Welcome to the main channel on the development of MoarVM, a virtual machine for NQP and Rakudo (moarvm.org). This channel is being logged for historical purposes.
Set by lizmat on 24 May 2021.
vrurg nqp's HLL::World is using @*comp_line_directives. But where is it declared? 15:24
nine vrurg: I think it isn't anymore since commit 0d4885d7450b22ccf2000823f9c2d7908fbe9019 16:04
MasterDuke vrurg: @*comp_line_directives was the first not-completely-trivial thing i did for moarvm/nqp/rakudo, so i'm sure it can be improved 17:10
btw, gist.github.com/MasterDuke17/97950...a030abcc54 is what i'm experimenting with right now. anybody have any thoughts on whether i should or should not PR that (and any further similar changes)? 17:12
for compiling CORE.e, it's reduced temporary allocations from 272k down to 124k 17:14
vrurg nine, MasterDuke: then how does `#line` currently works? I only found it being handled by HLL::Grammar/Actions where it refers back to HLL::Compiler which is using the dynamic. 19:57
timo are we certain that 16 is the longest capture path we will ever encounter? 20:16
nine vrurg: in RakuAST it doesn't at all. See also irclogs.raku.org/moarvm/2022-11-20.html 20:18
MasterDukeMobile timo: certain, no. But I saw a max of only 13 when compiling CORE.c, so 16 was sufficient for experimenting 20:33
Actual values would definitely be up for bikeshedding
timo shrug, stack is big :P 20:42
vrurg nine: that's what I'm working on. I encountered a bug yesterday which is hard to track down to exact location in a source. So, decided to do something about it. 21:09
nine MasterDuke: if we can't guarantee that we never run out of room there, I'd say it's not worth it. Those buffer overruns would be painful to debug. What do we gain in terms of time in return? 22:27