Dear monks,

please accept my sincere apologies if you find the following question trivial.

Could anyone tell me why a self-made sparse matrix multiplication without PDL is bound to fail, as moritz said? In my opinion, anyone can use twice-nested hashs in order to represent a sparse 2d-matrix:

$matrix{$row}{$column}=$value
[download]

This doesn't store the zeroes, and multiplication is also straightforward, by looping over appropiate key lists.

I don't see why this solution should be slower than using piddles.

Thank you very much for shedding some light on this issue.