IRC logs for #openrisc Friday, 2016-02-19

--- Log opened Fri Feb 19 04:11:44 2016
-!- Irssi: #openrisc: Total of 44 nicks [0 ops, 0 halfops, 0 voices, 44 normal]04:11
-!- Irssi: Join to #openrisc was synced in 16 secs04:11
wallentohi all10:14
wallentoanyone has an interesting project idea as a summer of code project?10:15
wallentowe are applying as FOSSi Foundation, mainly around OpenRISC and any ideas that can fit a three month period are welcome10:16
mafmwallento: were you selected as orga?10:51
wallentono, we prepare the application10:52
wallentoand put up a preliminary list10:53
wallentoso if you have an idea it does not need to be fully elaborated10:53
mafmnot really, I am not very involved with openrisc nowadays (I was with the Debian port)10:54
mafmjust asking because somebody was asking yesterday to the lowRISC project if they had applied to GSoC as orga10:55
mafmI don't know if FOSSi includes lowrisc nowadays?10:55
wallentono, they are complementary10:55
wallentolowrisc included many more generic projects10:56
wallentoand sponsored extra projects10:56
mafmthe Debian port would be a good project idea for GSoC, if GCC was upstreamed -- there were people accepted in previous years to work on the port of mips64el, for example10:56
wallentoso stuff like fusesoc can better be done under the FOSSi umbrella we think10:56
wallentoGCC upstream is a mandatory, right? Problem is we do not progress on this part unfortunately10:57
mafmyeah, I know :(10:57
mafmIIRC gsoc projects for lowrisc accepted were porting seL4 to risc-v or similar10:58
mafmmaybe that with openrisc?10:58
wallentoyeah, I mentored the seL4 port10:59
wallentowill put it up for OpenRISC this year :)10:59
wallentowe will also continue with that for lowRISC11:00
wallentoMaybe poke53281 wants to do another fast simulator :)11:01
wallentois there anything in the Linux part that needs to be done11:01
wallentoI unfortunately lost track of this11:01
mafmI think that it was poke53281 the one asking about this yesterday for lowrisc, BTW, but maybe I mistake the number X)11:04
mafmGCC support for openrisc (e.g. a clean reimplementation from scratch, to get rid of the copyright assignment problems -- if the person agrees to it from the start) would be a possibility for a project, but maybe too hard for 3 months and students jumping into the boat11:06
mafmif you want openrisc to continue to go ahead and be useful, I think that GCC is quite important to have (LLVM is another possibility nowadays, but still)11:07
mafms/I think/I guess/ -- you'll know better than me :-)11:07
mafmI guess that keeping the support of {binutils,glibc,gcc,linux,qemu}, either upstreamed or in unmerged ports (but working with recent versions of upstream) is good to have, if one wants to get good support in the Linux ecosystem11:10
wallentoyep, agree11:11
wallentoa clean GCC implementation is probably too much for a student in 3 months11:11
wallentoespecially as we need to keep them away from the current code..11:11
mafmI was wondering if they could reuse or inspire themselves in similar implementations, like risc-v?11:14
mafmwallento: in theory, if risc-v is a clean-room implementation (not based on MIPS or others, so they own the copyright and don't owe anything to any other implementation), and they release their part with the BSD license (it's Berkeley, after all) or GPL (the code will be GPL when added to GCC upstream), maybe the GCC support for openrisc would just need to be modified11:18
wallentonice idea, MIPS is probably the easiest11:18
wallentommh, okay11:18
mafmI am not a lawyer! :D11:19
wallentois there any problem with deriving from the MIPS port?11:19
mafmmaybe patents?11:19
wallentoBut RISC-V is also fine11:19
mafmI am not 100% sure how these things work in practice11:20
wallentoI think someone from the compiler guys needs to pick it up as a mentor as the student must be a really good one to do this task11:20
wallentoI have some basic knowledge of compilers, but of no help in this :)11:20
simoncookI would imagine if starting a new GCC backend, deriving from one that is already upstream would work best, because FSF already have the copyright on that?11:20
wallentoyes, that makes sense11:21
mafmhmmm, yes, makes sense, I just don't know if there's some kind of catch that might be problematic11:23
mafmbut I guess that with GCC being GPLv3 and insisting on the part about patents, people implementing new backends would be clear of problems11:24
mafmso there's at least risc-v (even if not upstreamed, their code is on github and should have some free license, I suppose), superh (supported by gcc, with patents expiring) and mips (supported by gcc, also some older patents expiring) that should be similar to openrisc11:26
wallentoYep, it may make sense to start a GSoC on this11:27
wallentoit just should be a closed project11:28
wallentolike with a result11:28
wallentoso it is probably a basic port11:28
mafmyeah, and make sure to get a good mentor knowing about the topic :)11:29
wallentoyep, thats the most critical11:29
mafmand yes, having realistic and clearly defined goals is quite important for the students11:34
mafmand for the orgas it is better if the project is not overly ambitious and doesn't extract 100% of the effort, but completely finished... a closed project, as you say11:35
mafmthe basic port can then be expanded, but if it's half-baked and it doesn't work at all by the time the project is finished, maybe it all goes to waste11:36
jeremy_bennettwallento: More ideas emailed to you11:42
wallentothanks, jeremy!11:43
poke53281mafm: Yes I was asking this. But didn't get a reply.12:44
poke53281At the moment I am just curious.12:45
poke53281Last year was funny, but also time consuming. Still wait for some Linux patches for MMIO so that I can finish the GSoC project from last year12:46
mafmpoke53281: what was the name of the project?12:46
mafmand yeah, I am afraid that most people actually involved in the project never participate in that channel12:47
mafmI don't know if it was always like that, but it has been for a few months :)12:47
poke53281Here are the details12:47
wallentohey poke5328112:47
wallentoyou already have some nice ideas?12:48
wallentoI will put up a fossi page this afternoon for collecting ideas12:48
poke53281To have finally a good QEMU port would be still nice.12:48
poke53281and doable12:48
wallentothat means bringing the current one into a proper shape, right?12:48
poke53281 + debugger + new interface + ....12:50
wallentothat sounds great12:51
mafmpoke53281: reading through the PDF... amazing :D12:51
poke53281port of OpenJDK12:55
poke53281Actually there is a good chance, that I start at a software company this year, which does mainly Java. So would be nice to start there with a port to OpenRISC ;)12:56
poke53281mafm: Thanks13:00
_franck__wallento: what about porting romcc (from coreboot project) to openrisc. It's a C compiler that only use registers. It can be useful to write rom code.13:17
wallentosounds good13:17
wallentocan you write a 2 sentence description of it?13:18
_franck__or SMP support in openocd for openrisc13:18
wallentoand send to my mail: stefan@wallentowitz.de13:18
_franck__wallento: ok I will13:18
wallentogreat, thanks13:18
wallentofeel free what you think is 3 months of work and a project with an actual outcome13:18
wallentonice page, funny that we never use it anymore13:21
wallentoso, I have started off with a few ideas I have the two sentences for already13:51
wallentoplease create a PR with your ideas to this file:
poke53281wallento: Send you two projects14:37
poke53281There were the projects last year.14:45
poke53281Oh, I forgot. What about gdb port?14:45
poke53281stekern or blueCmd would be a good mentor for this.14:46
poke532813.5 hours to deadline wallento.15:31
wallentoyeah, I am nearly finished with the application15:32
wallentothe ideas are not mandatory before the deadline15:32
wallentobut presumably are considered during evaluation15:32
poke53281the question is, if you can apply in some way as an umbrella organization together with lowrisc and RISC-V. This would increase the chances significantly. But it is too late now I guess.15:33
wallentoyeah, I think so15:35
wallentoactually, FOSSi should at some point in time be able to run its own summer of code15:35
mafmwallento: (reading the fossi website) -- librecores is nice, but given the name of the org, fossicores could be a good name, although maybe some people will try to find second meanings or puns :D15:58
mafmfoxycores, fossilcores, fussycores...15:58
wallentoyeah, and actually it went the other way around15:58
wallentolibrecores was there before fossi15:59
wallentobut we want it to be more then librecores15:59
mafmah, since there is no link I thought that it hadn't been created yet15:59
wallentoit is now15:59
wallentothe intention was there earlier15:59
wallentobut we just started with drafting the implementation15:59
wallentowe wanted fossi as the legal body16:00
wallentoto not have it under control of a company or a single person16:00
mafmis this a departure of, then?16:04
juliusbwe offered to take opencores over but that didn't work out16:06
juliusbso we've gone our own way16:07
mafmthat's quite nice16:14
mafmI was under the impression that everything would be shifted to risc-v now, and that these previous efforts would fade away16:15
wallentono, they can co-exist. In my humble opinion the RISC-V community and the OpenRISC community are organized differently16:19
wallentoand the ambition around librecores is somehow orthogonal to both16:19
wallentoits just that opencores is so 90s ;)16:20
mafmit is16:20
wallentospeaking of all this stuff17:27
wallentowe need to get the overview pages forward:
wallentoIts a petty we still need to reference the opencores Wiki as overview17:28
-!- Netsplit over, joins: foreverska18:29
-!- Netsplit *.net <-> *.split quits: rhythmx, simoncook18:29
-!- Netsplit over, joins: rhythmx18:29
olofk_franck_: Yeah!!! romcc for OpenRISC is the best idea ever :)22:15
--- Log closed Sat Feb 20 00:00:39 2016

Generated by 2.15.2 by Marius Gedminas - find it at!