Gabriel Scherer

This being said: we have been in discussion for years about replacing the current short-paths implementation with a better one that is in Merlin ( https://github.com/ocaml/ocaml/issues/6600#issuecomment-473035999 ). Given the lack...

> it also opens the door for performing optimisations that would be important in effect handler-oriented programs I don't think this is actually the case. The current API delegates the...

To clarify: I am skeptical that "separating effects from operations" has a strong impact on the optimization opportunities.

My gut feeling: this approach is also reasonable, but the use of a one-character buffer is a bit of a hack. (We could be a bit more explicit about the...

I rebased on trunk and added a Changes entry in the 5.2 section.

@shindere I am out of ideas for who would want to review this PR, and I would rather have someone else than myself do it. Would you by chance be...

@shindere : ping.

@Julow I wonder if you would be interested in helping review this PR. It looks like I will do it myself otherwise, but maybe we could split some of the...

Now that we don't support native compilation with x86 32bit anymore, most architectures have reasonable number of registers and frame pointers might affordable. However, it is not clear that the...

Oof, this is a substantial cost. I thought that only code with deep recursion would be affected, but in fact this is not the case, for example [qr_decomposition](https://github.com/ocaml-bench/sandmark/blob/5b97f1f2977b95feb5fc88a5b3dd5cb9c3d90989/benchmarks/numerical-analysis/qr_decomposition.ml#L4) and [LU_decomposition](https://github.com/ocaml-bench/sandmark/blob/5b97f1f2977b95feb5fc88a5b3dd5cb9c3d90989/benchmarks/multicore-numerical/LU_decomposition.ml)...