riggsdk
Joined 6 years ago
Joined 6 years ago
I just run a low-res (256x256) pre-pass and store the distances to a floating point texture. I then use that pre-pass texture as the starting point (minus some delta) when drawing full-screen. This makes it really nice and performant, even for complex SDF shapes.
I think a common misconception in GPU programming is that branching is slow. It is only really slow when neighboring fragments diverges on those logic branches.
The quick pre-pass step gets close enough to the SDF surface that more fragments are in lockstep with each other and terminates at the same time, eliminating the expensive re-runs the GPU driver has to do. More experimentation is needed on my end. I do this in the browser with WebGL so accurate profiling is sometimes difficult.
I experimented with different resolutions and number of pre-pass steps but found it was sufficient on most GPU's with a single prepass run (subject to change the more I test).
Either as a date in the example "4 days ago" or "in 2 days, 2 hours and 28 seconds" for future events. This requires some control for granularity to control for how precise you want it to be and what to omit.
"a few seconds ago", "3 seconds ago", "less than a minute ago".
Should support a shortform that can act as a countdown timer "00:00:56" or "00:56".
I would love to quickly install a small subset of the planet (say my home city or country) on a cheap webhost with a simple LAMP stack. Then have the map just use that server whenever it fits within the bounds of my subsets and then possibly fallback to other or the public servers if requesting data from outside.
Routing also seems like something that could technically be done (albeit less efficient than dedicated C++ implementations) on cheap servers but I guess it would require more work.
Rolling out map updates for all these self-hosted instances is a bit of another problem but I'm sure it can be done.
That’s just… I don’t know what to feel about that. I’d rather keep the websites we visit for humans first, LLMs second. Not the other way around.
> It's a single function that's now 100x faster, not the whole of FFmpeg
Seems to be in a color detection filter.
> avfilter/vf_colordetect: add x86 SIMD implementation
Just as I’m about to tap it, the other person ends the call and what I’m actually tapping is some other person on my call list that it then immediately calls. Even if I end the call quickly they often call back confused “You called, what did you want?”
Apple: PLEASE add a delay to touch input after the call screen closes.
You'd translate the "else" keyword to "ellers". "if then else" => "hvis så ellers"
Related, the word "or" would be translated to "eller" (without the 's') "and or not" => "og eller ikke"