| --- Log opened Mon Dec 16 00:00:11 2013 | ||
| -!- Netsplit *.net <-> *.split quits: vxe, ssvb | 09:52 | |
| -!- Netsplit over, joins: ssvb, vxe | 09:54 | |
| maxpaln | hi back again - and I have come unstuck with the burst write into the memory again | 14:11 |
|---|---|---|
| maxpaln | I haven't implemented the ability to do burst read and write int he DDR3 memory controller yet | 14:12 |
| maxpaln | in the OR1200 you can disable bursts through the or1200-defines.v file | 14:12 |
| maxpaln | I can see some parameters controlling burst operation in the ethmac-defines.v - but it isn't clear whether it is possible to disable burst memory access using these | 14:13 |
| maxpaln | does anyone know? | 14:13 |
| iorivur | Sorry, no. And I have another topic. Can I talk that here until someone who is expert about your problem come here? | 14:23 |
| maxpaln | running a simulation now to see if the behaviour can be reduced to classic cycles - not hopeful though! | 14:23 |
| -!- Netsplit *.net <-> *.split quits: mboehnert | 14:23 | |
| iorivur | I think there are few documents in official github release. | 14:24 |
| iorivur | I'm writing some templates and little bit contents but I don't know much about OpenRISC, so anyone else should write it. | 14:25 |
| iorivur | Will someone write and expand it if I send a pull-request to github/openrisc/orsoc? | 14:26 |
| maxpaln | well, it appears that by reducing ETH_BURST_LENGTH and ETH_BURST_CNT_WIDTH to 1 and undefining ETH_RX_BURST_EN I can disable bursts - going to try it in sim and HW now | 14:46 |
| maxpaln | I think adding burst functionality to the DDR3 controller might be the next priority once I get the processor running - will see a significant performance delta I suspect | 14:47 |
| iorivur | I sent a pull-req | 15:16 |
| _franck_web_ | iorivur: I wrote a tutorial on orpsocv3 here: http://www.elec4fun.fr/2011-03-30-10-16-30/2012-08-22-20-50-31/or1200-barebox-on-de1 | 15:22 |
| iorivur | _franck_web_: great! It'll be good for me. But I think official repository should explain how to use itself. | 15:26 |
| iorivur | Big thanks anyway | 15:26 |
| _franck_web_ | iorivur: you'll tell me if it's helpful | 15:30 |
| iorivur | sorry, how did you mean? | 15:34 |
| -!- Netsplit *.net <-> *.split quits: vxe, ssvb | 15:39 | |
| _franck_web_ | I mean if you find errors in the tutorial, please tell me | 15:39 |
| iorivur | ah ha | 15:39 |
| -!- Netsplit over, joins: ssvb, vxe | 15:44 | |
| stekern | iorivur: that's a good initiative, but I think the readme should explain what orpsoc is, not what openrisc is | 15:46 |
| stekern | and a second general comment on pull requests (or patches in general), you don't need to present your changes to stuff you are adding in previous patches | 15:47 |
| stekern | people are generally mostly interested in seeing your "finished product", not how you got there. | 15:48 |
| iorivur | stekern: Ah, yes, OK. It's reasonable. How do you think to explain it? | 15:49 |
| stekern | some info is already available here: https://github.com/openrisc/orpsoc/blob/master/INSTALL and that combined with what _franck_web_ pasted is probably a good start for a README | 15:52 |
| stekern | on a different subject, why am I such a sucker for things that are broken... I've wondered off hacking on this eco32 gnu toolchain, when there's so much other things I should be doing... | 15:54 |
| _franck_web_ | :) | 15:54 |
| stekern | it's something I'd like to dub olofk's law, "it's always most fun to do the stuff you are least supposed to do" | 15:55 |
| _franck_web_ | I also always need something new to play (something like this http://zet.aluzina.org/index.php/Zet_processor ?) | 15:59 |
| stekern | heh, yeah I've seen that | 16:00 |
| iorivur | Could I use part of what _franck_web_ pastes? I think workflow section is well-documented | 16:00 |
| stekern | but x86 is so boring | 16:00 |
| _franck_web_ | is it ? | 16:00 |
| _franck_web_ | iorivur: yes, you can copy what I wrote | 16:01 |
| iorivur | Thanks alot! | 16:01 |
| _franck_web_ | however, writing this kind of doc on the wiki would be better I think | 16:01 |
| stekern | _franck_web_: it's too mainstream, we softcpu hipsters loathe mainstream ;) | 16:04 |
| _franck_web_ | :) well it's mainstream but there not much x86 softcore out there. Ans this one misses a pipeline, 32bits support, cache, mmu | 16:05 |
| stekern | I really should fix the de0_nano simulation in orpsocv3 instead of this... | 16:05 |
| stekern | _franck_web_: ah, alot of missing features and defiencies, I take back all I said ;) | 16:06 |
| stekern | and it would suite well with another project I have in store, the opl2 clone | 16:09 |
| -!- Netsplit *.net <-> *.split quits: vxe, enghong, jonmasters_, ssvb, ysionneau, doomlord__, arokux, julzmb, ysangkok_ | 16:18 | |
| -!- Netsplit over, joins: enghong, julzmb, ysionneau, ysangkok_, vxe, jonmasters_ | 16:21 | |
| _franck_ | as I'm working on a or1ksim to vpi to rtl bridge, I followed the olof's idea to use etherbone... | 21:56 |
| _franck_ | ...when it doesn't work, it's not easy to get into this code. Have you ever seen such a big function: | 21:56 |
| _franck_ | http://svn.ohwr.org/etherbone-core/api/format/slave.c | 21:56 |
| _franck_ | ~400 lines | 21:56 |
| _franck_ | now I know why the Kernel's doc says "Functions should be short and sweet, and do just one thing" | 21:58 |
| _franck_ | "The maximum length of a function is inversely proportional to the | 21:58 |
| _franck_ | complexity and indentation level of that function" | 21:58 |
| * _franck_ swithes to his second idea: make an or1ksim to jtag to wb master bridge (using openocd tcl server) | 22:18 | |
| stekern | _franck_: if you want to see bi functions, go take look in bfd... | 22:36 |
| stekern | *big | 22:37 |
| stekern | https://github.com/openrisc/or1k-src/blob/or1k/bfd/elf.c#L4360 | 22:47 |
| --- Log closed Tue Dec 17 00:00:12 2013 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!