Post

Radiance compiles itself in 0.2s from scratch on the experimental amd64 backend. It's fast 💨

All benchmarks are run on a single thread. Radiance compiles its own source code, while the other compilers are compiling synthetic code. When allowed to use multiple cores, Go was the fastest. Radiance isn't optimized for parallelism currently.

4/4/26, 1:00 pm0 reposts1 like

Radiant Computer

@radiant.computer

4 Apr

Some of these compilers have a substantial fixed overhead, so a larger source code was used to make things more fair. The LOC/s is what counts.

Radiant Computer

@radiant.computer

4 Apr

Also note that most of these compilers are optimized for incremental compilation. Radiance is optimized for whole program compilation. The only other language we know of that by design does not do incremental compilation is Jai.