[toast0]

The BEAM JIT is all done when .beam files are loaded. It's a runtime JIT, because you can always load new code, but it's not a tracing JIT and it's designed to be pretty fast to do it while loading. For most applications of BEAM, I think the time spent loading code, including JIT, is probably immaterial.

If you have a special need, I haven't seen anyone do it, but I would expect it wouldn't be that hard to run the jit ahead of time. I haven't fully studied it, but I think the expectation is the output from jit should be pretty much the same every run, so it seems doable to generate it once and load many times.