Re: ANNOUNCE: Magpie 0.10 Utilities for multi-core execution

["Marc A. Criley" <mcNOSPAM@mckae.com>]

On 01/28/2011 08:25 PM, Shark8 wrote:

> Brad, Marc,
>    What is the difference between Magpie and Paraffin?

Magpie provides two generic functions and a generic procedure.  They're 
designed to operate on a container of data that is indexed or keyed by 
values of a discrete type.  E.g. an integer indexed array, or a vector.

The generic functions retrieve a value from the container, apply an 
application-supplied function to it, then combine ("reduce") each result 
with a running total, until all values have been processed, and the 
final value of the running total is returned as the generic function's 
value.

The generic procedure repeatedly invokes an application-supplied 
procedure, accompanying each invocation with an index value.  There is 
no value returned, the supplied procedure is simply repeatedly invoked, 
presumably progressing through each element of the container, and it is 
responsible for managing its results.

The technique I got from Brad's article had to do with partitioning the 
range of index/keys amongst application-designated CPU cores.

One of the functions simply evenly splits the range across the number of 
cores and kicks off the execution of each task associated with a core.

The other function, and the procedure, initially split the range evenly, 
but if one of the core tasks finishes early, it advertises that it is 
available to take on more work. When one of the other tasks notices 
this, it stops, splits its remaining work, and both resume execution.

There's more detail about all this in the README, along with a couple 
examples, including a Magpie version of Jakob Sparre Andersen's 
Mandelbrot set generator.

Marc