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.
Nicholas good *. * 05:06
strange bright thing in sky 05:33
nine I think I've seen that before 10:21
vrurg: I fail to see the difference in need for exception handling between low level locking hash and a high level implementation 10:24
jnthnwrthngtn The more general point is that we've already been down the "implement it in the VM, it'll be faster" and we already un-did a lot of that because it prevents spesh doing its job 10:27
Not to mention that writing stuff in HLL eliminates entire classes of bugs (can't get GC invariants wrong, for example) 10:28
MasterDuke i tried turning github.com/rakudo/rakudo/pull/4891...96f32R3344 directly into a call to `mul_I`, by making it `return QAST::Op.new(:op('mul_I'), $op[0], $op[0], QAST::WVal.new(:value($!symbols.find_in_setting('Int'))));`, but that just causes `P6opaque: get_boxed_ref could 10:34
not unbox for the representation 'P6bigint' of type Scalar` at runtime
and i haven't figured out why yet 10:35
jnthnwrthngtn Missing decont 10:38
MasterDuke oh...right...doh 10:39
hm, `nqp::decont($!symbols.find_in_setting('Int'))` still gives the same error. or you mean gen one? 10:46
jnthnwrthngtn The $op[0] 10:56
nine You need to decont $op[0] and/or $op[1]
MasterDuke doh (again) 11:12
huh. same 11:16
lizmat it just struck me that t/spec/S12-construction/destruction.t has been flapping a lot for the past 10 days or so 12:10
where the test is that a child DESTROY should be executed before a parent DESTROY
nine more like months 12:31
lizmat ah... ok... :-( 12:41
MasterDuke hm. i now have the mul_I generation working, but it obviously fails if the QAST::Var is not actually an Int. `nqp::istype($op[0].value, $!symbols.find_in_setting("Int"))` doesn't work. i'm obviously thinking even slower than usual today, what am i missing? 14:19
nine What do you expect to find in $op[0].value? 14:42
MasterDuke wants to plead the 5th 14:54
Nicholas chocolate? coffee? 14:55
(this answers can be used for either question)
these. pesky fingers
nine MasterDuke: you'll want to generate code that does the type check on that value (just like you need to do for the decont) 14:57
MasterDuke ugh. i'll probably just stick with genning the &infix:<*> then, that's a bit simpler 15:43
though i do need to add some more checks, because currently in the optimizer `2.1 == 2` is true 15:44

Powered by Cro and the Raku® Programming Language.

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

Thank you!