Re: Running on parallel hardware, was: Fun with LevelDB, LLVM and OpenCL

[Jean-Sebastien Delfino <js...@apache.org>]

On Sun, Aug 14, 2011 at 9:30 PM, dsh <da...@googlemail.com> wrote:
> Hi Jean-Sebastien,
>
> the old SpiderMonkey port [10] already contains support for nspr - so
> that end should be covered cause I am using the old port descriptor
> files as a basis for the new 1.8.5 port. That's the type of embedded
> board [11] I am considering. It comes with an FPGA processor that
> accompanies the main CPU that I want to leverage to optimize certain
> operations (maybe that scenario is similar to what you are doing with
> offloading CPU cycles to the GPU).
>
> [11] http://soekris.com/net6501.htm
> [10] http://www.freebsd.org/cgi/ports.cgi?query=^spidermonkey&stype=all&sektion=all
>
> Cheers
> Daniel
>

Yes it's similar. I'm targeting two different platforms:

- a cluster of energy efficient Atom + nvidia Ion fanless mini-itx
boards + flash memory (pretty cheap computing power)

- EC2 spot GPU instances loaded with high end nvidia Tesla Fermi gpus
(not cheap but great for spot stress testing)

--
Jean-Sebastien