libera/#shirakumo - IRC Chatlog
Search
14:22:57
Colleen
<shinmera> Now I just need about 15'000 more and we'll be looking at least somewhat alright.
14:26:01
Colleen
<|3b|> shinmera: i'd probably lean towards having that as an option (what is it keyed on currently?
14:27:20
Colleen
<shinmera> My proposition is this: make a KERNING function that takes two CHARACTERs and handle the lookup internally.
14:28:02
Colleen
<shinmera> An option for it would just make things even more of a mess, rather than less :)
14:29:31
Colleen
<shinmera> the kerning is a separate table indexed via a cons of the two characters.
14:29:59
Colleen
<|3b|> ACTION was just wondering if it would be worthwhile having a flag in that to indicate if a particular char is the first of any kerning pairs
14:31:01
Colleen
<shinmera> I suppose that could be worth it. Sounds like a micro optimisation though vs changing the table keys to fixnums.
14:32:01
Colleen
<shinmera> Anything to procrastinate from having to figure out once again why layouting with labels that have line breaking requires multiple resizes to stabilise to a proper sizing information for the labels :v
14:51:32
Colleen
<shinmera> |3b|: Pushed some more stuff to the PR. The writers should be translated now to use the structs, too, and I added the kerning table optimisation.
14:51:48
Colleen
<shinmera> there's still some GETFs in the code, around the page table and the sdf info.
15:58:47
Colleen
<shinmera> Of course, even getting a bloody test case together for this stupid Alloy layouting issue is a nightmare
16:00:50
Colleen
<shinmera> Hmm, ok. I think I got it. The clip-view seems to be necessary. But why?...
21:55:28
Colleen
<|3b|> possibly %kerning should just accept NIL instead of testing it in the caller as the comment suggested
21:59:29
Colleen
<|3b|> does seem to be a few ms faster when drawing 2848 or so characters in my quick (and not very good) testing
22:01:44
Colleen
<|3b|> ACTION is apparently consing about 300MB per sec in that test even without drawing the text though :p
22:05:49
Colleen
<|3b|> found a different place to push on it that possibly didn't push problems around quite as much :)
22:06:31
Colleen
<|3b|> though still have a bunch of a different sort of problem that i need to look at
22:14:15
Colleen
<|3b|> looks like might of those might be fixed by one of the other things i was planning, so i guess that's next
22:22:55
Colleen
<|3b|> ACTION reruns the full 4hr test of 2k fonts with that last fix, since i'm too lazy to think about the other one yet :p
22:25:23
Colleen
<|3b|> next part might be trivial, or i might have to go undo some other changes to make it trivial, which is less than trivial :p
22:26:38
Colleen
<|3b|> ACTION tried to make it minimize the # of contours it generated, but on further thinking decided it might be better to maximize instead... not sure if i actually switched to maximizing or not though
22:33:50
Colleen
<shinmera> |3b|: fwiw the case I had was opening the journal tab in Kandria when you have all quests. It took about 4s on my old machine to open that tab
22:34:08
Colleen
<shinmera> Still not great, and we do re-compute text probably much more often than we need to