 800 mflops in the dual DSP version and 1600 in the quad15:23
azonenbergI also am considering building a laptop around a slightly modified version of the same SoC15:24
azonenbergSince i've wanted to make a full custom laptop for a while and i think this thing will have the juice to do so15:25
azonenbergi'd probably use the quad core model in the laptop15:26
lekernelwith a FPGA?15:26
azonenberglekernel: yes15:26
azonenbergi want a full custom CPU + GPU15:26
azonenbergand MOSIS is a little pricey15:26
azonenbergThe goal of the project would be to build a laptop starting with the RTL and going all the way up to the mobo, case, and operating system + apps15:27
wpwrakazonenberg: oh, drones for iran. nice ;-)15:27
azonenbergwpwrak: Lol, this one isnt going to be able to stray that far off course15:27
azonenbergit communicates with the base station by 802.11n15:27
azonenbergand will force-land immediately if signal is lost15:27
azonenbergthis is an indoor micro-uav15:27
azonenbergAnd strictly research oriented15:28
lekernelazonenberg: what does it need compute power for?15:28
wpwrakheh, that's what they all say :)15:28
azonenberglekernel: It's autonomous15:28
lekernelKalman filters? radar processing?15:28
azonenbergactually it uses a kinect for 3d position sensing15:28
azonenbergmy roommate already has some pretty nice algorithms done up for point cloud processing15:29
azonenbergthen yes, a kalman filter15:29
lekernelazonenberg: do you know how the kinect works btw?15:29
azonenbergidentifying interesting targets in video feeds15:29
azonenberglekernel: yeah, stereo on a grid of projected IR points15:29
azonenbergThat's good enough, all it has to do is find walls and such15:29
azonenbergwe originally considered LIDAR but it was way too expensive15:30
lekernelit's only expensive if you buy the pre-built modules, no?15:30
azonenbergkinect gave slightly less good data for much less cost15:30
azonenberglol, homebrewing would be nice15:30
azonenbergbut is beyond our time budget15:30
lekernelthe Roomba has a cheap mass-produced LIDAR for example15:30
azonenbergthe main difficult part of hte project is the flight planning algorithms15:30
azonenberglekernel: 2D?15:30
azonenbergi'm pretty sure its only 115:31
azonenbergremember this is a UAV, we're working in 3-space here15:31
lekernelI don't know15:31
azonenbergI'm not on that team anyway15:31
azonenbergi'm doing the SoC15:31
lekernelbut it's the same basic technology, and adding another scanning axis doesn't justify a 100-1000x price increase imo15:31
azonenbergincluding the DSP and various support stuff like the camera controller15:31
azonenbergand hardware JPEG compression unit15:31
azonenbergoh, and a hardware driver for the kinect including associated usb stuff15:32
lekernelyou could try a SAR too :)15:33
azonenbergBetter position sensing would be nice, but the data we got off the kinect looked adequate15:33
azonenbergagain, all it's needed for is seeing whether we're headed for a door or a wall15:33
azonenbergand registering that data with our existing data to build a map15:33
azonenbergi do agree that lidars are way more expensive than they need to be15:35
azonenbergsame thing with, say, semiconductor fab equipment15:35
azonenbergbut i dont have time to reinvent EVERYTHING ;)15:35
lekernelhow's that DIY fab going, btw?15:35
azonenbergPut on hold for exams, just finishing up15:36
lekernelsounds more interesting to me than kinect "hacks" that everyone does :-P15:36
azonenbergThe choice to use a kinect wasnt mine15:36
azonenbergit was the flight control team15:36
azonenbergwe have like 15 people on the project now15:36
azonenbergRe fab, i have some very interesting ideas15:37
azonenbergThe first is fooling around with packaging, i want to try making a FCBGA15:37
azonenbergjust with some simple wiring patterns, no active compnents15:37
azonenbergto test if i can actually package chips i make15:37
azonenbergsecond is lithography related15:37
azonenbergi want to build a fairly complex mask with a lot of differnet designs on it15:38
azonenbergsend it out to laserlab to be fabbed15:38
azonenbergthey'll give me a 10x16 inch piece of film at 8000DPI which equals 12.5um design rules and 3.175um lambda15:38
azonenbergwhich comes out to quite a few 2" contact masks and 2cm projection masks15:38
azonenbergi should be able to easily hit single-micron design rules with optical reduction of a 12.5 or 25um mask15:39
azonenbergThe optics say that i can even hit 350nm15:39
azonenbergnot sure if my dust / photoresist / developer / etch combination will let me go quite that far15:39
azonenberg500 to 750 might be reachable15:40
kristianpaulis this going to be impleneted in the spartan3 and the boarb you showed the other at #homecmos ?15:41
azonenbergAt that point i could start thinking pretty seriously about transistor fab15:41
azonenbergkristianpaul: the SoC? No lol15:41
azonenbergthat was a 50k gate chip15:41
azonenbergi'm using the XC6SLX4515:41
kristianpaulyeah  :)15:41
kristianpaulNice15:42
azonenbergCurrent estimates say i can fit the tri-core version there15:42
azonenbergand the quad in the -7515:42
azonenberg the 5-core*15:42
kristianpaulown board or generic kit?15:42
azonenbergRight now i'm using the Atlys dev board15:42
azonenbergi expect to do my own in the future15:42
azonenbergone of the things i've been working on on the side is developing a BGA soldering process15:42
azonenbergi cant handle csg324 due to the 0.8mm pitch but fgg484 or ftg256 should be doable on affordable design rules (6 mil trace/space, 4 layers, no filled vias, no blind/buried vias)15:43
azonenbergmaybe 6 layers depending on how thigns go15:43
kristianpaulyou said case too,is any one working on this?15:44
azonenbergkristianpaul: the UAV or the laptop?15:45
azonenbergThe laptop is just me and a lower priority project15:45
azonenbergi'm quite handy with a milling machine tyvm15:45
kristianpaullaptop :), yes15:45
azonenbergThe laptop is just me15:49
kristianpaulbut the UVa will use this spartan6..15:49
kristianpaulUVA*15:49
azonenbergmy plan for the case is TiG welded 1/4" aluminum plate15:49
kristianpaulwhat about power management?15:49
azonenbergThe UAV may use the atlys or a custom board15:49
azonenbergwe're not sure yet15:49
azonenbergI havent though that far haead15:49
azonenbergmost likely going to be built around a pair of 5 Ah 11.2v li-po packs15:50
lekernelazonenberg: what's so special about the atlys btw? everyone's using that15:57
lekernelhdmi? ddr3?15:57
lekernelsomething else?15:57
lekernelhmm, it's ddr2 actually15:57
azonenberglekernel: Cheap :p15:57
lekernel$349.00?15:58
azonenberg19915:58
azonenbergfor educational15:58
azonenbergit was the only spartan6 board i saw that had a decently large fpga (lx45), a decent amount of memory (128MB), and a decent selection of peripherals15:58
kristianpaulhow do you program it?16:03
kristianpaulcan you tried in linux with urjtag?16:03
kristianpauls/can/had16:03
azonenbergthey have an 8051 based adapter built in but i use the xilinx platform cable16:03
azonenbergwhich runs fine on my debian and ubuntu system after i installed an open source driver16:03
kristianpaulnot bad16:04
lekernelstill with impact?16:09
azonenberglekernel: yes16:17
azonenbergi want to build my own programmer at some point16:17
azonenbergmuch cheaper, probably based on PIC32 and a small CPLD or FPGA16:17
azonenbergtarget price point $50 or less16:17
azonenbergwill include a 12V Vpp generator16:17
azonenbergand should be able to do xilinx JTAG, atmel in-circuit programming, and microchip ICSP16:18
azonenbergif not more16:18
azonenbergcross platform libusb based command line interface16:18
azonenbergand lol16:18
azonenbergi just wrote libjpeg instead of libus...16:18
kristianpaulis not that a bus pirate? :)16:20
kristianpaulwell sort of16:20
lekernelwhy reinvent urjtag?16:22
lekerneland modern chips do not have Vpp :)16:23
azonenberglekernel: bus pirates lack Vpp generators16:29
azonenbergand most PICs still need high voltage on MCLR#16:29
azonenbergeven if the real Vpp is charge-pumped onboard16:29
azonenbergthey use it as a "time to program" indicator16:29
kristianpaulPICs, yeah..16:30
kristianpaulazonenberg: there is a expansion boar for vpp tought i havent bouht it yet..16:31
kristianpaulbought*16:31
kristianpaulbut there are low voltage version of PICs that allow vpp to 5v, but i still dont source it at least for 18F familly :)16:32
Action: kristianpaul looks if urjtag can program its PICs with aditional hw..16:32
lekernelmy comment was about "modern chips", and I believe it is still correct =]16:33
kristianpaulxD16:33
wpwrakaren't most recent PICs low-voltage programmable ? (i'm avoiding the word "modern" here as well :)16:44
azonenbergwpwrak: low voltage programmable, yes, but i think you have to set LVP in the config bits16:51
azonenbergthey need HV the first time out of the box16:51
azonenbergunless i'm mistaken16:51
azonenbergthe project is still a long-term todo, i havent really read the programming specs etc yet16:52
wpwraki think it's the other way around: they support LVP out of the box but you can turn it off - and then you need high Vpp to turn it back on again16:52
azonenbergHmm, interesting16:52
azonenbergi'll have to check that16:52
azonenbergi've always used default settings which i think is high Vpp16:53
wpwraki think you lose quite a few pins for this luxury, though16:53
azonenbergon the low-pin-count parts, yes16:54
azonenbergbut not the ones with 64 or 100 pins16:55
wpwrakisn't it the same number on both ?16:55
azonenbergproprortionally, i mean16:55
azonenbergi think LVP uses one extra pin on the programmer, which i've never hooked up in any of my circuits16:55
wpwrak(PIC with LOTS seems a little weird anyway. at some point, the cost of the core becomes negligible. so you could as well use something decent)16:56
wpwraks/LOTS/LOTS of pins/16:56
azonenbergwpwrak: i've never used a lot of the cheap 8-bit ones16:57
azonenbergmost of what i've used lately is pic32s16:57
azonenbergwhich are completely different16:57
azonenbergthe ICSP is just JTAG muxed onto less pins16:57
wpwrakok :)16:57
azonenbergAnd the core *is* something decent16:57
azonenbergmips m4k16:57
kristianpaulwpwrak: yes some came witht this vpp bit already programed from factory others dont :|17:00
kristianpaulI buyed a M1 EDK when was thinking in PIC32s :)17:01
wpwrakyeah, M1 is so much more fun than a PIG :)17:11
lars_and it even has an built-in avr ;)17:13
wpwrakand you could probably synthesize half a dozen more, with all the space that's still free :)17:14
kristianpaulyes of course17:14
lars_but would probably timings would become critical17:17
lekernelif they don't have to communicate, there's no problem with timing17:22
lekernelyou can simply build one structure that meets timing, and replicate + translate it17:22
lekernelthe S6 architecture is homogenous17:22
kristianpaulhow do you actually (in commands) replicate and translate it?17:23
kristianpaulhackign bitstream?17:23
lekernelyou can do that, or build a NMC macro17:24
lars_yeah, but don't you want SMP AVRs? :/17:26
wpwrakwhy not NUMA ? :)17:27
lars_or NUMA17:30
wpwrakor maybe synthesize a few VAX CPUs and make a VMS cluster :)17:32
lars_hm, 150km and UTMS is already breaking down17:32
lars_km/h17:32
wpwrakbalancing the laptop on the steering wheel ? :)17:33
lars_hh17:33
kristianpaulhe17:34
lars_wpwrak: balancing it on the table of the train17:51
wpwrakah, safer ;-)17:54
lars_he, next train station, next bunch of emails arrives18:00
lars_and this ride is going to last another 5 hours :/18:02
wpwraktranssibirian ?18:03
kristianpaulat 150km/h !18:03
kristianpaulvisiting russia,cool :)18:04
lars_transgermany18:04
lars_all from the south all up to the north18:06
wpwrakah, visiting the CNC machines :)18:07
lars_right18:08
kristianpaulhmm, wait, why 5hrs germany is not that long..18:09
kristianpauli guess to much stops, but at 150km/h !18:09
lars_i think it is actually faster sometimes18:10
lars_it = the train18:10
lars_and the distance is over 700 km18:11
lars_but i think the route is not a straight line18:11
wpwraki guess there may also be snow. isn't it that the german railways suffer catastrophic breakdown when there's snow ?18:12
lars_mainly the berlin s-bahn18:13
wpwrakaah. lekernel should like that then. he's very fond of delayed public transportation :)18:14
whitequarkwpwrak: why does he?18:38
lekernelberlin s-bahn is actually quite OK when compared to the parisian RER18:38
lars_wait for the winter ;)18:39
lekerneland when it's not, I can use my car, which simply isn't possible amid the parisian traffic and paid parking lot thieves18:39
azonenbergNUMA AVRs? Lol18:40
azonenbergReminds me of the time i was thinking of building a PIC32 cluster18:40
azonenbergnice little blade maybe 1:10 scale of a 2U rack case18:41
azonenbergmake a little model rack18:41
GitHub69[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/7774ace7e1f4e5bf594edb5305687386f43e784721:19
GitHub69[migen/master] README: structure + FHDL description - Sebastien Bourdeauducq21:19
GitHub177[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/d9dc604c99984355634898a939f9e9f2cf99ee8322:29
GitHub177[migen/master] README: Core Logic, Bus, Bank - Sebastien Bourdeauducq22:29
GitHub156[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/47d321cd75fb3ed8c21ecf587063cba646898b3823:12
GitHub156[migen/master] README: Flow - Sebastien Bourdeauducq23:12
--- Tue Dec 20 201100:00

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