Saturday, 2021-08-21

lkclok that all seems happy (something borked with /etc/sudoers.d/{chrootname})00:00
lkclmikolajw, did a file /etc/sudoers.d/{yourchrootname} get created for you?00:00
mikolajwlkcl: yes00:02
lkclok good, whew00:02
lkcl./hdl-dev-repos should work fine then00:03
lkclbut do it under the user account (non-root) and let it run "sudo bash" for you.00:03
lkclsorry ... :)00:03
lkclrun "sudo bash" as the user, first00:03
lkcl*then* run the script00:04
lkclthat way, pip3 is run system-wide and the packages it installs are available to either user-run python or root-run python00:05
mikolajwlkcl: sent the key00:38
lkclgot it - you've not joined the list, i'll approve the message you sent, so it gets through.00:41
lkclmikolajw: added you to dev-env-setup and the wiki (yes, named libreriscv for historic reasons)00:42
mikolajwlkcl: pushed00:58
mikolajwfirst contribution complete00:58
lkclveera[m], those images look absolutely great, really clear. can you create an RfP based on the previous one you sent? send me a copy for review (without your bank details)11:13
veera[m]I will make a RFP. But can you correct or check in the Bugzilla NlNet Milestone, Total Budget and for task and Parent Task?11:23
veera[m]Bugs are 626, 634, 65111:24
veera[m]lkcl, eh ping11:25
lkclveera[m], yes, i'll deal with the bugzilla budget balancing12:04
veera[m]okay. I send a draft RFP to you12:08
Madan_KartheessaHi Sukhanshu12:09
Madan_KartheessaHi Mehul12:09
ghostmansdlkcl: I hope to dedicate some hours tomorrow; what's the next task we have? IIRC we discussed dynamic register width recently; do we have task on Bugzilla?12:19
lkclmatthewcroughan, Sukhanshu, hi, welcome12:33
lkclghostmansd, yyeah that one's going to be fun12:34
lkclthe entire pseudocode needs to be altered to use XLEN (which should default to 64)12:34
lkclcomplicating factor: in LSB0 numbering it's dead easy12:35
lkclRS[0..63] --> RS[0..XLEN-1]12:35
lkclMSB0 numbering... sigh if you have something that requires e.g. the least significant byte12:35
lkclthat's RS[55..53] in MSB0 numbering12:36
lkclsorry RS[56..63]12:36
lkclannoyingly becomes RS[MSB-8..MSB-1]12:36
ghostmansdCouldn't we define some macros?12:36
lkcli'll get it right eventually12:36
lkclnot in specification pseudocode, no.12:37
ghostmansdlol, hope I will too12:37
lkclwell, The Unit Tests Are Your Friend12:37
ghostmansdBut the specification doesn't talk of XLEN as well, does it?12:37
lkclnoo it doesn't - which means it needs to be added12:37
lkcli've been speaking to Paul Mackerras, the OPF ISA WG Chair, for several months12:38
lkclgiving him a heads-up12:38
ghostmansdDo you mean we'll send our updates to spec?12:38
lkclif the ISA WG...12:38
lkclif the ISA WG doesn't accept them, well, they don't accept them, and we have a divergent fork of the pseudocode12:38
lkclwe just have to live with that12:38
ghostmansdBut, if we're talking of updating the specs anyway... Couldn't we update it with macros?12:38
ghostmansdThey use pseudo-functions, e.g. this BCD/DPD conversion12:39
lkclthe issue then is that we actually have to teach the pseudocode compiler about macros12:39
lkclremember though, each change has to be justified, going through an Official Voting process that we're not even involved in the decision-making12:40
lkcland it operates entirely behind closed doors.12:40
lkclwe can't even see the non-voting *discussion* about proposed ideas.12:40
ghostmansdlkcl: I've been thinking of ISA functions12:49
lkclthey'll all need to be converted to "self.fn" btw12:50
lkclall the helpers. sigh12:50
ghostmansdBut OK, I'll keep it in mind. I like the functions since they keep the code similar, so that we don't need to fi the code everywhere in case the semantics are changed.12:50
lkclyehyeh good point12:51
lkcland the c compiler won't need hand-written variants12:51
lkclor, shouldn't12:51
lkcloooo annoying, the BNF doesn't support expressions in ranges12:57
mikolajwwiki points to a script yosys-et-all that doesn't appear to exist anymore. I think it was renamed to hdl-tools-yosys?13:26
mikolajwyup, that's what git says13:28
lkclmikolajw, nicely done13:42
ghostmansdlkcl: both issues are assigned to you; should I transfer the ownership?14:45
ghostmansd(I'll do it tomorrow, I'm AFK today)14:45
ghostmansd(except for IRC, which banned my IP again today, sigh)14:45
lkclyes do15:13
mikolajwshould I use upstream nmigen or from libre-soc repos?19:23
lkclmikolajw, mmm.... you shouuuld be ok with upstream.19:30
lkclthe only reason we use our own copy is to be able to apply tags to it for the 180nm ASIC19:30
lkclwe needed to pin a very specific version so as to have reproduceable builds19:30
mikolajwI think it would be good to say this in the wiki19:40
mikolajwthis appears outdated:
mikolajw"yosys -p "read_ilang; show dec31"" didn't work, there was no file created anywhere in my fs20:54
mikolajwand says "# moved to openpower-isa"20:55
mikolajwand, besides that, how do I start with development? Any example commands to get some results immediately?20:57
mikolajwalso, from `make test` in soc repo:22:21
mikolajwmake: svanalysis: Command not found22:21

Generated by 2.17.1 by Marius Gedminas - find it at!