| --- Log opened Fri Jul 01 00:00:18 2016 | ||
| olof | ZipCPU: I found this mentioned on twitter when I was on the way home from work yesterday https://parallel.princeton.edu/papers/openpiton-asplos16.pdf | 02:41 |
|---|---|---|
| olof | But they claim that OpenRISC was last update in 2012 and that it's not multi-core | 02:43 |
| Finde | apologies olof | 04:46 |
| Finde | we're thinking of making an online taxonomy with the ability for people to update to avoid issues like this | 04:46 |
| Finde | we only had one person looking for the info and with the number of cores it was tough to make sure we got every detail correct in January | 04:47 |
| Finde | I think the dates for a number of the cores were probably just taken from opencores | 04:48 |
| olof | Finde: Hi. You're from OpenPiton? | 04:52 |
| olof | And no worries. I understand that it's hard to get an overview of man different projects. I was planning to reach out to you for a correction, but having it as a public document would of course be even better | 04:54 |
| Finde | yeah the more I think about it the more value I think we could gain | 05:03 |
| Finde | and yeah I'm the first author on that paper | 05:03 |
| Finde | I'm hoping to come to ORCONF in October to meet you all | 05:07 |
| olof | Finde: That's great. I've seen your mails via Jose and Julius. Looking forward to meet you and learn more about the project | 05:10 |
| olof | Finde: Make sure to register on the orconf site once you have decided if you are able to show up | 05:11 |
| Finde | I think I registered the other day | 05:12 |
| Finde | currently putting together a talk blurb | 05:12 |
| olof | Ah yes. You're registered. | 05:14 |
| olof | I was thinking about registering for a presentation too | 05:15 |
| olof | Oh well. Sounds like you have it under control :) | 05:15 |
| Finde | well it can seem that way... | 05:23 |
| Finde | :p | 05:23 |
| Finde | but yes if you have any questions about openpiton, please feel free to ask me anytime (or email) | 05:23 |
| Finde | I imagine we're going to have a lot of questions in future | 05:23 |
| wallento | hi finde, nice paper, congrats | 06:07 |
| Finde | thanks wallento! | 06:07 |
| olof | Hi olofk_! Looks like my internet connection at home has been fixed :) | 06:55 |
| ZipCPU | olof Finde : And here I thought OpenSPARC was dead with no one working on it ... | 06:58 |
| olof | ZipCPU: We all did :) | 07:00 |
| olof | wallento: Want to challenge this with OpTiMSoC? :) "OpenPiton is the world’s first open source, general-purpose, multithreaded manycore processor." | 07:02 |
| ZipCPU | Go ahead, as long as I can be a fly on the wall. | 07:02 |
| wallento | optimsoc should probably have found its way into related work | 07:03 |
| wallento | but openpiton is good for my related work | 07:03 |
| wallento | good to see more open source in the area | 07:04 |
| olof | Absolutely | 07:04 |
| ZipCPU | "Qualitative comparison against prior open source processors." -- That's specifically been my struggle. I don't want to create "qualitative" comparisons, but rather "quantitative". | 07:05 |
| wallento | and openpiton beats optimsoc in google search "open source manycore" now | 07:05 |
| ZipCPU | I just googled "opentimsoc" and came up with nothing. | 07:06 |
| ZipCPU | Looks like I mispelled "OpTimSoC" | 07:06 |
| olof | Hey, what happened to googlefight? | 07:08 |
| olof | Ahh.. some spammer took it over :( | 07:08 |
| olof | I used to love a good google fight | 07:09 |
| ZipCPU | Finde : Table 3 states that the OpenPiton uses 75k out of 63k available LUTs in an Artix-7. Does this mean it never worked on an Artix-7? | 07:15 |
| ZipCPU | Okay, here it is: "Note that the default configuration is too large for the Artix 7" ... | 07:17 |
| olof | Nice! OpenOCD TCL interpreter can source files, so I can script stuff easily | 07:47 |
| ZipCPU | So they ran a multicore Sparc at 30MHz on an Artix-7. Fascinating. I never thought the ZipCPU was that unique, but I am running at a 100MHz clock on an Artix-7. I just didn't expect that to be very unique. | 07:50 |
| olof | _florent_: Are you running OpenRISC or lm32 on the Arty? | 07:55 |
| olof | Looking quickly at the source it seems like it's running at 100MHz at least | 08:02 |
| _florent_ | olof: I think I only tested with lm32, but mor1kx should also work | 08:12 |
| olof | Is this really the only way in verilog to avoid warnings for width mismatch when adding a constant 1 to a register? x <= x + {{WIDTH-1{1'b0}},1'b1}; | 08:36 |
| olof | Isn't 'd1 supposed to work too? | 09:01 |
| olof | or does that still requires the width of the operand to be known? | 09:04 |
| olof | Ah wait! I get rid of the warning by just making the constant smaller than the register | 09:11 |
| olof | So it's enough to do 1'd1 | 09:11 |
| olof | Which of course requires prior knowledge of the constant value | 09:11 |
| olof | Just learned about verilog array of instances too. Didn't know about that | 09:11 |
| olof | And instance name seems to be optional. That was unexpected | 09:13 |
| Finde | yeah 1'd1 should work olofk | 09:36 |
| Finde | oh he left never mind | 09:37 |
| Finde | I should've learned after using irc this long that you need to read the whole scrollback first | 09:37 |
| Finde | wallento: is OpTiMSoC HW multithreaded? | 09:58 |
| wallento | no, it is not | 09:58 |
| wallento | its based on the mor1kx OpenRISC core | 09:59 |
| wallento | loosely coupled clusters | 09:59 |
| Finde | ah gotcha | 09:59 |
| Finde | hard to search for something on a website when it's not there | 09:59 |
| Finde | I was scouring the pages :p | 09:59 |
| Finde | it looks really cool | 09:59 |
| Finde | I'll make sure we look into your infrastructure and try to cite you in future | 09:59 |
| wallento | yeah, nevermind, it was more a response to olof trying to get us into a fight ;) | 10:00 |
| wallento | we never really took off | 10:00 |
| wallento | more of a one man show for my phdthesis | 10:01 |
| wallento | I would have loved to find open source manycore at this time, but even nocs were hard to find | 10:01 |
| wallento | and are until today | 10:01 |
| wallento | I would also tend to give up OpTiMSoC in favor of something else, but need to finish my PhD thesis before trying to rebuilt the whole stuff on a different platform again | 10:02 |
| Finde | yeah I mean there wasn't much good out there that was multithreaded when we started in 2014 and there still isn't really | 10:04 |
| Finde | we needed the threading for a particular research idea | 10:04 |
| wallento | fully agree | 10:04 |
| Finde | and risc-v was just an isa at that point | 10:05 |
| wallento | its the case for most open source research platforms | 10:05 |
| wallento | you want to show something | 10:05 |
| wallento | for me its efficient on-chip message passing | 10:05 |
| wallento | so you end up with building the platform first.. | 10:05 |
| Finde | and that takes N years | 10:07 |
| wallento | finde: yeah | 10:12 |
| wallento | but as a side product I contributed to make OpenRISC multicore | 10:12 |
| wallento | I will write you a mail the next days, we should keep the discussion going | 10:12 |
| Finde | excellent | 10:15 |
| Finde | I'm sure you have some infrastructure that would be useful for us in some form or another | 10:15 |
| Finde | so we'll need to check out your code | 10:15 |
| Finde | wallento: if you happen to be going to orconf, hopefully we can meet there too | 10:18 |
| wallento | Finde: yes I already booked my travel | 10:19 |
| wallento | looking forward to meet you there! | 10:19 |
| wallento | one thing we currently cut out of optimsoc and make it reusable (mostly for lowrisc) is the debug infrastructure | 10:20 |
| wallento | http://www.opensocdebug.org | 10:20 |
| wallento | thats the most annoying part, maybe we can cooperate there | 10:20 |
| olofk | Ahh.. Finally! | 11:56 |
| SMDwrk | olofk: I never thought there are such troubles with internet in Sweden | 12:30 |
| olofk | SMDwrk: Turns out that the modem broke, which gave me a huge packet loss, and then someone tried to fix it, thought it was a problem outside of the house and managed to disconnect it all together. The guy who was here today was a bit angry at his coworker :) | 12:31 |
| olofk | stekern: I'm trying to silence some warnings in mor1kx that always turns up in my logs, but I'm not sure about what to do with decode_a and decode_b when CALCULATE_BRANCH_DEST != "TRUE" | 12:33 |
| stekern | mmm | 12:34 |
| stekern | that's there because of a difference betwwen cappuccino and espresso, iirc | 12:34 |
| olofk | ZipCPU: Oh, I forgot. You asked about accomodation at orconf. wallento asked our man in Bologna about that a while ago and we recently got a response. Will put it up on the web page soon | 12:42 |
| olofk | stekern: I fixed some stuff now. Had 171 warnings in total | 12:43 |
| olofk | ...now I have 233 :( | 12:43 |
| ZipCPU | olofk: Thanks! I'm hoping there's something nearby enough that I might be able to get by without a rental car, or a daily taxi. | 12:44 |
| wallento | ZipCPU: virtually anyhing works in bologna | 12:49 |
| wallento | hotel ren | 12:49 |
| wallento | zo got recommended | 12:49 |
| wallento | the location is in the southwest of the inner city | 12:50 |
| olofk | When you compile a C file, is the path of the C file always implicitly added as an include directory? | 15:46 |
| olofk | Looks like it | 16:02 |
| blueCmd | yo. do you guys have any interest in the freecores.net domain? | 16:27 |
| blueCmd | I just keep renewing it but not doing anything with it | 16:28 |
| olofk | blueCmd: I totally forgot about that one | 16:29 |
| olofk | I don't think we need it now that we decided on librecores, but I'll mention it on the FOSSi board meeting next week and see what the other say | 16:30 |
| blueCmd | it's like 80 kr per year and might be good to hold on to for whatever reason | 16:31 |
| blueCmd | I would feel bad letting it go for some link farmer to get their hands on it | 16:31 |
| olofk | Yeah, I feel that way too | 16:32 |
| olofk | It sucks that we lost openrisc.net | 16:32 |
| kc5tja | olofk: I'm here from Twitter, per your recommendation. I'll lurk now. :) | 17:07 |
| ZipCPU | olofk: It may depend upon the compiler. I know #include "file.h" is supposed to include the local directory, whereas #include <file.h> not so much. | 17:46 |
| --- Log closed Sat Jul 02 00:00:20 2016 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!