Very poor sparse matrix assignment performance
Reported by Roland
BUG DESCRIPTION:
----------------
The sparse matrix assignment seems to suffer from very poor performance when using the submatrix assignment syntax M( vec_row, vec_col ) = x where vec_row and vec_col are vector of indices.
ERROR LOG:
----------
None, very very poor performance.
HOW TO REPRODUCE THE BUG:
-------------------------
M = spzeros( 2000, 2000 );
M( :, : ) = 0; // takes more than 30 sec on my machine whereas it is immediate with scilab 5
M = spzeros( 10000, 10000 );
M( :, : ) = 0; // Operation aborted takes more than 3 hours...
OTHER INFORMATION:
------------------
1) The previous pieces of code are executed immediately on Scilab 5.5.2.
2) The memory use is increased when executing the previous command like Scilab tries to compute the full matrix before performing the assignment.