Friday, 2022-07-22

programmerjakelkcl: please don't add important details only by editing comments, no one looking through their emails sees any of the edits:
lkclprogrammerjake, i was on a call making notes, trying to concentrate on not losing track of the conversation.13:01
lkcland i don't expect people to read all the email notifications, i expect people to go to the bugreport and read all comments in full, so as to retain context13:01
programmerjakewell, I, at least, keep up-to-date on the bug tracker by reading my email, not by reading each bug's list of comments everytime I think something might have changed...I highly doubt I'm the only one who does that13:07
programmerjakecontext is easily seen by looking at previous emails13:08
lkcli find the bugzilla email almost impossible to read, it's so badly formatted13:08
octaviusIt's usually the emails that get me to check bugs, cause my client gives me pop ups13:08
octaviusI have a bit of time before lunch lkcl, what should I get started on? Move the svp64-proposal.tex and try to put the primer and spec in it?13:09
programmerjakeimho the formatting for comments in emails is fine, the formatting for other field changes is often terrible13:10
programmerjakewell, ttyl, 5am here13:11
octaviusgn programmerjake13:11
lkcleek! :)13:11
lkcli'd suggest instead starting putting in a new book into the spec .tex13:12
lkclcontaining the primer13:12
octaviusyes, good idea13:12
lkcldon't trash the ability to build the primer, though13:13
octaviuslkcl, are you able to compile the spec?13:20
octaviusCurrently I'm getting a 100 errors, but I can't identify the origin file13:20
octaviuslooks like simple_v_spec.toc13:21
octaviusline 681 has too many brackets {} XD13:22
lkclmake clean13:26
lkclyour first response on compile errors: make clean13:27
octaviusalready done that13:27
lkclbear in mind that "proper" latex build makefiles are fantastically complex13:27
lkclbecause they involve generating things like a toc13:27
lkclwhich needs to be re-read in a 2nd pass13:28
lkclwhich then needs to be read in a 3rd pass13:28
lkclso you *may* need to do "make pdf" *three times*13:28
lkclit's so complex that "proper" latex Makefile creation has a special script to help do it properly13:28
octavius10 times, same problem13:28
octaviusIs it working for you?13:29
lkcl1 sec13:30
lkclyep all good here13:32
octaviusgit reset, git pulled, cleaned, make pdf again 3 times, not working13:33
octaviusAlso updated openpower-isa repo13:33
octaviushmm, I'll try updating my packages I guess13:33
lkclno don't do that13:35
lkclstick with the known-good debian/10 set13:36
octaviusI meant apt update13:36
lkclyes.  don't do that13:36
octaviusUnfortunately I started it before you mentioned that13:36
lkclaside from anything if you are on debian/testing it will go horribly wrong13:36
lkclstop it right now13:37
lkclkill it before it starts the installs13:37
lkclif it's during the download you're fine13:37
octaviusAh ok13:37
lkclyou really, *really* must not do a full system upgrade except as a scheduled event13:38
lkclduring which time you expect to completely do absolutely nothing but deal with the potential problems that result13:38
octaviusOk, well I can't get the tex to compile13:38
lkcli.e. you must terminate all and any work, potentially for several days or weeks13:38
lkclstart looking back through git commit history13:38
lkclfind the last point at which it worked using either git bisect or do it manually.13:39
lkcli had the same error occur last week, it went away after i did "make clean"13:39
octaviusI agreed to meet with a friend for lunch, so I'll do some testing once I'm back.13:40
lkclyou also need to actually look at the log file rather than treat the log file itself as a "black box"13:40
octaviusThese are the lines I was planning to add:13:40
octavius\part{Scalable Vectors Primer}13:40
octaviusspeak to you in a bit13:40
lkcli think that would work great.13:45
octaviuslkcl, tried a commit from 6th Jul (last time I worked on spec), still the same problem of too many brackets {}. Which version of texlive-full are you using? Maybe at some point I installed the testing version15:38
lkcloctavius, use a new chroot15:41
lkcldpkg -l | grep texlive15:42
lkclii  texlive                                          2017.20171128-115:42
lkclbut i seem to recall there's some package needed15:42
lkcltemp files get created.... hang on...15:42
lkclrright.  i know.  do this:15:42
lkclrm *.aux *.out *.toc15:43
lkclyou could have inspected the log file and found that out.15:44
lkcltook me a while to remember that's what it was.15:44
octaviusI just wanted to check what your version was. Mine is from 2018, so probably mucked up the packages at some point. Hopefully the new chroot will fix that15:45
octaviusAh, removing the temp files worked15:46
octaviusPerhaps I should add "rm *.aux *.out *.toc" to make clean?15:46
octaviusSorry about that15:47
octaviusAh you did already15:47
octaviusProblem I just realised, is that latex doesn't allow including files above the root directory15:59
octaviusSo either the primer dir has to be moved where simple_v_spec.tex is, or a symbolic link to the primer summary has to be added (but that doesn't include the images)16:01
lkclyep move the primer, in its entirety16:10
lkcland remember to update this16:11
lkclopenpower/sv/overview.mdwn:![Single-Issue concept](/svp64-primer/img/power_pipelines.svg)16:11
lkclopenpower/sv/overview.mdwn:![Multi-Issue with Predicated SIMD back-end ALUs](/svp64-primer/img/sv_multi_issue.svg)16:11
lkclopenpower/sv/overview.mdwn:![image](/svp64-primer/img/svp64_regs.svg){ width=40% }16:11
lkcland check as well16:11
lkcldon't mess about, just move the whole lot into openpower/sv16:12
* lkcl still doing the table16:12
octaviuslkcl, moved svp64-primer and added primer to the spec16:51
octaviuscurrently it occurs before the Preamble and TOC, probably worth thinking about where to best place it16:52
lkclthat should work well18:41
lkcltplaten, the gram headless example should run under iverilog.18:42
lkclmake sure to use this though;a=blob;f=iverilog-install;hb=HEAD18:43
lkcloctavius, hmm the preamble needs to come first18:47
lkclalso hmm adding the summary has broken something18:49
octaviusAdding references didn't work either18:49
lkcllet me see if i can work it out18:51
lkclsuspect missing usepackage{}s18:56
octaviusAh, hm18:56
lkclgot it18:57
lkclcommittting in 1min...18:57
octaviusAdded the acronyms as well19:03
octaviusAlso in the primer, section 1.2.4, where I listed the hybrid workloads (CPU, GPU, VPU, 3D?), I left question mark. Possibly because I didn't know if the point was explained enough19:06
lkclyeah leave out questionmarks like that, it causes lack of confidence19:08
lkclacronyms good19:09
lkclacronyms need to go before heading, and without a number.19:11
lkcli know how to do that (use *)19:11
octaviusSure. I noticed page 59 (sv register example fig) is stretched19:12
lkcli did put in a width-scaling somewhere19:16
octaviusShall I add a contact section after the summary?19:16
lkclyes.  in the preamble19:18
octaviuswill do19:18
tplatenI've read that install manual. An its not the first time that I installed and used iverilog, that is maybe ten years ago or so.19:25
tplatenI only see ecp5 modules here, but I know that there is an emulation of the ecp5 ddr phy in iverilog19:28
tplatenls shows19:28  firmware  headless  uartbridge.py19:28
tplatendone, installed iverilog19:49
octaviuslkcl, added the contacts section. Not sure if the description is correct, or if everyone needs to be on there19:59
octaviusI was thinking konstantinos might be good to add given his experience with svp6420:00
lkclyes good idea21:30
octaviuslkcl, I didn't put konstantinos under the Libre-SOC team, is that correct? I seem to remember he works for a different organisation (but I guess we all do to some extent)21:45
lkclhis company is Vectorcamp.gr21:51
octaviusThanks, added his info21:58
octaviusTable looks really nice :D23:33
octaviuslkcl, may be worth to check with david and toshaan, but sections 3.1/3.2 (Other Vector ISAs) maybe a little sarcastic (I don't think it's drastically so, but worth checking)23:40

Generated by 2.17.1 by Marius Gedminas - find it at!