Re: [Bug-gnubg] MAT file bug

[Philippe Michel]

On Wed, 8 Oct 2014, Michael Petch wrote:

> I have a feeling this one problem might be enough to warrant a new
> release. I was going to consider an upissue to 1.03.002 however I
> noticed the cache related changes which might hve me thinking 1.04.000.
> How confident are you that the new cache code is sound?

I'm confident it gives the same evaluations, and that it is a little 
faster on x86 CPUs at least.

> I'm teetering on these 2 options:
>
> 1) Upissue to 1.03.002 excluding the recent cache changes (include all
> other changes/fixes).
> 2) Upissue to 1.04.000 including ALL changes since 1.03.000 (and
> 1.03.001 of course)

I don't think there's anything special in the cache changes. This is only 
a small speed improvement, unnoticeable without timing a match evaluation 
or a rollout.