Hello
Here is the latest OCaml Weekly News, for the week of December 20 to 27, 2016.
Archive: https://sympa.inria.fr/sympa/arc/caml-list/2016-12/msg00077.html
whitequark announced:I'm pleased to announce that opam-cross-windows[1], opam-cross-android[2] and opam-cross-ios[3] now feature an OCaml 4.04 toolchain, and all of the packages have been brought up to date. In addition, I would like to thank Hezekiah M. Carty for porting many packages to Windows. [1]: https://github.com/whitequark/opam-cross-windows [2]: https://github.com/whitequark/opam-cross-android [3]: https://github.com/whitequark/opam-cross-ios
Archive: https://sympa.inria.fr/sympa/arc/caml-list/2016-12/msg00091.html
Hendrik Boom asked and Evgeny Roubinchtein replied:I know there are people on this list who are way more qualified to answer these questions, but let me try. > Are there any tools available that could be used to generate ocaml > bytecode for other languages? I don't think you will get a definitive answer on this list. Here is a thought experiment showing why. Suppose J. Random Hacker decides to write a compiler from WhizBangLang to OCaml byte code. Under the assumption that OCaml developers are not omniscient, the way they would learn about J. Random Hacker's efforts is if s/he either: a) announces the new language in some venue that OCaml developers watch or b) finds [what s/he believes are] bugs in the OCaml byte code interpreter and files bug reports against it. It isn't clear to me that our J Random Hacker must needs do either of those things. > If I were to do that, by hand or otherwise, how would I interpret or > compile it? The ocamlrun program, shipped with the OCaml distribution and documented at http://caml.inria.fr/pub/docs/manual-ocaml/runtime.html is the standard interpreter for the OCaml byte code. For the OCaml compiler, the byte code is a target (as opposed to a source or an intermediate representation), so the existing OCaml tool chain does not support compiling byte code, to the best of my knowledge (AFAIK projects like Bucklescript and js_of_ocaml use the OCaml front-end and intermediate representation, but supply a different compiler back end). > Would the ocaml run-time system we available -- things like the garbage > collector, I/O libraries, etc. I think that question is answered in the documentation of ocamlrun. You probably will also want to peruse http://caml.inria.fr/pub/docs/manual-ocaml/intfc.html and specifically the discussion of static and dynamic linking of C code with OCaml code. > Is anyone else working of projects like this? I am not entirely certain what the intended antecedent of "this" is here. If "this" is "a compiler that targets OCaml byte code", then please see my answer above. If you feel that the current design of "ocamlrun for standard primitives + the '-custom' flag to the OCaml compiler for non-standard primitives" is failing to address a need, then a description of the need that isn't being addressed would be a good starting point for discussion. ;-)Yotam Barnoy suggested:
Check out Stephen Dolan's Malfunction (https://github.com/stedolan/malfunction). You essentially compile down to OCaml's internal representation, and the compiler takes it from there, making either bytecode or native binaries.Gabriel Scherer also said:
There is one known case of the OCaml runtime being reused, namely the Coq bytecode interpreter, which is a modified version of the OCaml bytecode interpreter that supports a different evaluation strategy. This is described in the article A Compiled Implementation of Strong Reduction Benjamin Grégoire and Xavier Leroy, 2002 http://gallium.inria.fr/~xleroy/publi/strong-reduction.pdf This implementation duplicates (and simplifies) the instruction set and interpretation loop, but it does reuse the same value representation and, I believe, the OCaml runtime system (the GC in particular). If you are interested in a documentation of the OCaml bytecode runtime, the two following documents could be of help: - Benedikt Meurer's article on his work on jitting the bytecode runtime begins with a very clear how-level description of how the whole thing works in OCaml today: https://arxiv.org/abs/1011.1783 - Xavier Clerc, while working on OCamlJava, produced a very detailed and up-to-date reference of the OCaml bytecode instruction set ( http://cadmium.x9c.fr/distrib/caml-instructions.pdf ) (A more high-level description of the principles behind the instruction set, in particular how it makes curried functions fast enough, can be found in Xavier Leroy's course notes on "abstract machines and compilation", along with comparisons with other designs of the 80s or early 90s and some correctness proofs: http://gallium.inria.fr/~xleroy/mpri/2-4/machines.pdf )Hendrik Boom also replied:
Having started to leaf through a few pieces of documentation about the OCaml compilation chain, I now figure that the lambda IR would probably be a better place to tap into the process. So I'll re-ask, this time sking about compiling from the internal lambda representation instead of the byte-code.Gabriel Scherer then added:
Maybe you should indeed consider Malfunction as Yotam mentioned in the previous thread: https://github.com/stedolan/malfunction It is a well-defined language made for producing code to go into the OCaml compilation pipeline. You can think of it as a subset of the lambda intermediate language, with a fixed (s-expressions) syntax and parser, and a documentation of the behavior of its programs. It is a bit simplistic (or at least it was last time I looked) but if you have extra needs you could get in touch with the author (Stephen Dolan) and work with him to extend it.
If you happen to miss a CWN, you can send me a message and I'll mail it to you, or go take a look at the archive or the RSS feed of the archives.
If you also wish to receive it every week by mail, you may subscribe online.