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.
dogbert11 t/10_base.t .......... ===SORRY!=== Error while compiling /home/dogbert/repos/CSV/t/10_base.t 12:19
===SORRY!=== Error while compiling /home/dogbert/repos/CSV/lib/Text/CSV.pm (Text::CSV)
MoarVM panic: Adding pointer 0x564e120de940 to past fromspace to GC worklist
dogbert11 tries to repro 12:20
japhb has ethernet cabling flashbacks at the mention of 10_base.t 20:01
timo ahaha :D 20:06
i wonder if support for the dispatcher-related stackframe bits would be good to have right now in the debugserver 20:13
it feels kind of not so great that when outputting the spesh log for an empty-ish program i get a billion invocations in the stats for frames from Exception.pm6 20:20
that's just the class bodies for the exception classes
i seem to recall this is something that rakuast can improve? 20:21
oh, sp_dispatch_o completely bailed jitting a frame, that means in theory adding that to the jit could make something somewhere faster and it should be easy to do so i've got something to occupy my time yay 20:26
why on earth is ack giving me results from a .o file 20:28
huh that's a lot of code, eh 20:31
timo there'll also want to be some handling for var-args ops perhaps 20:48
not being able to "make install" is kind of annoying, since ./rakudo-gdb-m doesn't seem to be very happy 21:03
timo i assume when we're left with an sp_dispatch_* we're already in a low performance situation anyway 22:53
my brain is kind of feeling a little noodly
tbh i don't even know what kind of code to run to test stuff :D 22:58
No resumable dispatch in dynamic scope 23:00
at SETTING::src/core.c/Mu.pm6:647 (/home/timo/raku/rakudo/blib/CORE.c.setting.moarvm:Str)
from gen/moar/World.nqp:3023 (/home/timo/raku/rakudo/blib/Perl6/World.moarvm:add_constant)
well, that looks kinda funny? (this is when trying to run json_fast's Fast.pm6
or rather compiling i guess)
30% interp_run, 12.6% disp_program_run :D 23:05
timo something seems to be seriously wrong with spesh, or with the spesh log ?!? 23:16
there's barely anything in it? in terms of being speshed? 23:17
the filename of my script doesn't appear once, so no functions or frames from in there
japhb timo: If you're trying to use it with new-disp, that's because (as jnthnwrthngtn said) spesh doesn't understand much of new-disp yet.
And there's no inlining yet either 23:18
timo ok, but this is almost nothing?
that's wild
japhb *shrug*
He says it's currently running about half speed even so. 23:19
timo OK fair enough
perhaps it was a little too much hubris to just jump in the code and think i can contribute something without asking around first 23:21
japhb timo: I think he said he's going to have time next week to work on it, so he may get it to the point that you can contribute (I'd ask, at least) 23:30
timo right 23:33