Ø I can say that the main part is coming up with a webpage full of ideas for projects.

If anyone wants to pursue anything like HLVM, please do get in touch with me.

I’ve been discussing student projects recently and a couple of ideas I’ve had are:

1. Merge core (stdlib) collections with the GC. For example, mutable stacks and queues are currently represented as arrays filled with NULL pointers whereas a collection-aware GC could skip the unreachable pointers and the collection wouldn’t have to NULL out its elements. Collections could also be shrunk by the GC.

2. Allow objects to morph their representation as they go from one generation to another. For example, long-lived immutable sets could morph from a tree of heap-allocated blocks into a sorted array.

I also think it would be interesting to study how fast purely functional data structures could be if their representation was optimized (within the confines of a memory-safe VM) for the performance characteristics of a naïve mark-sweep GC. For example, list cons in OCaml currently heap allocates a new block based on the assumption that the GC can do this quickly (which OCaml can) but the list implementation could pre-allocate a block of cons cells in an array and use a CAS to claim the next one, amortising allocations and greatly reducing the stress on the GC. I’ve implemented this for lists in F# where it is about as fast as the generational GC. I don’t know if the same can be done effectively for trees like Set and Map. The motivation is avoiding the pathological performance for programs that violate the generational hypothesis.

Cheers,

Jon.

From: caml-list-request@inria.fr [mailto:caml-list-request@inria.fr] On Behalf Of Yotam Barnoy
Sent: 15 January 2014 15:24
To: Adrien Nader
Cc: Ocaml Mailing List
Subject: Re: [Caml-list] Google summer of code

I actually don't think it's very hard to apply (in general). The requirements page is here (http://www.google-melange.com/gsoc/document/show/gsoc_program/google/gsoc2014/help_page?ModPagespeed=noscript#1._How_does_a_mentoring_organization). It's certainly much easier than applying for a grant.

Having participated as a mentor in another organization, and having seen the application process every year, I can say that the main part is coming up with a webpage full of ideas for projects. I think that already exists partially on the ocamllabs wiki. I'd really like to see that wiki be centralized and available from the ocaml.org site. The key thing is to figure out which projects are doable by people in one summer. The next step is that people need to volunteer for mentorship. Even 2 people are enough for a start, but 3-4 are better. Finally, one person takes on the process of filling out the application and submitting it.

One of the things that concern google, as can be seen in the webpage, is that there's enough effort from the mentors to
a. weed out the poor candidates and
b. be in touch with/demand enough from the candidates to make sure that they're actually working and not just slacking off for the summer. In my old organization, the participants had to post blog updates every week or so about their progress.

Other than that, there's not much to it.

-Yotam

On Wed, Jan 15, 2014 at 4:04 AM, Adrien Nader <adrien@notk.org> wrote:

Hi,

On Tue, Jan 14, 2014, Yotam Barnoy wrote:
> Dear List
>
> Are there plans to apply for GSOC mentorship this year? Searching online
> yielded only a rejected application from 2011. Applications for mentor
> organizations are due February 14th.

Then, if nothing has been started yet, it's probably already too late.
GSoC is fairly constraining and time-consuming. Nowadays, projects which
manage to do it have actually gotten experienced at it and this makes it
more difficult for new participants: they don't only have to do things
well, they have to do things better than projects which have been doing
it for 10 years.

There's also a feeling that Google doesn't care about what happens in
OCaml-land (that's mostly speculation).

I believe it might be worth trying again. Many things have changed and
the projects that could be worked on now might appeal more to the GSoC
organizers. That would be for 2015 in my opinion though.

Regards,
Adrien Nader