Search
Monday, 1st of June 2020, 13:06:37 UTC
13:12:23
heisig
beach: Yes, I read about your work on EQL specializers. But I didn't think it through yet.
13:12:45
heisig
As for relaxing - long walks or riding the bike work for me :)
13:13:12
beach
Yeah, I'll try that. Thanks.
13:14:19
beach
So if TYPEP contains EQL specializers, we don't currently cache calls, which means we call compute-applicable-methods, etc, in each call. These functions use things like copy-list, nth, etc.
13:14:35
beach
Those latter functions start by checking their arguments, using ... TYPEP.
13:15:54
heisig
Yes, that is definitely a problem.
13:16:12
beach
Besides we had to figure out how to cache calls to direct instances of standard-generic-functions with EQL specializers anyway.
13:16:24
beach
... for reasons of performance.
13:16:43
beach
So, might as well do it now.
13:17:08
heisig
I think you already wrote that bypassing COMPUTE-APPLICABLE-METHODS for standard generic functions is possible.
13:17:22
heisig
That could be one way around the issue.
13:18:23
heisig
The other one could be to follow the generalizers/specializers proposal. But I am not sure that is applicable here.
13:23:38
beach
Yes, my suggestion relies on the fact that we can bypass COMPUTE-APPLICABLE-METHODS for direct instances of STANDARD-GENERIC-FUNCTION.
13:23:51
beach
I don't remember that proposal.
13:26:36
heisig
This one: https://research.gold.ac.uk/9924/
13:38:51
no-defun-allowed
Hello (and goodnight) heisig!
13:40:21
heisig
Goodnight no-defun-allowed!
Tuesday, 2nd of June 2020, 1:06:37 UTC