On Thu, Feb 5, 2009 at 1:27 PM, Kevin Kofler <kevin.kofler(a)chello.at> wrote:
But several upstreams say they tried and it didn't change
anything or even
made things worse. And it'll be consuming more memory, both RAM and disk
space. I'm not convinced it is a good idea.
Can you distill that down into a set of prioritized benchmarks. If
someone were going to play around with optimization switches. What are
the minimum matrix of obtainable benchmarks that such an effort would
have to meet in order to be considered a credible test?
-jef