Friday, 2023-09-08

*** gnucode <gnucode!~gnucode@user/jab> has quit IRC00:31
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc00:31
*** psydroid <psydroid!~psydroid@user/psydroid> has quit IRC01:57
*** Lucretia <Lucretia!~laguest@user/lucretia> has quit IRC03:09
*** Lucretia <Lucretia!~laguest@user/lucretia> has joined #libre-soc03:13
*** Lucretia <Lucretia!~laguest@user/lucretia> has quit IRC03:20
*** Lucretia <Lucretia!~laguest@user/lucretia> has joined #libre-soc03:32
*** Lucretia <Lucretia!~laguest@user/lucretia> has quit IRC04:29
*** Lucretia <Lucretia!~laguest@user/lucretia> has joined #libre-soc04:50
*** ghostmansd <ghostmansd!~ghostmans@213.87.137.131> has joined #libre-soc05:44
*** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@176.59.174.147> has joined #libre-soc05:46
*** ghostmansd <ghostmansd!~ghostmans@213.87.137.131> has quit IRC05:49
*** Lucretia <Lucretia!~laguest@user/lucretia> has quit IRC07:07
*** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@176.59.174.147> has quit IRC07:08
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc07:12
openpowerbot[irc] <lkcl_> programmerjake, can i suggest that you take pow(x,y,mod), implement a test_caller_svp64_*.py for it, use parallel-prefix to create the sequence of 1 x x^2 x^3 x^4 etc. *only* if that is going to result in a small algorithm (no LD/ST, fixed pow-length fitting in regs)07:20
openpowerbot[irc] <lkcl_> and once that's done - as a REAL short unit test on which you should spend an absolute maximum of 6-8 hours - move on to bigmul REMAP as a "related subtask"07:22
openpowerbot[irc] <lkcl_> under the grant-end timeconstraint unfortunately we're not going to be able to get shriya up to speed on svp64.07:23
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC07:31
*** Lucretia <Lucretia!~laguest@user/lucretia> has joined #libre-soc07:32
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc07:36
openpowerbot[irc] <programmerjake> if we use shift-sub division instead of knuth's algorithm d, i might be able to get bigint modular exponentiation working in a few days to a couple weeks, did you see the budget estimates i put in https://bugs.libre-soc.org/show_bug.cgi?id=1153#c007:42
openpowerbot[irc] <programmerjake> those are higher since 2048-bit is big enough to need more complexity, but probably by no more than 1.5-2x07:43
openpowerbot[irc] <programmerjake> note that parallel prefix is not how to do modular exponentiation, instead you do one bit at a time of the exponent and repeatedly square and multiply, all multiplications must be followed by a remainder operation.07:48
openpowerbot[irc] <programmerjake> https://en.wikipedia.org/wiki/Modular_exponentiation#Implementation_in_Lua07:52
openpowerbot[irc] <programmerjake> like above, except using bigints07:53
openpowerbot[irc] <snowwolflair[m]> What to do with good ideas that are not of the moment? You will see I have recomended we capture all good ideas to be reviewed when there are less pressing priorities like completing grants before they expire.  FYI the new bug is https://bugs.libre-soc.org/show_bug.cgi?id=115208:14
*** JTL <JTL!~jtl@user/jtl> has quit IRC09:00
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC09:02
*** JTL <JTL!~jtl@user/jtl> has joined #libre-soc09:08
*** octavius <octavius!~octavius@92.40.169.9.threembb.co.uk> has joined #libre-soc09:12
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc09:15
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC09:23
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc09:27
*** psydroid <psydroid!~psydroid@user/psydroid> has joined #libre-soc09:49
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC10:22
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc10:27
*** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@176.59.56.241> has joined #libre-soc11:14
*** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@176.59.56.241> has joined #libre-soc11:14
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC11:15
*** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@176.59.56.241> has quit IRC11:16
openpowerbot[irc] <lkcl_> programmerjake, do not do LD/ST, do not do anything other than shift-sub and sub-loop for modulo-check, keep it BRUTALLY simple such that the answer *IS* "days" (absolutely not "weeks")11:25
openpowerbot[irc] <lkcl_> and move as quickly as you possibly can onto implementing bignum REMAP11:25
openpowerbot[irc] <lkcl_> fixed-width 256-bit pow-mod demo is perfectly fine.11:25
openpowerbot[irc] <lkcl_> if we then have time (or can find other budgets) then a carry-save bignum REMAP is also "on the cards", particularly given that both poly1305 and ed25519 need it11:26
openpowerbot[irc] <lkcl_> but please *always raise the bugreport* do not wait for me to constantly have to prompt you11:27
openpowerbot[irc] <lkcl_> that goes for everyone11:27
openpowerbot[irc] <lkcl_> https://bugs.libre-soc.org/show_bug.cgi?id=1044 - assigned to you, jacob.11:27
openpowerbot[irc] <sadoon[m]1> Hi guys, had a break yesterday will take time today to catch up11:48
openpowerbot[irc] <sadoon[m]1> I found the perfect solution to my email issues that does not involve protonmail bridges11:48
openpowerbot[irc] <sadoon[m]1> StartMail allows you to use your own domain name, which means they host my email using my domain and nothing has to change with my contact details11:49
openpowerbot[irc] <sadoon[m]1> Very awesome, only 2.85$ a month, going to do it in a bit11:50
openpowerbot[irc] <sadoon[m]1> Can also probably use my old email as another alias too11:50
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc12:00
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC12:03
*** ghostmansd <ghostmansd!~ghostmans@37.204.162.147> has joined #libre-soc12:12
openpowerbot[irc] <sadoon[m]1> My email should now work12:18
openpowerbot[irc] <sadoon[m]1> Yep, mxtoolbox verifies that it does12:19
openpowerbot[irc] <sadoon[m]1> Again sorry for the annoyance everyone12:19
openpowerbot[irc] <sadoon[m]1> <lkcl_> "caffeine triggers your liver..." <- That's actually interesting, will read into it thanks :)12:41
openpowerbot[irc] <sadoon[m]1> <snowwolflair[m]> "What to do with good ideas..." <- May I suggest a wiki page with sections for the main distinction of an idea (say FPGA, HDL, Simulator) and subsections for the ideas (say FPGA> run debian-sffs, HDL> implement so-and-so, Simulator> speedups)12:48
openpowerbot[irc] <sadoon[m]1> We could call it libre-soc.org/brainstorm12:49
openpowerbot[irc] <snowwolflair[m]> sadoon[m]1: 👍️ Whatever work, but capturing ideas is paramount.12:49
openpowerbot[irc] <sadoon[m]1> Great, I might start with that myself after I commit the verilator optimizations and start a proper plan for poly1305, also have some $DAY_JOB things to finish up so it's quite the busy weekend :)12:50
openpowerbot[irc] <snowwolflair[m]> sadoon[m]1: Or Libre-soc/havingabrainstorm 🤣12:51
openpowerbot[irc] <sadoon[m]1> I prefer brainstorm as it is shorter and easier to link to12:51
openpowerbot[irc] <sadoon[m]1> But either works12:51
openpowerbot[irc] <sadoon[m]1> Weird that I can't see emojis on my workstations, just saw that it was a laugh sigh12:53
*** ghostmansd <ghostmansd!~ghostmans@37.204.162.147> has quit IRC14:30
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc14:30
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.56.241> has quit IRC15:32
*** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC15:32
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.160.66> has joined #libre-soc15:32
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.160.66> has quit IRC15:51
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.164.225> has joined #libre-soc15:54
openpowerbot[irc] <sadoon[m]1> lkcl  I can't seem to push my new branch to microwatt.git, the git test passes though "ssh -v -p922 gitolite3@git.libre-soc.org"15:54
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.164.225> has quit IRC16:12
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.40.123> has joined #libre-soc16:28
openpowerbot[irc] <lkcl_> sadoon[m]1, what's the line in ".git/config"? i bet it's "https", isn't it? read HDL_workflow for the format, and make sure you have in ~/.ssh/config a Port 922 entry16:32
openpowerbot[irc] <lkcl_> it's all there https://libre-soc.org/HDL_workflow/#index7h116:33
openpowerbot[irc] <lkcl_> snowwolflair[m], #havingabadhairday16:33
openpowerbot[irc] <lkcl_> sadoon[m]1, "caffeine triggers your liver..." - bear in mind also that the actual caffeine which is mildly toxic gets stored *instead* in your liver, which then turns into bile, and dumps that into your intestines, later.16:35
openpowerbot[irc] <sadoon[m]1> <lkcl_> "sadoon, what's the line in ".git..." <- Yep I already followed the steps before which is why this is weird16:45
openpowerbot[irc] <sadoon[m]1> Host git.libre-soc.org16:45
openpowerbot[irc] <sadoon[m]1> Port 92216:45
openpowerbot[irc] <sadoon[m]1> And I already pushed changes to dev-env-setup from this same laptop with no issues16:45
openpowerbot[irc] <programmerjake> lkcl: sadoon doesn't have write permissions: the list of those that do:16:58
openpowerbot[irc] <programmerjake> admin adminbigmac adminfizzy adminbizi tpearson jacob1 jacob2 tobias andreym16:58
openpowerbot[irc] <sadoon[m]1> Yeah now it makes sense :)16:59
openpowerbot[irc] <cesar[m]1> sadoon, if you run "ssh gitolite3@git.libre-soc.org" it will print a list of your permissions on all repositories.17:08
openpowerbot[irc] <programmerjake> or, just clone ssh://gitolite3@git.libre-soc.org:922/gitolite-admin.git and read conf/gitolite.conf17:12
openpowerbot[irc] <sadoon[m]1> yep, I don't have write access17:15
openpowerbot[irc] <lkcl_> sadoon[m]1, you do now. both keys17:20
openpowerbot[irc] <sadoon[m]1> Awesome, thanks, pushed to branch "optimizations" (note this is based on verilator_trace obviously)17:21
openpowerbot[irc] <sadoon[m]1> Curious to see if you guys get any sizable improvements as well17:21
openpowerbot[irc] <lkcl_> sadoon[m]1, that's a branch off of "microwatt_verilator"? if you use anything else YMMV17:47
*** ghostmansd <ghostmansd!~ghostmans@176.59.40.123> has joined #libre-soc17:49
lkclghostmansd, thank you for asserting the value of your work. it means there's a clear justification if NLnet or the EU Auditors query it.17:50
*** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC17:51
*** markos_ <markos_!~markos_@static062038151250.dsl.hol.gr> has joined #libre-soc17:52
*** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc17:52
*** gnucode <gnucode!~gnucode@user/jab> has quit IRC17:56
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc17:57
openpowerbot[irc] <sadoon[m]1> <lkcl_> "sadoon, that's a branch off of..." <- I don't see a branch named microwatt_verilator, the one mentioned in the wiki is verilator_trace and so I based on that18:00
*** choozy <choozy!~choozy@75-63-174-82.ftth.glasoperator.nl> has joined #libre-soc18:05
*** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC18:14
*** markos_ <markos_!~markos_@static062038151250.dsl.hol.gr> has joined #libre-soc18:27
*** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc18:27
*** openpowerbot <openpowerbot!~openpower@94-226-187-44.access.telenet.be> has quit IRC18:28
*** openpowerbot <openpowerbot!~openpower@94-226-187-44.access.telenet.be> has joined #libre-soc18:41
octaviussadoon, there is no 'microwatt_verilator', I remember asking a similar question to you. Perhaps sometime ago there was, but 'verilator_trace' is the branch you should be using for libre-soc18:49
openpowerbot[irc] <sadoon[m]1> Yes that's the one I used :)18:49
openpowerbot[irc] <sadoon[m]1> The commits should be merge-able18:50
openpowerbot[irc] <sadoon[m]1> The only edited file is Makefile18:50
octaviusGood. Thanks again for finding this optimisation, I'll have to remember to pull before doing any more sims18:50
openpowerbot[irc] <sadoon[m]1> Of course :D18:50
*** gnucode <gnucode!~gnucode@user/jab> has quit IRC18:51
openpowerbot[irc] <sadoon[m]1> I haven't tested which ones specifically make it this much faster, could be a combination or just a single one, I'll have to do more testing of course, as well as document this in a bug report once I'm free18:51
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc18:51
openpowerbot[irc] <sadoon[m]1> Currently researching Kubernetes so I don't look stupid in my $DAY_JOB18:52
octaviusWait, the git.libre-soc.org website shows a separate branch called 'optimisations'18:52
octaviusin the microwatt repo18:52
openpowerbot[irc] <sadoon[m]1> Yes18:53
openpowerbot[irc] <sadoon[m]1> I didn't want to mess with the known-working one18:53
octaviusWas that the intention?18:53
octaviusAh ok18:53
openpowerbot[irc] <sadoon[m]1> We can merge once we're sure it's all good18:53
octaviusAh fair enough18:53
openpowerbot[irc] <sadoon[m]1> "Never git push to upstream on a Friday" or so it were :P18:54
octaviusAh yes, sounds like a reasonable way of thinking XD18:54
openpowerbot[irc] <sadoon[m]1> Agh got bored of k8s (fast) will work on testing the different verilator options and poly1305 instead19:08
openpowerbot[irc] <programmerjake> sadoon: note that -mcpu is what generally works, -march for setting the cpu is mostly x86-specific19:34
openpowerbot[irc] <sadoon[m]1> For some reason using -mcpu on my intel laptop made the compilation quite a bit slower19:36
openpowerbot[irc] <sadoon[m]1> Plus the gcc warnings that it is deprecated are annoying19:36
openpowerbot[irc] <programmerjake> actually, now that i'm looking that up, it seems -march is more common...-mcpu is also used on arm19:37
openpowerbot[irc] <sadoon[m]1> If we have any devs using ARM servers we'll add that to the if statement then :P19:37
openpowerbot[irc] <sadoon[m]1> markos might consider it haha19:37
openpowerbot[irc] <sadoon[m]1> But I am seeing an anomaly that is bothering me, give me a bit19:38
openpowerbot[irc] <programmerjake> well, i sometimes do stuff on my aarch64 phone...19:41
openpowerbot[irc] <programmerjake> not that you need to add special arm code19:41
openpowerbot[irc] <sadoon[m]1> Reminds me of testing poly1305 on my Galaxy S10e to compare with an FPGA, fun days19:44
openpowerbot[irc] <sadoon[m]1> It was userland or whatever it was called19:44
openpowerbot[irc] <programmerjake> i use termux for that19:44
openpowerbot[irc] <sadoon[m]1> Feels like ages ago, was only a little less than two years ago19:44
*** gnucode <gnucode!~gnucode@user/jab> has quit IRC19:45
openpowerbot[irc] <sadoon[m]1> Yeah termux works too19:45
openpowerbot[irc] <programmerjake> https://termux.dev/en/19:45
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc19:45
openpowerbot[irc] <sadoon[m]1> Funnily I got invalid data only on that phone19:45
openpowerbot[irc] <programmerjake> weird...19:45
openpowerbot[irc] <programmerjake> reminds me of https://www.mono-project.com/news/2016/09/12/arm64-icache/19:46
openpowerbot[irc] <sadoon[m]1> oh boy19:47
openpowerbot[irc] <sadoon[m]1> I can't seem to replicate the performance gains for some reason, at least on hello_world..19:47
markos_nowadays -march is the norm and -mcpu is the exception20:01
lkcloctavius, thank you - yes i forget the names20:06
octaviusnp20:10
openpowerbot[irc] <sadoon[m]1> Retesting full linux boot without the optimizations but still on 3 threads, let's see how this goes20:13
openpowerbot[irc] <sadoon[m]1> Bah, ran out of disk space21:27
octaviusghostmansd, I created a new bug #1154, funded by the cavatools grant, for the task of autogenerating the binutils code21:41
octaviusplease adjust the description as you see fit (since I only have surface level understanding)21:41
octaviusHopefully 2500EUR is enough at least for an initial exploration21:41
*** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC21:48
*** markos_ <markos_!~markos_@static062038151250.dsl.hol.gr> has joined #libre-soc22:02
*** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc22:02
*** choozy <choozy!~choozy@75-63-174-82.ftth.glasoperator.nl> has quit IRC22:08
*** octavius <octavius!~octavius@92.40.169.9.threembb.co.uk> has quit IRC22:22
ghostmansdoctavius, you triggered me for a large post on the mailing list, I had so many unstructured thoughts on insndb that I have to use that format22:44
ghostmansdThank you for bug #1154; I'll adjust the description so that the task affects only vanilla assembly; for SVP64 we have a damn big prerequisite, it'd be a waste of time to even start generating SVP64 code w/o this prerequisite22:46
ghostmansdguys, when you have time, please check this thread: http://lists.libre-soc.org/pipermail/libre-soc-dev/2023-September/005633.html22:46
ghostmansdit's not urgent, but critically relevant code-generation-wise22:46
*** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@37.204.162.147> has joined #libre-soc23:37
*** ghostmansd <ghostmansd!~ghostmans@176.59.40.123> has quit IRC23:41
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.40.123> has quit IRC23:56
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@37.204.162.147> has joined #libre-soc23:56

Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!