The State of Multicore Software Development

I've just come back from the Multicore Expo in Santa Clara and it was striking how far multicore has progressed since we started looking into it. The number of processor companies and tools companies was a big surprise.

One of the big issues in multicore is specialism vs generalism. Some multicore processor providers design a multicore processor for a specific job (e.g. Ageia made a processor for simulating physics in computer games, Icera make a software defined modem for 3G). The advantage of making a special-purpose processor is that you can cut out all the features of the processor you don't need for the application and just concentrate on getting the maximum performance for the desired application. You can have your own software developers working on optimizing the software specifically for your application. The advantage of a processor is that you can improve the software later, and maybe even apply the processor to a different application which has similar requirements. Some companies (like our neighbours Critical Blue) take this even further and have tools that take an application and make a processor specially for it.

Sieve vs. OpenMP: Custom reduction operations

Reduction is the process of combining elements of a vector (or array) to yield a single aggregate element.  Examples of reduction include: summing or multiplying all elements of a list, and computing minima/maxima.  In fact, reduction can basically performed using any associative operation [but see Sam Lindley's comment below about monoids].

Both Sieve and OpenMP provide some support for reductions: Sieve via its accumulator classes, OpenMP via its reduction clause. However, the reduction clause of OpenMP is very limited - certain reductions operators are available over basic types (float and int), but there is no way to define your own custom reductions for specific applications.

In this post I will illustrate the respective approaches to reduction taken by Sieve and OpenMP, and show how Sieve's accumulator mechanism can be used to define a custom reduction over structures - leading to parallel reductions which are not possible using OpenMP.

Consolevania

Check out Consolevania's "Fluff" piece:

http://cv.robbo.com/torrents/fluffpiece550.torrent

It's about 500 Mb, and we feature in it!