bartoszmodelski

Thank you for highlighting this and contributing a test. Indeed, both of the writes should never succeed. I had a stab at fixing this and the diagnosis here is correct...

[kcas](https://github.com/ocaml-multicore/kcas) is now based on the paper above (Practical Multi-Word Compare-and-Swap Operation), so I believe this is completed

To add to that, at the moment offers are not removed on cancellation - bonus points for out of order removal.

I spent a while looking at this, I think it's a rather non-trivial change (compared with e.g. ``). Code is in paired-composition. # The issue Turon's works (e.g. [Reagents](https://aturon.github.io/academic/reagents.pdf)) introduce...

Addressed in https://github.com/ocaml-multicore/reagents/pull/33

Closing as kcas and reagents build on 5.0.0 (although reagents are not in opam yet). Feel free to reopen if there's any issues.

stale since 2017, closing do let me know if you want to work on any of it

> This isn't quite right: the threshold for large allocations is 129 _words_, which is 8256 bytes on 64-bit systems. Rip me. Thanks for pointing out. > In fact, I...

I’m happy to shoot a PR following @stedolan’s suggestion in a week or two (on holidays now). On Fri, 24 Mar 2023 at 19:49, Florian Angeletti ***@***.***> wrote: > cc...