Re: Regression test

[Simon Waters]

Tim Bedding wrote:
> 
> Do we have any kind of regression test for
> GNU chess?

I replied to Tim personally.

For the list I keep some tests in the CVS repository chess/test
see chess/test/README.

We won't reject changes that lower performance on regression
testing for as long as they take the project where it is
heading.

Sometimes you get changes that produce different results on
different computers, changes that affect memory usage in
particular, where different CPU cache sizes produce different
results, or the compiler can perform architecture specific
optimisations, so it isn't always apparent what effect a change
will have.

Similarly I wouldn't particularly want to incorporate changes
that provide assembler level bit population counts, which purely
make the code run faster, but otherwise make life harder, and
the benefits aren't portable. Crafty does these kind of things.
Algorithmic improvements will exists when 8086 architectures
really are a thing of the past ;-)