| --- Log opened Wed Mar 16 00:00:36 2016 | ||
| olofk | andrzejr: Yeah, I never pushed the coregen provider stuff to FuseSoC, but I think the other patches are applied | 04:23 |
|---|---|---|
| andrzejr | olofk, there are a few new commits in my repo but they need reviewing (haven't looked at them for ~4 months). Are there any features providing a function similar to coregen provider? | 04:57 |
| wallento | andrzejr: cool, thanks! | 05:00 |
| wallento | will have a look soon | 05:01 |
| wallento | olofk: which gdbserver? | 05:01 |
| _franck__ | juliusb: FYI search doesn't work anymore on you IRC log server | 05:18 |
| olofk | andrzejr: No, I haven't implemented anything equivalent to your coregen provider. It would be interesting to see if it can be solved by doing it with a pre_build_script instead | 06:00 |
| olofk | I still haven't decided if the coregen provider is genious or madness :) | 06:01 |
| olofk | wallento: This one http://opencores.org/or1k/Linux#Status | 06:02 |
| wallento | ah, its inside Linux userland, right? | 06:04 |
| olofk | That's how I understand it | 06:47 |
| olofk | Wanted to deprecate that whole page, but wasn't sure about what to say about gdbserver | 06:48 |
| shorne | olofk: in terms of those opencores pages, I see they are all backed up to https://github.com/openrisc/community-wiki/blob/master/Linux.mw | 07:08 |
| shorne | should we try to get them all into openrisc.io ? | 07:09 |
| shorne | then put links on opencores.org to openrisc.io ? | 07:09 |
| shorne | is that the plan? | 07:09 |
| shorne | or are there issues with copying that text | 07:09 |
| wallento | yes, thats the idea | 07:19 |
| wallento | filter out the currently important information | 07:19 |
| wallento | lists.librecores.org is up now, will need some testing, but we can go live then | 07:19 |
| olofk | I've been fixing up a lot of the documentation on the opencores wiki over the years, but much of the stuff would probably make sense to have as stand-alone pages, like the ones we have for Debugging and Linux now in the tutorials repo | 07:56 |
| wallento | https://lists.librecores.org/listinfo/openrisc | 08:13 |
| wallento | the new mailing list is up | 08:14 |
| olofk | wallento: Great. Let's tweet :) | 08:14 |
| wallento | I have made julius and you administrators too, but it does not require moderation or any other s**t | 08:17 |
| rah | is this *another* openrisc mailing list? :-) | 08:18 |
| olofk | rah: Yeah, it's a mess. But openrisc.net died because the maintainer ran away, and opencores doesn't seem to care, so that one is down too | 08:20 |
| rah | oh right | 08:21 |
| olofk | Which is all the more reason to move away the rest of the important stuff from opencores | 08:22 |
| rah | aye | 08:23 |
| rah | I was aware that there were issues with opencores but I didn't realise the openrisc.net maintainer ran away | 08:24 |
| wallento | yes, and librecores.org is now community controlled | 08:24 |
| wallento | by FOSSi Foundation | 08:25 |
| wallento | so, it should be _the_ OpenRISC mailing list and not yet another :) | 08:25 |
| olofk | rah: I tried to get control over openrisc.net when it was about to expire, but lost it to domain squatters :/ | 08:27 |
| rah | :-/ | 08:28 |
| rah | wallento: aye, I saw FOSSi | 08:28 |
| rah | and #fossi :-) | 08:28 |
| rah | this is good | 08:28 |
| olofk | Need to do something about the pulpino boot. The program I run only takes ~100us, but there's a 15ms startup time before I get there | 08:30 |
| olofk | Not sure exactly what it does. It seems to send a reset and some other stuff over jtag | 08:31 |
| olofk | and half of the times, I forget that I need to copy some files manually, so I wait 10 minutes just to get an error message. Doh! | 08:34 |
| olofk | But I'm damn close now. Probably just some timescale fuckup in the uart printout | 08:42 |
| olofk | Haven't done much systemverilog, but at least it looks like they have added the option to explicitly set the time unit, as in VHDL. That alone would be enough to switch over :) | 08:46 |
| olofk | verilog timescale are horrible horrible horrible | 08:46 |
| olofk | Unfortunately, timescales are heavily used in things like RAM models, and I don't expect them to swtich over anytime soon | 08:47 |
| shorne | wallento: thanks for setting up that mailing list, joined. Do you think we would need speparate lists for separate discuttions? like linux, gdb, etc? | 09:04 |
| shorne | or all in one place? I would think just one list is fine for now | 09:04 |
| wallento | yes, we should stick to one place | 09:05 |
| wallento | not much traffic | 09:05 |
| wallento | and openrisc discussions have the tendency to diverge into several topics :) | 09:05 |
| shorne | olofk: for the memset patch I am just doing some testing, in case there are questions about performance, how did you test that its actually faster? | 09:13 |
| olofk | shorne: I fired it up in or1ksim, enabled instruction tracing and noticed that it used 10% less instructions during boot and made 13% fewer memory accesses | 09:21 |
| olofk | Not sure about the exact numbers, but I hopefully wrote it down somewhere | 09:21 |
| olofk | During Linux boot that is | 09:21 |
| shorne | ok, thats what I thought, you have to look at a kernel load | 09:22 |
| shorne | I was thinking to write some test, but it would be a bit hard to from user space | 09:22 |
| shorne | ill trace the boot | 09:23 |
| olofk | shorne: Yeah, I was a bit unsure about the proper procedure to test it. More testing would be most welcome | 09:23 |
| olofk | It sucks that our searchable documentation database is down :( | 09:24 |
| olofk | juliusb: Heeeeelp! | 09:24 |
| olofk | shorne: http://www.juliusbaxter.net/openrisc-irc/%23openrisc.2015-02-06.log.html#t12:31 | 09:31 |
| shorne | olofk: yeah, I found lost of good stuff in there | 10:13 |
| shorne | thank you for the link | 10:13 |
| olofk | hmm.. ok. So the pulpino UART for some reason sends with the wrong baud rate | 11:35 |
| olofk | hmm.. actually.. it's supposed to do that.. hmm | 11:37 |
| olofk | aha. Got it! | 11:49 |
| olofk | timescale calculation was quietly rounded, so a period time of 1280 became 1000 | 11:50 |
| olofk | I tell you. They are nothing but trouble those damn timescales | 11:50 |
| andrzejr | olofk, I'm OK with pre_build_script instead of coregen provider. I remember you preferred not to call scripts for portability reasons, though. Hence the provider. | 17:12 |
| andrzejr | IMHO provider is fine but it would be better if it could be defined by the user. The same is true for synthesis and simulation tools. | 17:14 |
| andrzejr | One (minor) issue with the provider is naming of methods and status flags ("fetch", "downloaded"). The API could be renamed or extended if we cared about that. | 17:20 |
| shorne | Does the mor1kx support write back cache now? I am reading pdf spec and it says write through only, but I see the kernel code supports flushing for write back. But I cant really see where it happens/is configured in the mor1kx verilog code | 18:34 |
| -!- Netsplit *.net <-> *.split quits: killertux, rokka | 22:19 | |
| -!- Netsplit over, joins: killertux, rokka | 22:25 | |
| --- Log closed Thu Mar 17 00:00:38 2016 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!