| --- Log opened Tue Oct 27 00:00:43 2015 | ||
| olofk | poke53281: Time to rewrite jor1k :) https://gitlab.brokenpipe.de/stettberger/avremu/tree/master | 10:21 |
|---|---|---|
| wallento | haha | 12:09 |
| robtaylor | just seen how cheap the lattice pcie card is http://www.digikey.co.uk/product-detail/en/LFE5UM-85F-PB-EVN/220-1916-ND/4976291 | 12:10 |
| robtaylor | would there be any sensible approach to talking pcie with the yosys toolchain? | 12:11 |
| LoneTech | yeah, "cheap" with a limited-time offer to try the tools. last I checked it was about $10k/y if you wanted to keep using it. | 12:11 |
| robtaylor | hence the question ;) | 12:11 |
| wallento | I would assume you need some soft IP for PCIE | 12:11 |
| robtaylor | mm, looks like i'd have to purchase Lattices :( | 12:17 |
| ysionneau | olofk: omfg.... | 12:18 |
| olofk | robtaylor: I think that might be the card that maxpaln used to implemented his OpenRISC SoC | 13:13 |
| olofk | robtaylor: yosys is only for the iCE40 chips. Not ecp5 | 13:14 |
| olofk | Well, actually yosys is the synthesis part, so what I just said is wrong | 13:15 |
| olofk | But we only have P&R for iCE40 | 13:15 |
| wallento | is there any estimate of complexity of ecp5 to ice40? | 13:16 |
| olofk | Lattice got iCE40 when they bought Silicon Blue a few years ago, so they are totally different products | 13:17 |
| wallento | in months of reverse engineering? | 13:17 |
| wallento | ah, okay | 13:17 |
| olofk | ecp5 is more like virtex/stratix | 13:17 |
| wallento | yeah, so I assume it gets ugly in the bitstream with clocks, I/O, etc. | 13:18 |
| olofk | But aren't thre any open source pcie cores? | 13:18 |
| olofk | yeah. That must be nearly impossible without real specs | 13:18 |
| wallento | RIFFA is a pcie core | 13:19 |
| wallento | I don't remember what layers it covers | 13:19 |
| wallento | ah, it is not pcie but only the communication | 13:20 |
| wallento | channels to fifos | 13:20 |
| wallento | it needs xilinx IP clocks | 13:21 |
| olofk | Yep. Saw that too just now | 13:21 |
| wallento | but they are free with vivado IIRC | 13:21 |
| olofk | But that won't help for lattice devices | 13:21 |
| wallento | just for abstraction | 13:22 |
| robtaylor | olofk: ah, thats good info, thanks | 14:16 |
| ysionneau | https://github.com/enjoy-digital/litepcie | 14:25 |
| robtaylor | ysionneau: nice! | 14:28 |
| ysionneau | doesn't support lattice/altera though | 14:30 |
| ysionneau | but I guess you can fund _florent_ for doing it :) | 14:30 |
| robtaylor | that reminds me question, what are people's thoughs on migen? | 14:30 |
| ysionneau | I've been using it for more than 1 year now | 14:30 |
| ysionneau | and it's great! | 14:30 |
| ysionneau | I don't want to touch verilog anymore | 14:31 |
| robtaylor | seems similar to the chisel approach | 14:31 |
| ysionneau | I guess | 14:31 |
| robtaylor | ie. you get a buncgh of primitives for building up hardware designs as a data structure | 14:31 |
| robtaylor | oh, maybe i'm misunderstanding - it isn't event driven?! | 14:32 |
| ysionneau | you just describe your synchronous/combinatorial design using python syntax/constructs | 14:33 |
| ysionneau | and you have Python syntaxic sugar + object oriented stuff as a bonus | 14:34 |
| robtaylor | how does it stack up against myhdl? | 14:34 |
| ysionneau | it's really different from myhdl from what I understood | 14:34 |
| ysionneau | but I can't say much I never used it | 14:35 |
| ysionneau | also you have to take into account the existing ecosystem : is there lots of useful cores in myhdl? chisel? | 14:37 |
| ysionneau | with migen you have already an entire SoC + lots of modules (FIFOs, wishbone bus, bus width converter etc) | 14:37 |
| ysionneau | DDR controllers, uart, spi flash controller | 14:37 |
| ysionneau | etc | 14:37 |
| ysionneau | pcie, ethernet, usb | 14:38 |
| robtaylor | there's a fair bit in myhdl, also you can wrap verilog cores in both | 14:38 |
| ysionneau | yep | 14:38 |
| robtaylor | but pcie and ethernet are quite cool | 14:38 |
| robtaylor | is there a community? | 14:39 |
| ysionneau | M-Labs -> #m-labs | 14:39 |
| robtaylor | hmm, that's a company | 14:39 |
| ysionneau | it started off as the Milkymist community, and it got incorporated | 14:39 |
| robtaylor | thats kinda a bad mark if it's an 'open product' model | 14:39 |
| ysionneau | but the corporation existence does not change anything about the community, does it? | 14:40 |
| ysionneau | making a corp does not change gold into shit instantaneously :p | 14:40 |
| ysionneau | anyway, m-labs is the channel where anyone interested in Migen/MiSoC can discuss | 14:41 |
| ysionneau | regardless of any commercial stuff | 14:41 |
| ysionneau | and btw, the pcie ethernet and usb cores are also made by a company ... -> enjoy-digital | 14:41 |
| ysionneau | the fact that several companies use Migen also proves it's usable :) | 14:42 |
| robtaylor | ysionneau: companies involved is good. project controlled by one company bad | 14:51 |
| robtaylor | marginally concerning if the comminuty channel is also the company channel, as a businessman myself | 14:51 |
| robtaylor | shows a lack of thought of govornance | 14:52 |
| ysionneau | what issue do you see with the company channel being the same as the community one? | 14:52 |
| robtaylor | ysionneau: it shows a lack of thought of govornance | 14:53 |
| ysionneau | I see more a concern about being totally open | 14:53 |
| robtaylor | it says 'this projet is this company's' and puts me off as a company | 14:53 |
| robtaylor | which is actually the reason i hadn't looked into migen earlier - all the info on it was on a corportation's page | 14:54 |
| robtaylor | If you have an interest in the whys and wherefores of open source govornance, I'd reccomend Jono Bacon's book | 14:55 |
| ysionneau | ok :o | 14:58 |
| robtaylor | ysionneau: i'm actally kinda saddened, as I'd like to consider using migen, but it'd be too big a risk commercially | 14:58 |
| ysionneau | actually, the project didn't change overnight when the main contributor incorporated | 14:59 |
| ysionneau | the source code is still the same | 14:59 |
| ysionneau | with same license | 14:59 |
| ysionneau | on the same github repo | 14:59 |
| robtaylor | code and license aren't govornance | 14:59 |
| ysionneau | are you referring to the directions taken by a project? | 14:59 |
| robtaylor | If i do anyting to promote it, I'd be paying to promote another compant | 14:59 |
| robtaylor | which is then a commercial linkage | 15:00 |
| ysionneau | what about 1-men companies? are they still companies? | 15:00 |
| ysionneau | like free lancers and such | 15:00 |
| ysionneau | anyone is a company basically ... or could become one in 1 week | 15:02 |
| robtaylor | if you are passing ownership to that company and that company is espressing controlling rights on the work, yep | 15:04 |
| wallento | what exactly is the problem with promoting another companies free and open source product? | 15:04 |
| robtaylor | its more if I'm building something on top of it | 15:05 |
| robtaylor | so less of an issue for, say, a core | 15:05 |
| robtaylor | than the language you're using (effectively) | 15:05 |
| robtaylor | its not a problem if there's a clear govornance model | 15:06 |
| robtaylor | but in this case there isn't | 15:06 |
| ysionneau | maybe you should talk about it to sb0 | 15:06 |
| wallento | okay, got it | 15:06 |
| robtaylor | - its about knowing the rules of the game - if the rules aren't clear then I need to worry about that companies actions | 15:06 |
| robtaylor | ysionneau: happy to do so, I've got quite some experience in this area ;) | 15:07 |
| robtaylor | (done it right and wrong...) | 15:07 |
| ysionneau | :) | 15:08 |
| sb0 | robtaylor, what is your question about "govornance" (sic)? | 15:10 |
| robtaylor | sb0: its not a question, more an observation | 15:10 |
| robtaylor | sb0: there isn't a clearly defined relationship between migen/misoc and m-labs | 15:11 |
| robtaylor | http://oss-watch.ac.uk/resources/governancemodels | 15:12 |
| sb0 | oh, yawn, red tape ... | 15:12 |
| robtaylor | sb0: I'm creating a startup, and currently feel I can't utilise migen as this isn't defined | 15:12 |
| sb0 | so what is the risk exactly? | 15:13 |
| robtaylor | sb0: i can't trust m-labs, and m-labs (appear to) hold all control | 15:13 |
| robtaylor | (i can't trust m-labs as I'm in no contractual relationship with them, not because they are untrustable) | 15:14 |
| sb0 | how is that different from, say, linux or worse, windows (if you're using that)? | 15:14 |
| robtaylor | linux is not controlled by one company | 15:14 |
| robtaylor | windows does indeed have exactly that problem, which is why folk stopped using it.. (at least in my world) | 15:14 |
| sb0 | what about the linux foundation? is that not a company? | 15:15 |
| robtaylor | well, not quite, you did have a contractual relationship with MS, and could ask for more | 15:15 |
| robtaylor | sb0: well, its a not for profit, but yes, it has a clear governance relationship with linux | 15:15 |
| wallento | I think governance models are rare in the open source hardware and EDA world | 15:16 |
| robtaylor | yep, its young | 15:16 |
| robtaylor | you haven't had enpough people and companies around to hit the problems yet | 15:16 |
| wallento | which is not a problem of age (its not that young actually) but of the structure, I think | 15:17 |
| robtaylor | true, its not that young, but the size of the overall community is small | 15:17 |
| wallento | yes, so its hard to find governed projects | 15:18 |
| robtaylor | yeah, you don't need it when gthings are small, you have similarly minded folk working towards common goals | 15:18 |
| robtaylor | and mostly its hobbyist and single-man companies providing services | 15:19 |
| robtaylor | (or small product companies) | 15:19 |
| sb0 | also, all this red tape is boring | 15:19 |
| wallento | sb0: its different when stuff grows | 15:19 |
| sb0 | plus, linux has its problems too. didn't Sarah Sharp complain about it quite loudly, recently? | 15:20 |
| robtaylor | sb0: indeed, and actually thats becasue Linux's governance model is pretty shoddy | 15:20 |
| robtaylor | its basically 'what Linus says' | 15:20 |
| robtaylor | better examples abound elsewhere | 15:20 |
| robtaylor | sb0: so is m-labs you? | 15:21 |
| sb0 | I founded it, yes | 15:21 |
| robtaylor | sb0: ownership wise? | 15:22 |
| robtaylor | sorry, public channel | 15:22 |
| sb0 | yeah, I don't have external investors | 15:23 |
| juliusb_ | robtaylor: when it's a one-man-band backing something that is all open source licensed and publically available stuff, I don't see the big deal; a formal governance about the development of that thing isn't necessary. Eg. if it goes the wrong way, you fork and have your own thing. No biggie. If it's got commits from more than half a dozen places, some of which are commerical, fine, you might need it, but before then (or somewhere around then) it doesn't matter right? | 17:29 |
| juliusb_ | I would argue it's a bit rash to avoid all projects which might be done by a few men who also have a commerical entity | 17:32 |
| juliusb_ | what about cocotb? Going to avoid that too? | 17:32 |
| juliusb_ | it's likely once you | 17:33 |
| juliusb_ | it's likely once you get big enough, you'll need something formal, but even then it's not obvious; are not there loads of projects where you might have a couple commerical contributors plus a community, and it's all worked out via a reasonably informal system of perhaps one man in charge, or a leadership council? What changes if you codeify that? | 17:34 |
| juliusb_ | an interesting area, anyway, and hopefully something the open source hardware area will need to sort out one day ;) | 17:38 |
| robtaylor | juliusb_: its more that it lives within a commercial entity at the moment | 17:54 |
| robtaylor | juliusb_: i'd raise the same concerns with cocotb :) | 17:54 |
| robtaylor | juliusb_: but indeed, at this early point one can go have conversations with the company owners if there are issues. Its good to be aware that if we're sucessful that causes problems (as we had with opencores and the toolchain) | 17:56 |
| poke53281 | olofk: No, with a tex port I can't offer a network. | 22:29 |
| --- Log closed Wed Oct 28 00:00:45 2015 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!