|
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. |
|||
|
01:35
lizmat_ joined
01:38
lizmat left
02:14
MasterDuke joined
|
|||
| MasterDuke | so the two bytecode dumps from t/02-rakudo/reproducible-builds.t have the same number of lines. and the difference doesn't start showing up until the lexicals. but *many* of them are in different frame numbers | 02:23 | |
| hm. probably unrelated, but something i noticed while investigating stuff. during the build of CORE.c, we do a lot of allocating via MVM_free_at_safepoint. and then in MVM_alloc_safepoint (which frees all the entries) the max size of the list was 22k | 03:26 | ||
| i wonder if it would make sense to statically allocate a giant array of pointers and just assign and increment an index instead of adding to a linked list | 03:29 | ||
|
07:39
lizmat_ left,
lizmat joined
08:16
sena_kun joined
|
|||
| jnthn | Booleans tend to be quite transient anyway (produced and soon tested, rather than storing loads of them in arrays). An MVMRegister is a union of different types that are maximally 64 bits, and CPU registers are also 64 bits. | 09:00 | |
|
13:05
Woodi left
13:09
rypervenche left
13:19
Woodi joined
13:23
rba left,
rba joined
|
|||
| ab5tract | jnthn++ thanks for the explanation, it makes a lot of sense to me | 14:01 | |
| japhb | There *are* uses for byte-boolean arrays that are situationally more efficient than bit-packed or fully-unpacked bools, but I think we just haven't hit them yet. | 14:34 | |
|
20:11
rypervenche joined
22:15
sena_kun left
|
|||