Boost logo

Ublas :

Subject: Re: [ublas] [uBLAS] What's inside the uBLAS matrix multiplication algorithm?
From: Gunter Winkler (guwi17_at_[hidden])
Date: 2010-06-14 15:12:45

Anders Kabell Kristensen schrieb:
> Thanks for your answer Gunter, I will try some other libraries the, but
>> It is O(n^2) for full dense matrices. It is O(nnz) for sparse matrices
>> and it is O(n*k) for band matrices. You can check matrix_expression.hpp
>> for the class matrix_:matrix_product (or similar name) and
>> operations.hpp for axpy_prod().
> don't you mean O(n^3) for dense matrices? If not, can you explain why?
> Surely, naive mat. mult. should be O(nm^2) or (n^2m), right?
you are right. I gave the numbers for matrix*vector mult. For
matrix-matrix it is n^3 and k*n^2.


__________ Hinweis von ESET NOD32 Antivirus, Signaturdatenbank-Version 5196 (20100614) __________

E-Mail wurde geprüft mit ESET NOD32 Antivirus.