Hi folks
I'm solving linear equations like A*x = b, where A is sparse matrix,
I try to describe sparse matrix A with the std::map<int, map<int, float> > and realize the solver by LU method.
Now i'm trying to realize the solver with ublas::sparse_matrix/compressed_matrix, ublas::lu_subsitute, lu_factorize.
But I found that this version was about 3 times slower than std::map<int, map<int, float> > !
( where A is about 180*180, non zero element count is 1051, ublas version solver use time is about 0.26s, CPU :Intel PM 1.3G, Redhat Linux)
so i wonder if the ublas::lu_* use the character of sparse_matrix/compressed_matrix ?
Hoping for you advice.
Thanks and best regards!
Yan Yaqin 2004-09-03