Sunday, 2021-08-22

lkclmikolajw: this means you haven't run the dev-env-setup script10:40
lkclah, yep, found a spelling error (on a command that is never used because it's run from a different location)10:41
lkclgenerally we don't run "make test" unless having about 2 days spare to wait for it to complete10:43
lkclyep, that's out-of-date, i moved to openpower-isa several months ago10:44
lkcltherefore it does nothing10:44
lkcli've removed svanalysis from the soc Makefile because it's part of openpower-isa (moved several months ago)10:53
lkcladded a page with links to nmigen tutorials11:00
ghostmansd-pclkcl: shouldn't we also add XLEN to parser?17:15
ghostmansd-pcI see that you updated ISACaller namespace with XLEN=64, but, since the whole pseudocode is parsed before ISACaller enters the scene, we should teach parser and lexer XLEN, shouldn't we?17:17
veera[m]Hi lkcl17:17
veera[m]lkcl: Did RFP mail reached you?17:25
lkclghostmansd-pc, it's treated as simply "yet another variable"18:46
lkclso, no18:46
lkclveera[m]: yes.... i spam18:47
lkclveera[m], you need to adjust the Phase/Milestones18:48
lkclyou have the old field :Phase/milestone(s): 384 389 626 634 65118:48
lkclthat needs to be changed to: 626 634 65118:48
lkcloh ok you have 2 extra ones, 384 and 389 - take those out18:48
ghostmansdlkcl: already found it, thanks18:49
ghostmansdlkcl: let's spam in two places simultaneously :-D18:49
lkclghostmansd :)18:50
lkclveera[m], everything else looks good18:50
lkclsend it to the same 2 email addresses as last time18:50
ghostmansdlkcl: I _think_ that minus sign should not be part of number token18:50
ghostmansdI suspect should be handled at parser level18:51
ghostmansdi.e. the parser should differ (a - b) and (-b), based on context. In fact, if I drop the sign from regex, it looks like it already does. I haven't tried any complicated cases, though.18:52
ghostmansdlet me try something like XLEN - -118:53
ghostmansd-pcFWIW, since recent I see a lot of warnings from, about unused PowerDecoder object. Is it a known issue, or maybe this is not an issue at all?19:25
lkclyeh ignore it19:28
lkclthe reason is, you're supposed to use nmigen objects in a module19:28
lkcland one of the very common mistakes in nmigen HDL is to instantiate an object but not actually link it to anything19:29
lkclso whitequark added a check "is this nmigen object actually used, if no produce a warning"19:29
lkclbut we use nmigen objects in the *emulator*... without actually linking them to anything.19:29
lkclhence the warnings... which can be ignored19:30
mikolajwdoes anyone have a suggestion what could be a good first contribution to the SoC itself? I need something simple and self-contained. I have no prior experience in Nmigen and SoC design20:22
ghostmansd-pclkcl: can we disable it e.g. via some environment variable? If test fails, these appear below the log.20:38
lkclmikolajw, can i suggest going through the "first steps" page that ghostmansd-pc wrote last week?20:41
lkclghostmansd-pc, there is - i think - it's somewhere in nmigen source code20:43
lkclclass Elaboratable...20:44
lkclclass MustUse:20:46
lkcl    _MustUse__silence = False20:46
lkclin nmigen._unused.py20:46
lkclso you can monkey-patch nmigen._unused.MustUse and set _MustUse__silence=True20:48
lkclghostmansd-pc, yyeahh i was just experimenting with that MINUS comparison and yeah it seems to work21:05
ghostmansdYep, though I'm still far from understanding what it has to do with comparisons.21:10
lkclit's just a moniker21:17
lkclBNFs are often full of things that have the wrong name, esp. if they're hacked together21:17
lkclor, as is most likely in this case, two BNFs were merged into one, from a formerly "good" BNF that had too many features to implement in a simple example21:18

Generated by 2.17.1 by Marius Gedminas - find it at!