| *** klysm is now known as klys | 01:21 | |
| ghostmansd-pc | Hi everyone. $(git pull origin master --rebase) on libreriscv repo fails. Did something changed? | 05:11 |
|---|---|---|
| ghostmansd-pc | FATAL: R any libreriscv dmitry3mdeb DENIED by fallthru | 05:12 |
| ghostmansd-pc | $(ssh gitolite3@git.libre-soc.org) does not show anything related to libreriscv, only riscv-isa-sim and riscv-tests. | 05:13 |
| ghostmansd-pc | Hm, I see the name's changed. Is the problem related to underscores? | 05:16 |
| ghostmansd-pc | I've updated the docs with the new chapter depicting the transition from `test_caller.py` to `caller.py`; this chapter is intended to serve as a bridge between the test and the decoder. | 06:46 |
| ghostmansd-pc | Please, ping me, when the git is available again; I'd like to push the changes. | 06:47 |
| ghostmansd-pc | The new chapter depicts the flow up to the point we invoke `info.call`. | 06:47 |
| lkcl | yeah i know, i have no idea.... arghh, i know what it is: i upgraded gitolite3. | 10:27 |
| lkcl | to try to "fix" that security problem | 10:28 |
| lkcl | i've downgraded to 3.6.6-1 again, it should be back | 10:33 |
| lkcl | ghostmansd: it was "available"... just not for the libreriscv repo. | 10:38 |
| lkcl | regardless of what was in the conf file, the entire repo did not appear - at all - in the ssh access. | 10:38 |
| lkcl | confirmed it's fixed by a downgrade. | 10:39 |
| ghostmansd | lkcl: pushed, I'll continue on extending the docs. | 15:21 |
| ghostmansd | lkcl: I think it might be a good option to extend the doc from there on in course of bug 656; what do you think? | 15:23 |
| lkcl | doh 656 not 646 1 sec | 15:23 |
| lkcl | yes great idea | 15:24 |
| ghostmansd | I mean, the doc will still be about add instruction, but I think I need to spend some time really investigating the decoder :-) | 15:24 |
| ghostmansd | The discussion was quite fast on this part, perhaps too fast for a newbie like me | 15:25 |
| lkcl | yehyeh | 15:25 |
| lkcl | it was a glimpse. | 15:25 |
| lkcl | just looking at the wiki, it's not in "recent changes", will look at the git logs | 15:26 |
| ghostmansd | Ah, yep, I need to cross-link it | 15:29 |
| ghostmansd | Would [link_test](full_url_as_in_wiki) be sufficient? | 15:30 |
| lkcl | ikiwiki just needed a poke to rebuild the recentchanges.mdwn | 15:31 |
| lkcl | if the URL is part of the wiki just use standard [[wikipagename]] | 15:31 |
| ghostmansd-pc | a quick question on class MicroOp: it uses manually-entered "enum" values. Why not auto()? | 15:31 |
| lkcl | if it's somewhere else in the tree hierarchy use [[subdirectory/actualpage]] | 15:32 |
| lkcl | because the numbers matter when it comes to reading them and trying to check against the HDL simulation gtkwave output | 15:32 |
| lkcl | if it was "auto", and there was some random number inserted into the VCD trace files, how would you know what that number actually meant? | 15:33 |
| lkcl | then, we add another one, and accidentally someone inserts it in the middle, and oh whoops all the numbers changed, "wtf moment" | 15:33 |
| ghostmansd-pc | Perhaps I could have a map between enum values and strings. But fair point, yeah. | 15:34 |
| lkcl | we already have that with "asmcode" which is the index into the list of instructions, there's over 256 of them, we have absolutely no idea which number refers to which actual asmcoed | 15:34 |
| lkcl | power_enums.py _insns, lists all instructions, it wasn't originally intended to actually go into the HDL | 15:35 |
| lkcl | "from the src/openpower/decoder/isa/caller.py" - these could do with linking to the actual URL in git.libre-soc.org | 15:35 |
| lkcl | in that case... | 15:35 |
| lkcl | https://git.libre-soc.org/?p=openpower-isa.git;a=blob;f=src/openpower/decoder/isa/caller.py;hb=HEAD | 15:36 |
| lkcl | which i will do, i have the URL now | 15:36 |
| lkcl | ha, it got the font right but the colour is borked, because it's a link. doh :) | 15:38 |
| ghostmansd-pc | :-D | 15:38 |
| ghostmansd-pc | https://libre-soc.org/3d_gpu/architecture/decoder/ how about cross-linking here? | 15:38 |
| ghostmansd-pc | * First steps for a newbie developer: [[docs/firststeps]] | 15:38 |
| lkcl | yes. ok so that would be [[3d_gpu/architecture/decoder]] | 15:39 |
| ghostmansd-pc | yep | 15:40 |
| ghostmansd-pc | lkcl: I'm wondering if we have some script to convert from microwatt's decode1.vhdl to our CSVs? | 17:56 |
| ghostmansd-pc | or at least legend to CSV form, like decode1.vhdl has | 17:56 |
| ghostmansd-pc | -- unit fac internal in1 in2 in3 out CR CR inv inv cry cry ldst BR sgn upd rsrv 32b sgn rc lk sgl rpt | 17:56 |
| lkcl | ghostmansd: they should be the same fields, except sigh, microwatt added one new field last month, and i added a new one. | 18:43 |
| lkcl | but the ones that are the same names are the same fields | 18:44 |
| lkcl | about the startup tutorial: i thought about it, and actually i'm happy for you to a completely separate one with a separate budget for the BCD instructions, "how to add new instructions" | 18:44 |
| lkcl | if you can add a section to the first one, "modify the pseudocode CSV to be a subtract instead of an add" and everything else we went through | 18:46 |
| lkcl | then i'm happy to declare that one done | 18:46 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!