Reorganisation of the repo structure

[Kerl13]

• Reference C implementations of the different RNGs are moved inside their respective OCaml modules;
• OCaml implementations are tested against the corresponding C reference.

[Niols]

What is the status of this? Can I help? Review, maybe?

[Kerl13]

What is the status of this? Can I help? Review, maybe?

Yep, review and tell if you think of a better organisation

[Niols]

Where would you put the TestU01 bindings?

I don't know so much what to think about the organisation, because this repository features very different things, namely:

• Bindings for TestU01 (+ ProbDist + MyLib). They would probably deserve their own packaging, either as an independent repository, or at least an independent directory in this repo.

• Implementation of several PRNGs, with testing against their C reference. This whole thing looks also like a quite independent library providing PRNGs for OCaml (as well as a way to reproduce the interface of Stdlib.Random) and would deserve imo an independent packaging.

• Built on top of these two (and possibly PractRand soon-ish?), a "research" work to compare PRNGs, by benchmarking them and running test batteries on them. I think this is also an other part of the work, quite independent from the others.

So I guess I would expect these three aspects of this repository being separated in three directories at the root of this repo? testu01/ containing the definition of the whole testu01 library, prngs/ (or probably something better) containing the definition of the lib containing all the PRNGs and bench/ (or something better) containing the work about benchmarking and crushing PRNGs.

Those are simply ideas; and opinion on this?

[Niols]

I put a few comments, but their are just very random things and actually the PR can be merged whenever we want, from my point of view.

Oh, also, something that actually hasn't changed in this PR is the benchmarking tests. This is probably a job for an other PR, but shouldn't we seed the Random generators (on similar seeds) before testing? Like go from:

ocaml-rng-experiments/prng/src/bench.ml

Lines 28 to 29 in e29fb07

	B.Test.create ~name:"Random.float" (fun () ->
	Random.float 1.);

to something like:

B.Test.create_with_initialization ~name:"Random.float" (fun `init ->
    Random.init 4; (* please don't *)
    fun () -> Random.float 1.);

(untested; from memory; might not be exactly this)

[Kerl13]

Oh, also, something that actually hasn't changed in this PR is the benchmarking tests. This is probably a job for an other PR, but shouldn't we seed the Random generators (on similar seeds) before testing?

Probably yes!

But I take the opportunity to postpone this to another PR ;)

[Niols]

OK, I moved this to #10!

[Niols]

That all looks good to me! I sent quite a few comments; most of them are just here because I felt like trolling.

[Niols]

Yes!

[Niols]

Two conversations that I think can just be marked as resolved and then I have nothing else to say!