GPU-accelerated computation in ATS

[Vanessa McHale]

Hi all,

I have an example of calling Futhark from ATS:
https://github.com/vmchale/ats-stats - it depends on OpenCL and it runs
relevant computations on the GPU.

So far it's pretty spartan (dot products, matrix math and a few
descriptive statistics), but I hope to expose a wider API once the
upstream has statistical tests and such.

Cheers,
Vanessa McHale

[gmhwxi]

Very nice!

Those dimension indexes can be very useful in practice to catch bugs.

[Brandon Barker]

I'm curious, how does Futhark handle modeling IO? I mean, at a certain level it is all about IO I assume, but it claims to be pure. And it is an ML-based language as I recall, and I think ATS is the only ML language I'm familiar with that can model IO effects (as linear types, iirc).

On Tuesday, February 25, 2020 at 8:11:41 AM UTC-5, Vanessa McHale wrote:

[Vanessa McHale]

It doesn't! As far as I know :)

It's "pure" in that it's for array computations, then they're run on a GPU (or CPU or whatever backend). I can't think of many languages that model IO with linear types; I think part of the "ATS advantage" is how it handles memory.