Blend2D performance

[dumblob]

I wonder whether raqote could leverage some of the performance benefits Blend2D (C++/C) brings. Any insights?

[jrmuizel]

Is there a particular workload that you're interested in? Fixing #158 is likely needed as a first step in improving performance further. I have some local work done on that but haven't had a chance to fix it yet.

[dumblob]

Well, there are at least 2 fundamental differences AFAIK. First, Blend2D highly paralelizes everything (both SIMD and os-level threading). Second, it doesn't use "buffers" between operations but rather merges all operations into one huge loop automaton in runtime using JITting.

Some techniques could be though ported to raqote.

It's not urgent for me, but I wanted to point it out as it might further shape the vision of the raqote future.