On 9/8/05, Lucas Galfaso <lgalfaso@gmail.com> wrote:
A lot of people are talking about performance, how whatever is proposed should have the best performance ever and to be able to compete with the best. That is absurd, is like not having a car and your first car should be a Ferrari. Did someone _really_ benchmarked BGL? This library is not able to compete with commercial products and nobody ever asked for this. What about uBlas, this library has mayor improvement to be made. Some people asked for benchmarks for the Parameters library that I believe where never made. The for_each library defines 'extra' variables that are not always used. Does that mean that those libraries can not be part of Boost?
there is some truth in what you are saying, but i think this is a special case; most users of (unlimited precision) arithmetic are interested in speed above all else of course boost doesn't have to be better than everyone else (and it depends on the application field what is faster), but imho it has to be in the same league as existing implementations to be accepted -- things like ease of use won't help here br, andras