LAPACK Archives

[Lapack] Diagonally dominant linear solvers


Here is a pointer on some recent experiments from Jack, Marc and Stan on 
nonpivoted LU on GPU.
See: Stanimire Tomov, Jack Dongarra, and Marc Baboulin. "Towards Dense 
Linear Algebra for Hybrid GPU Accelerated Manycore Systems." LAWN 210.
http://www.netlib.org/lapack/lawnspdf/lawn210.pdf
See Figure 8.

(That does not answer for Jason though :)).

--julien.


On Tue, 19 May 2009, Chuvelev, Michael wrote:

Jason,
?
did you consider an introduction of diagonally dominant functionality in 
LAPACK? That is, A=L*U (general) and A=L*D*L' (symmetric) with no
pivoting.
?
I observe a growing interest to tridiagonal solver with no pivoting, in 
particular. In case of this solver the speed-up is dramatic - up to times
- when using no pivoting. ScaLAPACK has already have it in for some reason - 
I mean p*dttrf/p*dttrs. I think it's worth while having it in LAPACK
too.
?
Best regards,
?
Michael.
?
?

 --------------------------------------------------------------------
Closed Joint Stock Company Intel A/O
Registered legal address: Krylatsky Hills Business Park, 
17 Krylatskaya Str., Bldg 4, Moscow 121614, 
Russian Federation

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.



<Prev in Thread] Current Thread [Next in Thread>


For additional information you may use the LAPACK/ScaLAPACK Forum.
Or one of the mailing lists, or