github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
samcv the coverage git has gotten pretty big 02:03
brrt .tell lizmat I saw one mention that said that 'perl6 on MoarVM is starting to grow a JIT" 06:28
yoleaux brrt: I'll pass your message to lizmat.
lizmat .
yoleaux 06:28Z <brrt> lizmat: I saw one mention that said that 'perl6 on MoarVM is starting to grow a JIT"
brrt ohai lizmat
lizmat yeah, I saw it too :-)
I think the "starting to grow a JIT" is not doing the MoarVM JIT justice, though 06:29
brrt heh, maybe it'd be worthwhile to have a post that says 'Perl6 has a JIT, and it is useable and in production, read here for more'
(and I'd include all of spesh into that discussion)
lizmat be my guest, although maybe a simple comment with a link would suffice
brrt meh, nobobdy will see that 06:30
the whole discussion is ... light on facts unfortunately
(probably nearly everybody else who knows how to hack one is busy actually doing so :-P)
i did think of the following though 06:31
you know how all problems in CS can be solved by adding another layer of indirection?
and how layers of indirection are the worst thing for performance?
well, perl6 solves a lot of problems for you ;-)
Geth MoarVM/jit-expr-refactor: cf775ddf41 | (Bart Wiegmans)++ | 5 files
[JIT] Remove 'value type' enum

Only remaining use for it is to distinguish C arguments (and our use there is not really required either since we already maintain the VM type).		 06:46
brrt another one bytes the dust...
I've decided I'm moving in the direction of making the expression template compiler (perl) smarter and the expression template application (c) dumber 06:47
samcv well i got the docs building on travis finally heh 06:55
sadly everything on travis seems broken. gradle was broken. and asciidoctor was broken too. so had to dl both those. (this will only be in our coverage branch that we'll end up doing this extra stuff) 06:56
and openjdk 7 isn't compiled with enough crypto algos so gradle was failing because it couldn't access the https page :P 06:58
lizmat any objections against me bumping Moar / NQP ? 08:19
samcv none here 08:40
brrt lizmat: what would be brought in? 11:05
brrt I think I can move cast insertion to the expr template compiler as well, with a bit of work 12:34
If so, and if I can encode type + size in the expr compiler, I can completely remove MVM_jit_expr_tree_analyze 12:35

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!