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.
nine Shouldn't we abandon November? I mean, it's been December for several days now 07:26
Nicholas good *, #moarvm 07:42
jnthnwrthngtn Maybe the module stops working at the month boundary... :-) 10:06
moarning o/
Nicholas \o 10:08
lizmat hmmm... looks like I borked Geth 10:33
Nicholas quick, everyone commit Easter Eggs whilst nothing is watching 10:34
lizmat Too late! 10:36
MasterDuke just performed an interesting experiment. i replaced the `value % 10` on this line with lemire's fastmod implementation from lemire.me/blog/2019/02/08/faster-r...libdivide/ 10:38
*this line github.com/MoarVM/MoarVM/blob/mast...erce.c#L43
but `my str $s; my $n = now; for ^10_000_000 -> int $_ { $s = $_ }; say now - $n; say $s` was ~0.1s slower (~0.53 before, ~0.63 after) 10:39
lizmat ok, Geth and the RakuIRCLogger should now be more resilient to losing connection to the IRC server 10:43
if the client now does not receive a PING within 10 seconds of when a PING is expected, it will do a reconnect to the IRC server 10:44
Geth MoarVM: MasterDuke17++ created pull request #1618:
Use faster itoa implementation		 14:25
MasterDuke oh interesting, it looks like gmp can use custom allocators, so we could use the fsa for them... 16:51
s/them/bigints/ 16:52
nine On darn...we've already got 8 different callsite flags 19:10
But interesting. The comment above MVM_CALLSITE_ARG_INT's definition has read "Argument is a native integer, signed." from the start. 19:12
And there used to be a MVM_CALLSITE_ARG_UINT
Which got removed in commit 873f102b8a18251 because "We just treat it as int. The only place the distinction matters is in generating certain instructions that work differently, and we don't do this distinction elsewhere in the VM, so it's silly to have it in this one place, not to mention useless." 19:14
Well, useless no more...
nine I wonder why MVM_CALLSITE_ARG_FLAT_NAMED got its own bit when there's MVM_CALLSITE_ARG_FLAT and MVM_CALLSITE_ARG_NAMED and the whole point of using bits like this is that they can be combined 19:16
Looks like it won't be difficult to free up that flag. 19:23
Except that of course it causes bootstrapping issues... 19:32
[Coke] nine, I'll buy you a curry if we also get a blog post on this huge change. :) 19:58
(that is probably not sufficient reward. :) 20:00
vrurg_ Let's live thug life: two currys and a pizza? :) 20:24
[Coke] Sold. 20:27
Oh wait, are you their agent? :)
vrurg_ is insidiously chuckling... 20:56

Powered by Cro and the Raku® Programming Language.

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

Thank you!