On 5/6/2010 7:15 AM, Daniel Carrera wrote: > > I'm working on an series of benchmarks to compare PDL, NumPy, Octave > and Scilab. So far, it looks like PDL and NumPy have broadly similar > performance characteristics. Usually there is a slight edge for PDL, > but not always. In particular, PDL is slow at matrix multiplication.
I believe NumPy can use ATLAS for its linear algebra which would explain this difference. > It is difficult to design good benchmarks. My approach is to design > very simple tests which measure only a small amount of functionality > (e.g. matrix multiplication). Then, have several of those. What you see > then is that different tools are better for different tasks. None of > them is absolutely the best at everything. Benchmarks of small kernels are useful when those small kernels are the dominant part of a computation. For scripted languages with interactive development, I would expect that expressiveness and ability to get work done would be more interesting metrics. That said, if PDL is relatively slower than another package at matmult (e.g.) then that could indicate a direction for optimization. --Chris _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
