close Warning: Can't synchronize with repository "(default)" (/var/svn/tolp does not appear to be a Subversion repository.). Look in the Trac log for more information.

Opened 17 years ago

Closed 17 years ago

#624 closed task (fixed)

Configurable drop value for design matrix of BSR linear blocks

Reported by: Víctor de Buen Remiro Owned by: Víctor de Buen Remiro
Priority: normal Milestone: Numerical methods
Component: Math Version: 2.0.1
Severity: blocker Keywords: BSR, CHOLMOD
Cc:

Description

Sometimes design matrix of BSR linear blocks (main or missing ones) has a lot of cells very close to cero.

If we drop it all, matrix sparsity will be bigger and Choleski factorization and triangular system solving will speed up.

User must be able to control this action by mean of a configuration parameter.

Change History (3)

comment:1 Changed 17 years ago by Víctor de Buen Remiro

Status: new → assigned

comment:2 Changed 17 years ago by Víctor de Buen Remiro

Milestone: → BSR Numerical methods
Version: → 2.0.1

comment:3 Changed 17 years ago by Víctor de Buen Remiro

Resolution: → fixed
Status: assigned → closed

There is a new configuration parameter but it doesn't seem to be much faster when there is ARIMA block.

Drops linear block matrix to force maximum sparsity
Real bsr.linBlk.X.drop = 1.E-16;

Note: See TracTickets for help on using tickets.