Optimal Optimizations: A Very Preliminary Report

[Scott Robert Ladd]

Nathanael Nerode wrote:
> Is there a way to set the default optimizations differently on a 
> per-architecture basis?  If there isn't, should there be? ;-)

Yes, there should. Assuming, of course, that someone from each 
architecture has a clear idea of how various options affect their 
platform. Discovering this information is part of the motivation for my 
genetic algorithm.

My testing is ia32-based; I'm also starting test runs on my ultraSPARC; 
preliminary runs suggest the its optimization characteristics differ 
from those of the Pentium 3 and 4 systems at my disposal.

The GCCGA testing harness is nearly finished for an initial release; I'm 
encapsulating machine and version characteristics into a set of classes 
to ease porting to various architectures and languages. I'll spend the 
next couple of days getting the code in order for general release; then 
others can try it on their systems, and we can compare results.

-- 
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing