Proper README.

[lyrm]

No description provided.

[polytypic]

This looks basically good to me.

The README explains what things can go wrong with a backoff, which is a good thing to mention, and it mentions that backoff can reduce contention. It might be useful to slightly expand on how contention forms (i.e. quadratic amount of shared memory traffic caused by cache invalidations) and how backoff reduces such contention, which is by forming a kind of "queue" such that not all retry at the same time (kind of like an application of probabilistic sleep sort) so that people not already familiar with backoffs could also understand when it makes sense to use a randomized exponential backoff.

[lyrm]

For contention is this enough ?

Contention is what happens when multiple CPU cores try to access the
same locations in parallel. Let's take the example of multiple CPU
cores trying to perform a CAS on the same location at the same time.
Only one is going to success at each round of retries. By writing on a
shared location, it invalides all other CPUs' caches. So at each round
each CPU will have to read the memory location again, leading to
a potential quadratic O(n²) bus traffic.

Or should I do something more details ? (In this case, should I (and can I) directly take your paragraph in your kcas bloc post ?

(I am also working on a paragraph about the exponential part of exponential backoff :p)

[polytypic]

That paragraphs looks good to me. You could then continue by explaining that with the exponential backoff the CPUs will then not all attempt to perform the CAS at the same leading to reduction in the amount of bus traffic.

[polytypic]

Should we also include e.g. .prettierrc for automated formatting of the markdown?

[lyrm]

Yes, I am going to set it up.

[polytypic]

I added a couple of minor nits and there is the comment about explaining contention and how backoff helps with that, but this looks good, so I'll mark this as approved. Feel free to merge once you feel done. We can always fine tune things later.

Also, would be great to get a release out soon on opam, so we can then remove the backoff from Kcas and Saturn and use this library!

[lyrm]

I will finish that tomorrow morning 👍 Thanks.

[polytypic]

Speaking of using this library, I already created a draft PR to Kcas to use backoff from this library, during which I noticed that OCaml 4 is not supported.