#qi-hardware IRC log for Thursday, 2012-11-01

panda|zxiangfu: morning!03:11
kristianpaulha is amazing i couldbuy a 1.2V regulator locally..04:03
kristianpaulperhaps i could use a diode ;)04:03
wolfsprauldvdk hey David, still sometimes following here?04:05
wolfspraul(he used to read the backlog, so I try :-))04:05
wolfspraulI spent a whole day the other day trying to track down what colorforth and OKAD and those ga144 chips are really about, but couldn't...04:05
wolfsprauldo you know more?04:05
wolfspraulit looks like a refusal to reuse/use any other software, resulting in 'no source code' and in the end just some web pages at colorforth.com ?04:06
wolfspraulor is there more?04:06
wolfsprauldoes OKAD run? are there OKAD source codes?04:06
wolfspraulare the chip designs for those GA144 chips published? do they exist anywhere but in Chuck's brain? :-)04:06
wolfspraulwe have to track him down in his cabin in the western midlands somewhere? :-) 04:07
kristianpaulthinking about the next forth single porpuse  computer (after the wikireader?) 04:07
wolfspraulI was comparing vlsi cad software, came across electric, magic, alliance and toped. and the elusive OKAD...04:07
kristianpaulahmm04:07
kristianpaulok04:07
wolfspraulnot really, I just want to compare what's around04:07
kristianpaulsure sure04:07
kristianpauli'm to new for forth, but i remenber mcus optimized for java :p04:11
wpwrakhmm. http://www.ultratechnology.com/okad.htm04:11
wpwrakwas unhappy with the existing tools. too bulky and don't do what he wants to do. so far, that sounds familiar.04:12
wpwrakthen he built his own operating system, "OK". with one thing running on it, that OKAD. erm, what was that about the cabin again ? :)04:13
kristianpaulcome on, give it a chance :)04:14
kristianpaulyou could had write lilo in forth ;-)04:15
wolfspraulIs there someone here who can help xiangfu through the next step of debian bureaucracy?04:16
wpwrakat least i didn't "entered the kernel of [it] using a debugger "04:16
wolfspraulsome packages like fped (and fpgatools) need a 'sponsor', I believe with "DD" rights04:16
wolfspraulit's amazing to see how much bureaucracy there is, he he04:16
kristianpaulyes 04:16
wolfspraulthat is definitely worse than a lot of small country governments I dealt with04:16
kristianpaulDD <- bureaucracy, yeah... :-/04:16
wolfsprauland then they are surprised why the number of bug reports has declined by almost 50% in the last 6 years04:16
kristianpaulafaik my debian mentor is a bit away since two yeards ago... let me see what can i do04:17
wpwraki have no clue about debian bureaucracy. i think in this case ignorance is bliss :)04:17
wolfspraulbureaucracy04:17
kristianpaulperhaps he can join #debian-co and ask :-)04:17
wolfspraulthe problem with the bureaucrats is that all orgs are self selecting04:17
kristianpaulthere are two friendly DD there04:17
wolfsprauleventually only those types are left that start their computers to find typos04:17
wolfspraulhttp://mentors.debian.net/packages/uploader/xiangfu%40openmobilefree.net04:18
wolfspraulI see it quite fatalastic, same as I would deal wtih any US/China/Germany-type bureaucracy04:18
wolfspraulthen it's fun, no problem04:18
wolfspraulhow many years are we trying to get fped uploaded now?04:18
wolfsprauljust one or two04:18
wolfspraulpah04:18
wolfspraulnothing in the bigger picture of the universe04:18
xiangfuwpwrak, Hi 04:18
wolfspraulpeople are born, retired, whatever04:18
qi-bot[commit] Xiangfu: modules: bga: add 324 footprint (master) http://qi-hw.com/p/kicad-libs/87db42504:19
wolfsprauland fped is in status XYZ... ;-)04:19
xiangfuwpwrak, I am create the csg324 chip footprint. but I have problem on the 'row', you can see that in (this^) commit.04:19
wolfsprauloops http://www.donarmstrong.com/posts/bug_reporting_rate/04:20
wpwrakwolfspraul: on my ubuntu, fped is available as a package. so the/some result is there04:20
wolfspraulgood :-)04:20
wolfspraulkristianpaul: thanks!04:20
kristianpaul#debian-co <--- debian irc servers04:21
wolfspraulI believe the link I posted is where the "sponsors" can assert their powers04:21
xiangfuwpwrak,  fped is real great. once I have the 484 footprint. we you edit ~10 variables . then we get csg324. 04:21
kristianpaulwell he needs ask anyway04:21
wolfspraulpabs3: do you know anyone?04:21
kristianpaulhe if join debian.co i cna try convice DD's04:21
wolfspraulwe get there04:21
wolfspraulbtw, after some X years xiangfu has "dm" power now?04:22
wolfspraulxiangfu: was it "dm"?04:22
wolfspraulWOHOOO!04:22
wolfspraulDEEE EMMM04:22
wolfspraulthat is it04:22
pabs3wolfspraul: know anyone?04:22
wolfspraulsomeone who could help with the 'sponsor' thing?04:22
wolfspraulhttp://mentors.debian.net/packages/uploader/xiangfu%40openmobilefree.net04:22
xiangfuwpwrak, it's old package. I have a new version here: 'https://mentors.debian.net/package/fped'. and now I am a DM. so I need a DD upload a new version with one line add (DM-Upload-Allowed: yes)04:23
xiangfuthen I can upload/update the package by myself.04:23
wolfspraulDM!04:23
kristianpauldm is okay for fped04:23
xiangfuwolfspraul, yes. it's dm (debian maintainer)04:23
kristianpaulbut he needs sponsor for fpgatools04:23
wolfsprauleasier to get a US green card04:24
kristianpaullol ;)04:24
wolfspraulit is04:24
wolfsprauldebian can be proud04:24
wolfspraulthey are so big they are approaching the bureaucracy of the largest country on earth04:24
wolfspraulthat must be a sign of victory!04:24
kristianpauluniversal victory :p04:24
xiangfukristianpaul, fped, it still need a DD upload a new version with one line add (DM-Upload-Allowed: yes)04:24
pabs3xiangfu: FYI, DMUA is deprecated: https://lists.debian.org/debian-devel-announce/2012/09/msg00008.html04:24
kristianpaulxiangfu: aahmmm i tought was un-reviwed... :/04:25
pabs3kristianpaul: xiangfu has a sponsor for fpgatools already04:25
wolfspraulyup04:25
pabs3kristianpaul: http://lists.debian.org/20121009121100.230230@gmx.net04:26
kristianpaulah good !04:26
pabs3looks like Jan Luebbe <jluebbe@debian.org> sponsored fped04:27
wpwrakxiangfu: (row) problem. ah yes, messy. best to move the index outside the table04:27
wolfspraulthat is a wonderful mail there (new DUMA thing)04:28
wolfspraulI stand by my bureaucracy observation04:28
pabs3btw, there is an electronics software packaging team that might be interested in helping: http://wiki.debian.org/PkgElectronics04:28
wolfspraula change in US green card processing could not be any more intricate04:28
xiangfuwpwrak, you DO not sleep. it's 1AM  Buenos Aires :-)04:28
wolfspraulof the 5 "advantages" listed there one is "DM can't give another DM upload rights for his package anymore"04:29
wolfspraulthat's an advantage?04:29
wolfspraulanother bottleneck created04:29
wolfspraulgreat :-)04:29
wpwrakxiangfu: hehe ;-)04:29
wolfspraulthe pope himself should sign (and stamp!) a holy letter before any bit moves at all!04:29
xiangfuwpwrak, the first/last/inner frames is useless for bga484/324 right? is that for further use? like some BGA don't have balls at center.04:30
wolfspraulthe stamp is only valid if stamped exactly at midnight every year Easter Sunday04:30
wolfspraulmakes it safer!04:30
wpwrakwolfspraul: removal of one of the remaining loopholes to avoid bureaucracy :)04:30
xiangfuwpwrak, (best to move the index outside the table) please give me more advise. :) what about those rname?04:30
wolfspraulpabs3: what if a sponsor looses interest, drags his feet or becomes unresponsive?04:30
wolfspraulcan you 'unsponsor' a package to look for a new sponsor? can there be two sponsors?04:31
wolfspraulcan a sponsor marry another sponsor?04:31
wpwrakxiangfu: there are a few things wrong there ... fixing ... just watch :)04:31
Action: pabs3 gets bored and goes to watch tv04:31
wolfspraul:-)04:31
qi-bot[commit] Werner Almesberger: modules/bga.fpd: make "row" a loop and use the "row" table only for lookups (master) http://qi-hw.com/p/kicad-libs/e5c8bd904:32
wolfspraulon a serious note, does it matter that Debian's bug activity is 40% less now than it was 6 years ago?04:32
wolfspraulmaybe it's becoming so mature we should just consider the whole thing as more or less stable...04:32
wpwrakthis one gets the geometry right. now the Nr-plication of stuff that's needed only once ...04:32
wpwrakwolfspraul: yeah, meaningless statistics04:33
wolfspraulI was surprised about the magnitude of the decline, although it is quite obvious that the whole scene becomes more and more sleepy04:33
wolfspraulwpwrak: how so?04:33
wolfspraulpeople report bugs differently? not at all? there are less bugs?04:33
wpwraktime is an insufficient correlator. well, you can use this sort of statistics to plan the disk space for the bug database04:33
xiangfuwpwrak, ?row. great. we do need more documents on fped. :-D04:34
wpwrakxiangfu: isn't it self-explaining ? ;-))04:34
wolfspraulbugs per second over 8 years is 'insufficient'?04:34
Action: xiangfu have csg324 footprint now. ready to design some PCB for csg324.04:34
wolfspraulthe bug tracker is used less, for sure04:34
wolfspraulanyway we won't stop there bigger trends here04:35
wolfspraulthese04:35
wolfspraulI won't go watch TV, but enjoy some sunshine outside :-)04:35
qi-bot[commit] Werner Almesberger: modules/bga.fpd: don't draw outline Nr time. Once is plenty. (master) http://qi-hw.com/p/kicad-libs/ef5fb2004:41
wpwrakxiangfu: now the looping is also a bit better. and yes, inner/outer aren't used. not sure what the idea for them was. (bga.fpd comes from adam)04:42
wolfspraulso my vlsi cad comparison led me to toped04:42
wolfspraulhttp://www.toped.org.uk04:43
wolfspraulwill play with that a little04:43
wolfspraulmaybe I can abuse it as my visual fpga editor :-)04:43
wpwrakthere are some BGAs that have a "stripe" pattern but then, they usually don't assign column numbers to unused column. there are bgas with a empty square at the center, but then you'd also need to skip "inner" columns. so may be this just wasn't finished.04:43
wolfspraulthe other tree I looked at were magic and electric, which seemed to be very old (from the 80's) and without much new development04:43
wolfsprauland alliance which is newer (some french university), but also it seems recently they stopped developing/funding?04:44
wpwrak"very stable" ;-)04:44
wolfspraulelectric was sponsored by sun but that seems to have stopped around 200904:44
wolfspraultoped seems like a really new and fresh thing wiht active development (though just 1 or 2 people mainly, I think)04:45
wolfspraulso I start there...04:45
wpwraktoped looks promising. i wonder if the thing about the scripting language means that you get human-readable data files as well (like it is the case with fped)04:53
xiangfuwpwrak, (bgg "inner") ok. I think I will just remove it now. if we needs some bga have a empty square at the center. then we think about it.04:53
wpwrakxiangfu: sounds reasonable04:54
wolfspraulprobably not [human readable]04:55
wolfspraultheir native file format is some sort of binary database04:55
wolfsprauljust keep in mind that the size and efficiency of those files is paramount04:55
wolfspraulthat was the main reason for the switch from GDSII to OASIS in recent years04:56
wolfspraulwhen you are dealing with billions of transistors, 'human readable' may just not be that important04:56
wolfspraulhttp://www.semiwiki.com/forum/content/445-dawn-oasis.html05:00
wpwrakhmm ... if they're really that large, then binary would make sense05:00
wolfspraulnice article about the move from gdsii to oasis05:00
wolfspraulit would be nice to have a way to export to a human-readable format or otherwise make the data accessible to line-based tools05:03
wolfspraulbut that's for that specific purpose then, and for other purposes you need other optimizations05:03
wolfspraulit does make a difference whether the file I send around to a manufacturer is 2 GB or 20 GB...05:04
wpwrakyeah. if things have to be that big, then that becomes an issue05:04
qi-bot[commit] Xiangfu: bga.fpd: remove useless frames (master) http://qi-hw.com/p/kicad-libs/3d3be9209:26
dvdkwolfspraul: when it comes to OKAD, i think there at most a handful of people who know how to operate it.10:14
dvdksame with the GA144 "dev-kit" (i.e. colorforth)10:15
dvdkI even installed a free (original) version of colorforth under linux, but it's impossible to use.  The source-code is mostly in assembler, with a little color-forth code put on top.  Very ugly and undocumented (at least the asm part).10:16
Guest55383dvdk: that was roughly what I thought10:52
Guest55383oops, nick... one sec10:52
wolfspra1ldvdk: but isn't that a pity? so there is all this talk about how great it is, but then there is nothing really...10:55
wolfspra1lthere are some strange colorized assembler codes for DOS (!) which may or may not run under early-90's 'Windows'? urgh...10:55
wolfspra1las for OKAD? it exists? or is it just a myth? I still don't get it10:55
wolfspra1l'it' is something as tangible as "source code"?10:55
wolfspra1lit seems totally lost, this whole thing. I shall continue with toped unless I feel some boost of enlightenment around OKAD one day :-)10:55
dvdkOKAD does in fact exist.  You just need some form of autism to being able to operate it.  And I don't think that you can buy it, nor that there is much written documentation.10:56
wolfspra1lthe download of all sources from digitalarrays comes with a disclaimer that their software is only allowed to run on their ga144 chips10:56
wolfspra1lanother step 20+ years back10:56
wolfspra1land that is supposed to be "public domain"?10:56
wolfspra1lman man :-)10:56
wolfspra1lthey should really all just retire to some cabins in Montana...10:57
dvdkik okad public domain?  i know that colorforth is.  But okad is like 1 MB of (compressed ) color-forth source, I never saw a copy online.10:57
wolfspra1lah ok10:57
wolfspra1lwhy not?10:57
wolfspra1lsomething strange about this Interweb thing?10:57
dvdkdid you read this stuff about colorforth http://www.colorforth.com/cf.htm10:58
wolfspra1lthe ga144 chip was 'designed' in/with OKAD?10:58
wolfspra1lyes I read most of the colorforth.com/... pages10:58
wolfspra1lthat's the most tangible stuff I could find10:58
dvdkah, ok.10:58
dvdkso I tried to understand the few colorforth sourcees that come with colorforth.10:58
dvdkIt even claims to have a FAT export to floppy disk.  Which is required to output those mask files okad generates.  10:59
wolfspra1lI knew I had to ask you to find someone who went deeper into this path of insanity than I would have wanted to :-)10:59
dvdkI somewher had some info how it operates.10:59
wolfspra1lFAT floppy disks...10:59
wolfspra1lis everything OK with those people?10:59
wolfspra1lhttp://www.greenarraychips.com/home/support/download-02a.html11:00
dvdkIt's like: you have to create an enmpty file of size XXX on the floppy disk using windows, then colorforth's "file output" routine can dump a blob of data there.11:00
wolfspra1l"software may only be used with chips manufactured by GreenArrays"11:00
wolfspra1lcan I dismiss all this stuff as "crazy stuff of some old hippies" and move on and forward with the real world?11:00
wolfspra1lor is there more to it?11:01
dvdkcolorforth even has a (approx 20-line) implementation of a "png export".  also it generates a fixed header that amounts to no compression (like a pseudo huffman table, dunno), then dumps the binary data in a way that it looks like a png to an average png reader.11:01
viricthe forth world11:02
wolfspra1lany reason why OKAD was never uploaded/published?11:02
dvdkIf you are a friend of extreme minimalism, than OKAD and CF may be interesting.  But you want to build real-world chips, with real registers and cache and mmu etc., that's somewhat incompatible with the OKAD and CF philosophy11:02
wolfspra1lis all this stuff serious in any way?11:02
wolfspra1lhonestly it strikes me as mostly some form of insanity11:02
dvdkwolfspra1l: well, they did design an asynchronous multi-core CPU with it that actually works, so judge yourself :)11:02
viricit's a bit like the Amish.11:02
wolfspra1lwell11:02
dvdkbut that CPU has only like 3 registers and 128 bytes of RAM.11:02
wolfspra1lwhat software runs on those chips, and when?11:02
wolfspra1lin reality, not as yet another 'grand plan'11:03
wolfspra1land who is using them?11:03
dvdkthese chips implement something like a few color-forth-primitives in hardware.11:03
wolfspra1land what's the point/roadmap?11:03
wolfspra1lsure11:03
wolfspra1lself-referential as in any good religion, no? :-)11:03
wolfspra1la is b because b is a11:03
wolfspra1llet's get drunk now11:03
viric:)11:03
dvdkso looks like forh-code when you write assembly code.  they have a color-forth based "IDE" that allows you to input (a subset of) color-forth code that's turned into binary code for the GA144.  They even have an emulator etc. 11:04
wolfspra1ldvdk: come on you are real? what is the real thing one can expect out of this - today or tomorrow?11:04
wolfspra1lor it just leads nowhere?11:04
dvdkwell, self-referential sure.  it's like a second evolution that's not using Carbon but Silicon.  Kind of an alternate eco-system.11:05
viriclots of closed designs around that, though11:05
wolfspra2lok but is that stuff even open source?11:05
wolfspra2lthe okad 'sources' are not published?11:06
viricwolfspra2l: all that philosophy is pre-FSF I think :)11:06
wolfspra2lgreenarraychips.com software downloads are "unless specifically noted otherwise" only allowed to run on their chips11:06
wolfspra2lwhy would I even bother with the dysfunctions of those guys... and try to find some crazy couple dozen "lines" of source codes that maybe can write FAT floppy sectors?11:06
wolfspra2lI don't get it :-) I just don't...11:07
dvdkcolorforth is, the GA144 is pretty well documented.  but programming it feels like programming an FPGA by manually filling LUTs and setting the routing matrix.11:07
virichaha11:07
wolfspra2ldo you have one or programmed it?11:07
dvdk(like 123 bytes ram per core is somewhat like a bug logic slice :)11:07
dvdks/123/128/11:07
viricyes11:07
wolfspra2ldvdk: what is a good colorforth download url?11:08
dvdkno, but I read about guys who did program, I read some programs and I tinkered around with colorforth (http://mosquito.dyndns.tv/opensvn/free/trunk/forth/glcolorforth/)11:08
virichttp://www.colorforth.com/install.htm     '2001 version'11:08
wolfspra2lviric: *download*11:09
wolfspra2lin that world, there's a lot of *talk* about download, not so many things you can actually download11:09
Action: wolfspra2l checking david's link...11:09
dvdkno, on linux try my version.  chuck's version only runs on bare (old) hardware from floppy, other linux ports have been abandoned11:09
virichaha11:09
whitequarkcolorforth...11:09
whitequarkit seems for me that everyone who uses forth becomes a little bit brain-damaged11:09
dvdkso I had to do some hacking to make it compile&link on 64-bit linux.11:09
whitequarke.g. TeX11:09
whitequarkit's absolutely awesome11:09
whitequarkbut so freaking alien11:09
viricwhitequark: I've used a bit of forth, but got out on time ;)11:10
dvdkwolfspra2l: you may want to try to download the official GA144 SDK.  That's build around a much newer non-public domain version of colorforth.  It even has QWERTY layout and can run from within windodws :)11:10
dvdk(but never tried it)11:10
viricI've a friend that wrote parts of retroforth... (clisp implementation, among others)11:10
dvdkbut then I think you'd still waste your time. 11:11
wolfspra2l"non-public domain" what does that mean?11:11
viricdvdk: qwerty support? does it require an XT keyboard? :)11:11
whitequarkviric: it's an interesting _concept_ indeed. I enjoyed reading about a "functional Forth" implementation11:11
whitequarkviric: but if you're going to use it in production you're probably insane11:11
dvdkwolfspraul: what i wanted to say: colorforth=public domain, GA144 SDK = closed source11:11
wolfspra2lah then why would I bother :-)11:11
dvdkbut both based on color-forth, where GA144 SDK is much much newer11:11
viricwhitequark: I always thought of Forth as a language for programming AB8@0;L=K5 <0H8=K11:11
wolfspra2lthere must be thousands of one-off proprietary this-and-that chips/archs in the world11:11
wolfspra2lamong those that would be one of the stupidest to explore, most likely11:12
wolfspra2lbut your link is great, finally11:12
viricwolfspra2l: work better for an antropologist, I think11:12
wolfspra2lthanks!11:12
wolfspra2lthose few files look like something real, and I know there is probably not much real in all this11:12
wolfspra2lso this must be it! :-)11:12
whitequarkviric: hehe11:13
whitequarkyou're not too far off, it's used in PostScript printers11:13
whitequarkas PS is basically forth11:13
wolfspra2ldvdk: do  you think the ga144 has much future?11:13
dvdkwolfspra2l: I wouldn't dismiss thit as stupid.  It's just alien to what we're used to deal with.  As i said, an alternative eco-system that's incompatible to ours.11:13
wolfspra2lyes but wants to be incompatible too11:14
viricwell, all those stack based things... RPN in calculators, forth, ... even jvm is a similar thing.11:14
dvdkwolfspra2l: ask the marsians (or alpha centaurians) about GA144's future11:14
wolfspra2lincompatible to anything that exists today or may exist in the future11:14
wolfspra2land that's where it gets strange11:15
wolfspra2lso the fact that this tech serves very few/no real people is almost by design11:15
wolfspra2lat which point I reserve my right to just not care anymore...11:15
dvdkwolfspra2l: a niche is a niche. 11:15
viricdvdk: there are also the hardware java virtual machines (no more virtual?), or the lisp-on-hardware, ... there are plenty of hw/sw places far from c-programmers. :)11:15
wolfspra2ldvdk: yes :-) and that's an extreme one :-)11:15
dvdkas I said, think of it as silicon-based life.  it just can.t thrive on earth, and ther won't be a place wher it can be a competitor to us carbon-based critters11:16
wolfspra2ldo you know anyone who has one of those ga144 boards or is programming on it?11:17
wolfspra2lare any 'source codes' of such software being published anywhere?11:17
wolfspra2lgod forbid on something as ...11:17
wolfspra2lI hardly dare to say it11:17
wolfspra2l... github11:17
wolfspra2l? :-)11:17
viricdvdk: since those chips have been released years ago... there should be some writings about their impact, is there?11:17
dvdkyes. not personally.  a few on comp.lang.forth usegroup write about it.11:17
viricdvdk: remove the dates from comp.lang.forth random posts, and try to guess the decade.11:18
viricdvdk, wolfspra2l: http://prog21.dadgum.com/128.html11:19
wolfspra2lI would be interested in what software runs on them, and source code of that of course11:20
wolfspra2ldvdk: thanks a lot for your insight and source link!11:20
wolfspra2lthat boosted the level of what is 'real' by about a factor of 10 for me :-)11:20
viricI always remember that post, when I see forth reviving again in some room. :)11:20
dvdkwolfspra2l: there's a colorforth google-group i just found out.11:20
wolfspra2lyes but I just wanted to make sure that there wasn't this vibrant OKAD github community somewhere, with dozens of people contributing and taping-out last-gen chips etc.11:23
wolfspra2lwhich of course there isn't, so I'm done :-)11:24
wolfspra2lI have a pretty clear idea/imagination of what that OKAD thing actually is/was11:25
dvdkviric: no, the original color-forth implements kind of a dvorak-layout that uses only about 30 keys, other functions are activated by cycling layouts.11:26
dvdkah, and it's not ASCII, it's a proprietary 64-char character set that's usually huffman-compressed in chunks of 32-bit words 11:27
dvdkAFAIR the original version even didn't have separate 'i' and '1' glyphs.  why bother when they *look* the same :)11:27
wolfspra2lI shall continue with toped...11:27
dvdks/glyphs/charcodes/11:27
wolfspra2lagain, thanks a lot for the feedback! that is really hard in that scene to find anything that relates to something real...11:27
wolfspra2lit just makes you search harder and harder11:27
dvdkno prob.11:28
dvdka lot of people are excitedly waiting for fpgatools to become less vaporwareish11:28
dvdk, btw11:29
viricdvdk: haha   typing machines had 'l' and '1' the same, not 'i' and '1'. :)11:33
viricdvdk: why think of another character-number coding, if we already have EBCDIC?11:34
viricbtw, I always wanted to program the Saturn chips of HP 48, that I heard had 20-bit words with 5-bit nibbles.11:35
wpwrakviric: pre-FSF or pre-gnu (those with the horns) ? :)11:49
viricI think the gnu with horns was first though, yes. :)11:52
wolfspra1lyes fpgatools has a steep curve now11:53
wolfspra1ldid some more directional wires, vertical and horizontal11:53
wolfspra1lnow the diagonal ones11:53
wolfspra1lthen lots more clock routing, bufgmux, and and and11:53
wolfspra1l:-)11:53
wolfspra1ldvdk: is there software for the ga144 that is open source/published anywhere?11:54
dvdkwolfspra1l: do you mean dev-kit or software running *on* the GA14411:54
dvdkhere is e.g. a md5 hash algorithm implementation for GA144. 11:56
dvdkhttp://www.greenarraychips.com/home/documents/greg/AN001-120510-MD5.pdf11:56
dvdkyou'll see it's more FPGA-like than CPU-like11:56
wpwrakwolfspra1l: colorforth and github ... wouldn't they have their own revision control system, completely alien to anything we know ? and of course, not communicating via TCP/IP. maybe some highly efficient DIY protocol over 300 bps modems, to go with the floppies.12:00
wpwrakhmm. an application note. implementing a well-known hash (nothing DIY). i smell heresy :)12:02
lindi-wolfspra1l: should blinking_led.c already work?12:02
viricwpwrak: what makes you think they use a RCS?12:02
dvdkwpwrak: they have no text-file format that you can read, so you can only commit thnem as binary blobs12:03
dvdkso that means yes12:03
dvdkI also read that htey think about doing version control from within colorforth12:03
dvdkwpwrak: about the app-note: don't worry it's the only GA144 app-node of it's kind :)12:04
wpwraki think they learned their lesson from the internet. remember how great and full of freedom it was before big business and governments noticed it ? now that would be a way to make sure those old enemies never even get close :)12:05
viricif even nodejs could become popular, I think Forth could achieve something similar, once mastering the art of pastel-color CSS in Helvetica web pages.12:05
viricpastel colors and helvetica are the most determining factors to a broad success of a project, I think. :)12:08
viric- I feel a bit too nihilist today12:08
larscand I always though blink tags were the way to success. damm, had it wrong all these years, but that may explain why my projects never took off ;)12:17
wpwraklarsc: you mean you didn't put enough emphasis on obscurity and the arcane ?12:24
larscwpwrak: I sure did! The projects are hidden 5 paths deep on a obscure domain name. And I think I only gave the link to a handful of people12:26
larscOh and the page only renders correct with firefox 0.97-0.9912:26
wpwrakyou put it on the web ? i keep my stuff on a mailbox, only accessible via direct dial-in.12:28
wpwrakdownload via my own enhanced (incompatible) variant of kermit.12:28
larsc_I_ put it on the web, indeed.12:29
wolfsprauldvdk: sorry was away and lots of disconnects. do you think it may make sense to try to run the mini-cpu from inside the g144 in an fpga?12:43
wolfspraulI think the instruction set is documented - but is it usable for anything at all?12:43
wolfspraulI am looking for some small test-'cores' for fpgatools, and a big cpu like lm32/openrisc may still be too much for me at the beginning12:44
wolfsprauldo you have any experience with that cpu, can you recommend it at least for further study?12:44
wpwrakhow about making your own mini-cpu ? the basics are very simple.12:45
wolfspraullindi- no, not working yet12:46
wolfspraullots of detours, a little depressing but I try to just steer right through12:46
wolfspraulI have to spend more time around the clocking stuff first12:46
wolfspraulI'm a little anti-forth here, I guess. if there is something I can reuse, why not reuse it?12:46
wolfspraul:-)12:47
wpwrakany "real-life" core is likely to be a lot more complex than the minimum you need for demonstration purposes.12:47
larscmaybe use the navre core12:48
wpwrakof course, the first question should be what this core would actually be expected to do. blink a led ? control a robot ? run linux ?12:49
larschow about 'do something'12:50
larscit doesn't matter that much if the goal is to show that fpgatools is working12:50
wpwraksounds like a case for a core with two registers, maybe 8 bit word size. 8+something instruction size. harvard arch.12:53
wpwrakor maybe accumulator plus register array is easier12:57
wpwrakadd Ri: A = A+Ri12:58
wolfspraulanything good about MISC? http://en.wikipedia.org/wiki/Minimal_instruction_set_computer ?12:58
wpwraksame for and, or, xor.12:58
wolfspraulor maybe I search the other way round and look for interesting/new llvm backends, since I don't want to be stuck with something where no sw runs...12:59
wpwrakmmh, that would put you into the feature-rich core segment12:59
wpwrakif you;re looking for something that just gets "something" done, but you have to program it in assembler, then you can make a super-simplistic architecture13:00
dvdkwolfspraul: just coming back from lunch.13:15
dvdkwolfspraul: the GA144's mini CPU is an asynchronous design: no clock.  don't think this will easily port to an FPGA.13:15
dvdkHowever, there is a core that follows very similar design ideas, developed by a german forther.  The B16 core, with a even smaller "b16-mini" version13:16
dvdkhttp://bernd-paysan.de/b16.html13:16
dvdkit's GPL.  Uses a very similar, highly efficient instruction packing that allows to build large programs with very little instruction RAM13:17
dvdkoriginally this was developed for a real ASIC for the company Bernd was working for, then the FPGA version open-sourced.  The documentation describes various ASIC-specific optimizations like replacing registers with latches.13:18
dvdkthe compiler is written in Gforth, and the assembly code reads like normal forth code (i.e. not color-forth, and much saner)13:19
dvdkthere is an even simpler Stack-based core called J1, but it's a 32-bit architecture and needs moer instruction RAM:13:21
dvdkhttp://www.excamera.com/sphinx/fpga-j1.html13:21
dvdkit's open source, and the forth-written software stack for this thing supports even tcp/ip13:21
dvdkthis may be easier to work with than B16, architecture is simpler.  13:22
dvdkno instruction decoder at all, instruction word directly switching RTL data flow13:22
dvdkoops, mistake, J1 is also a 16-bit CPU just like B1613:23
lekernelwolfspraul: are you in Germany near the end of the year?13:24
wolfspraulgreat links, thanks!13:26
wolfspraulwhich of those is a good starting point for getting whatever 'other' software to run?13:26
wolfspraulwhy would an asynchronous cpu design not work in an fpga?13:27
lekernelbecause it has plenty of D flip flops and global clock networks which would go unused, resulting in a quite suboptimal design13:29
lekerneland the vendor software doesn't have good support for asynchronous stuff13:31
wpwrakwolfspraul: what's the objective for implementing a core ? it sounds like a bit of a detour to me, at least as far as gaining knowledge is concerned.13:33
wolfsprauloh sure, no worries. I am not doing this now. the point is to guide me a bit through the really large amount of features in the fpga13:34
wolfspraulwithout focus, I can spend years there13:34
wolfspraulI am implementing the blinking led now :-)13:34
wpwrak(years) hmm, sounds bad ... in an exciting way :)13:35
larscwell, software is never done13:36
larscit can only become obsolete13:36
viriclarsc: luckily gnu hello is maintained to avoid becoming obsolete.13:37
wolfspraulas the links from david show there may well be some nice & tiny experimental cores out there13:37
wpwrakyeah, the software life-cycle. planning (optional), first prototype, alpha/beta version (repeat as many times as necessary), abandonment.13:37
wolfspraulthat may be a better first step or focus generator at some point than a larger core13:38
wolfspraulmay13:38
wolfspraulI'm just happy I can read a bit through those cores and think13:39
rohmorning14:14
wolfspraulgood morning14:14
kristianpaulwolfspraul: small cpu like lattice mico8?14:40
kristianpaulmorning14:40
wolfsprauldvdk: ok I looked briefly at the b16 and j1 pages, will read more tomorrow14:48
wolfspraulwhat is the future of those projects?14:48
wolfspraulany plans you know about?14:48
wolfspraulor are they basically finished?14:48
wolfspraulb16 looks older (2002-2005), j1 more recent (2010). can't immediately tell other differences in stated goal or roadmap (if any)14:49
wolfspraul200 lines of verilog (for the j1) sounds good14:50
wolfspraulthat gives me plenty of lines to play with (in how they translate into the fpga...), which is all I am looking for right now14:50
wolfsprauland there is a url right there on his page, what a relief after the colorforth day :-)14:51
wolfspraulhttp://www.excamera.com/files/j1demo/verilog/j1.v14:51
dvdkwolfspraul: b16 is more a CISC processor and J1 more a RISC-style processor 15:08
dvdk(bernd would disagree, actually b16 is not cisc-like, but compared to the J1 it does more instruction decoding and has lower throughput)15:08
dvdkso b16 is more for controller applications, and J1 can do more media processing (like moving data from camera to network interface)15:09
dvdkb16 is to J1 like ARM to MIPS15:09
dvdkI think both are somehow used commercially: B16 in ASIC (with FPGA version used for rapid prototyping), while J1 is FPGA-only, AFAICS15:10
dvdkBoth use Gforth-written software for software development (assembler/compiler)15:11
dvdkI guess b16 is harder to grok, it's has more optimization quirks that are difficult to follow, while J1 is very straight-forward15:11
dvdkIf you want to know stuff about ASIC-targetting design, then the B16 documentation / papers will most likely be interesting to read15:12
kyakwpwrak: could you please give a brief status of atusb support in kernel? I'm looking to build it as a module (for linux-3.6) and package it for for my distro (using dkms)18:54
kyaki'm looking at instructions in ben-wpan/install/INSTALL-PC and wonder if they are still valid18:54
kyakfor kernel part, that is18:54
kyakbtw, they are not valid any longer for user space tools. lowpan-tools are now ported to libnl v.3, so libnl1 is no longer needed18:55
kyakanyway, user space is not bothering me at all.. i already packages the lowpan-tools.. Now wondering how do it with kernel. My stock kernel has some of the mentioned options as =m, and some are missing18:57
wpwrakkyak: i haven't touched it in a very long while. i think more than a year18:58
kyakthe missing options i guess are provided by patches in qi-kernel.git. Makes me wonder if ti possible at all to avoid rebuilding of stock kernel18:58
kyakheh, the "software life cycle" you mentioned earlier :)18:59
viricI thought the same :)18:59
wpwraki think xiangfu did some integration for more recent kernels. so that should be in the qi kernel.18:59
kyakyeah, i remember he did things for a router with usb.. i wonder how intrusive this integration is19:00
kyaki'd like to have atusb.c and atusb.h :)19:00
kyakand then simply atusb.ko19:01
kyakok.. thanks anyway19:01
wpwrakthe integration may merely be tracking of the linux-zigbee work. that provides the WPAN infrastructure.19:02
hellekinhola... Can you guys beat that? http://oggstreamer.wordpress.com/2012/10/25/oggstreamer-production-costs-a-rough-estimation/23:55
wolfsprauldvdk perfect, that is very helpful!23:58
--- Fri Nov 2 201200:00

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