Re: Bi-CGSTAB.m


[ Follow Ups ] [ Post Followup ] [ Netlib Discussion Forum ] [ FAQ ]

Posted by Henk van der Vorst on May 23, 1998 at 15:36:09:

In Reply to: Bi-CGSTAB.m posted by Barya Venansius on May 04, 1998 at 12:12:16:

: In the routine bicgstab.m under mltemplates.shar,
: r_tld is constant throughout the iterations.
: I can not figure out why this is so.
: I personally expected it to change at every iteration.


MY COMMENT: The vector r_tld should be kept
constant throughout the iteration. In Bi-CG this
vector is changed, but the key trick in Bi-CGSTAB
is that the changes to r_tld are transferred to
r. This makes the method potentially faster than
Bi-CG. The vector r_tld is still required in
order to be able to computed several innerproducts.
Further information on B-CGSTAB may be obtained
from my lecture notes on my WEB page:
http://www.math.ruu.nl/people/vorst

good luck with Bi-CGSTAB,
Henk van der Vorst



Follow Ups: