| stekern | juliusb: yes, I think I'll put in that kind of signal to hold off ctrl_branch_occur until it's really certain that the load/store have finished without causing an exception | 05:11 |
|---|---|---|
| stekern | ... or then I introduce the pipeline nop bubble at this phase in the patch merging and only use it to squeese in a nop between lsu ops followed by a jump | 05:12 |
| stekern | since I will want to introduce that at some stage anyway | 05:13 |
| stekern | because holding off ctrl_branch_occur and squezing in a nop in between is essentially the same | 05:13 |
| stekern | ah, no. I found out another thing that makes this work lurking in my unmerged stuff | 06:28 |
| stekern | when I'm doing the load/stores in ctrl/mem stage I have changed the exception logic so that lsu_valid is not asserted for misaligns and dbus errors | 06:30 |
| stekern | and instead of padv_ctrl the exception is kicked off by registering those signals (since that's what is stalling the pipeline) | 06:30 |
| stekern | of course that logic can still work before I have moved the load/stores into ctrl/mem stage | 06:31 |
| stekern | (and it does according to the tests) | 06:34 |
| stekern | a bit more work to merge that change in at this stage, but probably worth it to get those things in in nice small incremental steps | 06:35 |
| stekern | bah, except for the debugexceptsteptest | 06:59 |
| stekern | think I've got that sorted out now as well | 07:44 |
| mor1kx | [mor1kx] skristiansson pushed 8 new commits to master: https://github.com/openrisc/mor1kx/compare/d5231383c5c9...422d17116c07 | 08:15 |
| mor1kx | mor1kx/master 8c75a3f Stefan Kristiansson: cappuccino/ctrl: add missing 'input' to except_trap_i | 08:15 |
| mor1kx | mor1kx/master 53e4820 Stefan Kristiansson: cappuccino/ctrl: simplify delay slot detection for exceptions... | 08:15 |
| mor1kx | mor1kx/master b6dbd4e Stefan Kristiansson: decode: add detection of branches to immediates... | 08:15 |
| stekern | juliusb: is it ok if I just push stuff I'm kind of certain should go in without question (like I just did) and put up pull requests only on stuff that I think you should take a look at first? | 08:53 |
| -!- Netsplit *.net <-> *.split quits: @juliusb, gxti | 10:19 | |
| -!- Netsplit over, joins: gxti | 10:19 | |
| -!- Netsplit over, joins: juliusb | 10:20 | |
| juliusb | stekern: yeah no worries | 11:11 |
| juliusb | I trust your judgement | 11:11 |
| stekern | even when I haven't took my medicines? ;) | 11:45 |
| zewan | Hi all. Has anyone seen this? http://lkcl.net/articles/fsf_endorseable_processor.html | 12:21 |
| zewan | Is that your job? :) | 12:21 |
| juliusb | zewan: very interesting, thanks for pointing thato ut | 12:59 |
| juliusb | stekern: hah what medicine? Finnish medicine (vodka?) | 12:59 |
| juliusb | zewan: that project isn't just about processor design, it's about SoC design with a bunch of cores which there aren't open source variants of | 13:00 |
| juliusb | where they gettign this $10mil from? | 13:01 |
| juliusb | for $10mil you could maybe get a SoC done | 13:01 |
| juliusb | not sure why you would, you're just better bribing some employees of qualcomm to release the full software and specs for one of their chips | 13:01 |
| juliusb | you should get a bit for $10mil | 13:01 |
| juliusb | That "The justification" paragraph doesn't make sense. | 13:07 |
| juliusb | what is "the average factory"? | 13:07 |
| juliusb | but streamlining and simplifying hasn't got anything to do with it being "open" in this sense (free for them to do with as they wish) | 13:08 |
| juliusb | (at a lower level than other SoC products) | 13:08 |
| juliusb | my other comment is that, commercially, they're not going to be onto a winner - SoCs are a lot more about what CPU and GPU core you have, connectivity is a lot more important nowdays then ever - phone manufacturers don't want to put down many ICs, and increasing integration of wireless and baseband (TI, Qualcomm, Samsung, Renesas) into APs means that no one will care about something that is fully open if it doesn't have | 13:14 |
| juliusb | so they can make it, it's not going to be a success other than within the free software community | 13:14 |
| zewan | Who they are? | 13:16 |
| juliusb | whoever does this | 13:18 |
| stekern | juliusb: nah, then it probably would have been "even when I have took my medicine?" | 13:50 |
| stekern | (just to make things clear, I'm not really on any medication, not finnish nor any other ;)) | 13:52 |
| stekern | juliusb: on the other note, I'm not saying you're wrong, but remember that there are other areas than phone manufacturers | 13:55 |
| stekern | the industrial segment usually have very different needs | 13:56 |
| stekern | maybe I should have said consumer electronics instead of phone manufacturers | 13:57 |
| stekern | ... but reading the link more carefully makes me realise my comment was a bit ot | 14:10 |
| --- Log closed Thu Dec 06 18:52:04 2012 | ||
| --- Log opened Thu Dec 06 18:52:20 2012 | ||
| -!- Irssi: #openrisc: Total of 21 nicks [0 ops, 0 halfops, 0 voices, 21 normal] | 18:52 | |
| -!- Irssi: Join to #openrisc was synced in 23 secs | 18:52 | |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!