How about a competition?

[nickoppen]

Hello,

With all the buzz about the Parallella at the moment I'm sure that there is going to be a huge amount of effort focused on getting projects going. There'll also be a big learning curve and getting over the learning curve is always a combination of enthusiasm and the availability of leaning materials.

If Adapteva ran a competition over the next 6 months or so this could help with both of those. My thought is that the entry criteria would be for a significant computational problem with publicly available source code (running on the Parallella of course). The judging criteria should be for the system that best utilises the Parallella's capabilities in a meaningful way.

In the longer term this would leave a legacy of examples for people to download, pull apart and learn from.

The prize would have to be something meaningful (a Parallella-64 springs to mind) but mainly it would be about winning.

How about it?

nick

[LamsonNguyen]

Nice idea! I think there should also be prizes for optimal code and writing the best "Intro to Parallella" tutorial; the E16 comes to mind.

[9600]

Agreed, definitely a good idea! And a first draft of an outline for a competition for schools and universities was to the Parallella website around a year ago.

Once the boards have gone out to backers I think we'll almost certainly look at organising something.

As they say, "the devil is in the detail". In order to make a competition accessible to as wide an audience as possible, I have wondered about the possibility of having a Parallella cluster online, to which entrants can schedule code for execution — as you would with a typical shared HPC facility — and get the results back. With certain limits and checks, of course. I'd be interested as to your thoughts on the idea. Practical? Useful? Might it stimulate wider interest?

Cheers,

Andrew

[nickoppen]

Hi Andrew,

I was thinking that for the broadest coverage of solutions and techniques there would not be a single set task. If that is the preferred model then maybe a couple of competitions covering different tasks. Then the library would contain a bunch of examples all addressing the same problem which would be interesting from the educational point of view.

Also agree that there is a need for well written documentation. Perhaps an outline of the algorithm and implementation would be part of the entry.

The more I think about it the more I agree that the devil is indeed in the details. I guess it boils down to what outcomes Adapteva wants to achieve.

nick

[Gravis]

i think people having a parallella to develop on is a bit of a prerequisite from the competition. fill the orders that are on the books and then have the competition otherwise you are excluding a large amount of possible participants.

[LamsonNguyen]

[CIB]

Yeah, if there's a way to make it somewhat safe, I'll be happy to offer access to my board when it arrives. I don't plan to use it for anything than testing my own programs anyway, so there'll be plenty of time when it won't be in use.