freenode/#sicl - IRC Chatlog
Search
19:28:45
Bike
i did a quick survey. on clasp, discriminating functions for about 28 of all 1089 defined generic functions (or 577 generic functions that have ever been called) could be improved by reduction
21:02:22
Bike
i put in something that counts how often generic function outcomes are reached. kind of interesting. for a start I can see that cleavir-ir:inputs has been called over twenty five million times
21:04:31
Bike
looking through random cleavir functions, it seems reasonably common for there to be one outcome that's almost always used, and then a few that are rarer
21:04:48
Bike
for example, cleavir-ir:inputs has twenty five million calls using that one outcome, and then a few hundred thousand using the other
21:05:07
Bike
(there are different outcomes because slots are ordered differently so they have different instance locations... wondering if it's possible to fix that...)
3:35:59
Bike
if done properly, counting the branch selection for each argument, you could optimize the dispatch a bit. check the most common stamps first and all
5:04:40
alandipert
oh, i wasn't sure if i would actually record myself either. they can't make me :-)
5:05:12
beach
Am I supposed to adjust the slide window with the mouse and count the pixels until I get exactly 1024?
5:06:38
beach
Going to ELS and meet my friends, have a glass of wine, chat with them, etc., is very enjoyable. Now the joy has been replaced by endless pain.
5:11:57
alandipert
yeah. i dunno, i'm gonna give it my best effort, but if things are blurry or whatever because i didn't get the resolution right... that will have to do. it's not like we have a month or two to become A/V experts
5:12:45
alandipert
i'm envious that you're thinking that far ahead, i actually had to completely rewrite my paper, doing that now and i owe it to them tomorrow x)
5:12:54
beach
Right, and I have no desire to become such an expert, especially since I am very busy and I would rather spend my time on more important things.
5:14:49
alandipert
yeah i shared my draft on social media, it was posted on reddit, and a commenter pointed me to a project that already does the tagbody optimization i discuss. so i let ELS know my thing wasn't scholarly and i'd either need to shorten/weaken substantially, retract, or do something else they suggest
5:14:50
beach
Well, I can write (or rewrite) a paper in a day or so. Dealing with crap like screen recording and such typically involves fighting with software that I usually don't understand, so it is going to take days of stupid activity.
5:17:31
alandipert
thanks, but i'm glad i learned of the other project in time. and demo is maybe a better format for my work anyway
5:23:06
alandipert
now i must continue with the rewriting. beach if you get hung up on any A/V stuff don't hesitate to loop me in, i have some battle scars in this area. might be able to point you at helpful tools etc if you're on linux or mac
5:43:30
beach
My idea is to do a slide show with audio, and the pointer to point to various things on the slides.
7:02:21
splittist
Telling myself to be intrigued rather than my natural response of being cynically dubious; to consider the format a new type of performance, rather than just someone reading their slides at me; value the fact that no live demos can possibly go wrong if they are pre-recorded...