solidpixel

This is a bit of a brain dump on what a BC7 equivalent in ASTC would look like for targeting runtime compression for virtual texturing - “astcenc-rt”. This is an evening thought experiment over a beer, so beware I’ve not tried to actually build this (yet) …

After two years of incremental improvement the astcenc 3.6 release is around 20 to 25 times faster than the 1.7 release I started with. This blog is a look at some of the techniques I used, and a few that I tried but found didn’t work.

Shader texture sampling is a really interesting area of graphics technology, and one of the areas where there are some nicely hidden gotchas in terms of performance. This blog explores some of the issues …

Computer floating-point maths is the evil twin of sensible real-world maths. It provides great flexibility, being capable of high precision and high dynamic range (although not both simultaneously), but comes with some nasty boundary conditions that can cause major issues with algorithmic stability.

One of the requirements we have for astcenc is ensuring that the output of the codec is consistent across instruction sets. Users quite like having the choice of SSE4.1 or AVX2 when their machine supports it – faster compression is always good – but no dev team wants game builds that look different just because a different machine was used to build it.