freenode/#shirakumo - IRC Chatlog
Search
7:58:59
Shinmera
Which would mean either the patented technique is different, or the patent is not valid due to prior publication, yeah?
8:16:49
Shinmera
For text layout I mean stuff like alignment options and an algorithm that actually selects line break options.
8:49:25
Colleen
github.com/servo/pathfinder Website (HTML), Title: GitHub - servo/pathfinder: A fast, practical GPU rasterizer for fonts and vector graphics
8:53:33
Shinmera
Though figuring out which part of that stuff to tear out to get glyph rendering is giving me a headache
9:07:31
|3b|
doesn't sound too bad, looking at https://nical.github.io/posts/a-look-at-pathfinder.html
9:07:31
Colleen
nical.github.io/posts/a-loo... Website (HTML), Title: Eight million pixels and counting – A look at pathfinder
9:09:12
Shinmera
Maybe it'll be ready by the time I've worked myself up from the gutters to the top of the stack
9:10:06
Shinmera
Yeah, thought so. I just really like the idea of something that doesn't need an atlas.
9:10:54
|3b|
ACTION is also wondering about whether it would be reasonable to generate sdf atlas on GPU at runtime
9:11:30
Shinmera
Maybe, but needing a texture large enough to hold all glyphs in a font can still be bad
9:12:43
Shinmera
Though having to recompute atlases at runtime when the set changes might introduce unwanted stutters if it's not fast enough.
9:12:46
|3b|
seems like for most normal uses, you have maybe 1-2k active at any point (and that's for something like chinese)
9:14:33
|3b|
if set of characters doesn't change rapidly, calculating a few new glyphs per frame is probably not much, and GPU bandwidth is probably high enough to make a copying GC for space in the atlas. particularly with sparse textures
9:15:55
|3b|
https://github.com/3b/first-light/blob/msdf-wip/examples/shaders/text.lisp is where i got adding it to virality
9:15:56
Colleen
github.com/3b/first-light/b... Website (HTML), Title: first-light/text.lisp at msdf-wip · 3b/first-light · GitHub