Random Ideas

[Aaron J. Seigo]

On Wednesday, August 13, 2014 22.44:16 Samuel Gaist wrote:
> I've attended a presentation about a new web platform designed to improve
> Reproducible Researches by providing a mean to create experiments
> (algorithms etc…). And If I understood correctly, one of the most used
> language for that is Python. Then I've thought: wouldn't be something where
> funq would be an interesting language to use ?

Python has become quite popular in this area because it is well supported, 
easy to use and has reasonable performance.  That has led to a number of 
python libs that are aimed at the scientific community. Hopefully funq can 
reach those same goals while adding the benefits of functional programming to 
the mix. Being able to access C++ libraries easily, a topic I haven't dug into 
enough yet, could be a big help in terms of easily adding features specific 
communities need.

> From that point, it went a bit wild: gpufunq or funqCL? Already thought
> about ? Heresy ?

I've spent a bit of time thinking about JIT-style rewriting of the generated 
bytecode. Obviously it is a bit early for that, but it influences the design 
of the VM. 

One nice by-product of the multi-process approach is that it may be possible 
to run specific processes on specialized hardware (e.g. GPU), perhaps with 
some restrictions as to the library calls they have access to.

With appropriate runtime bytecode rewriting support, it should be possible to 
target some funq processes to GPU and leave some on the CPU, allowing just the 
computationally intensive parts to run on GPU (e.g.).

This is all a ways off from where we are now, but it's good to keep such goals 
in mind ...

-- 
Aaron J. Seigo
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://kde.org/pipermail/funq-devel/attachments/20140815/dfb1da4d/attachment.sig>