|
14:51
MasterDuke joined
|
|||
| MasterDuke | hm. i have a change to the NQP static optimizer that changes `if @` or `if %` into `if nqp::elems`, but it dies when building NQP with `elems requires a concrete object (got a NQPMu type object instead)`, which is actually coming from the elsewhere in the static analyzer... | 14:53 | |
| github.com/Raku/nqp/blob/main/src/...er.nqp#L71 | 14:54 | ||
| so i guess the generic optimization has to be `if nqp::isconcrete && nqp::elems` | 14:55 | ||
| [Coke] | and is that still a win? | 14:59 | |
| MasterDuke | dunno, haven't done it yet | 15:00 | |
| [Coke] | heh | 15:01 | |
| MasterDuke | i feel like that's what the plain `if` probably ends up doing, so it might not be. the speed in lizmat's pr might be from not doing to extra check of concreteness | 15:02 | |
| so i'm trying to think of a way the optimizer can know if doing just the elems is safe | 15:03 | ||
|
15:08
MasterDuke left
|
|||
| lizmat | yeah, the difference in the PR is the concreteness check, which can be skipped for slurpy hashes | 15:22 | |