KC Sivaramakrishnan

The motivation for doing this change is to avoid introducing the predefined `Continuation_already_taken` exception. Thanks to #11423 with which `Continuation_already_taken` no longer needs to be pre-defined. Hence, I do not...

Would it be possible to push the rendered page of the web manual somewhere?

Overall the documentation looks good. Consider my reorganisation comments as optional.

> I had the impression that you were grouping the most frequently used variables at the beginning of the array for cache locality purposes. We haven't measured the impact of...

Done #120.

We've massively refactored the stack layout in order to make it more efficient. Getting Arm64 back online would require rewriting `runtime/arm64.S` by replicating the changes from `runtime/amd64.S`. This requires 2-3...

- [x] runtime/amd64.S:589: `jl` (jump if signed less) should be `jb` (jump if unsigned less), since you're comparing pointers. In asmcomp/amd64/emit.mlp: - ~[ ] The `preproc_fun` function looks independent of...

Thanks @xavierleroy. I've moved them to a separate issue #754.

> Did I miss something? Is NULL now a valid value despite NO_NAKED_POINTERS? null_stk is `Val_ptr(NULL)`. So it is ok.

stdlib/fiber.mli: - [x] line 33, Continuation_alread_taken should be Continuation_already_taken - [x] line 52, the type "handler" is defined as "('a,'b) handler" but is then used at line 60 as "('b,...