LAPACK Archives

[Lapack] Question about Lapack

Good morning,

DSYGV needs A to be symmetric and B to be symmetric positive definite.

Largest size for DSYGV? On my laptop, I can do 500x500 in 0.1 second, 
1,000x1,000 in 0.5 second,
2,000x2,000 in 4 seconds, 3,000x3,000 in 15 seconds, 5,000x5,000 in a bout a 
minute.  This is an n^3
operation so for 10,000x10,000 that would take 8 minutes. (I did not try.)

LAPACK can get challenged when you start having matrix larger than 30,000 say.

If you start having this large matrices, it might be worth looking into 
software like PLASMA.  I do believe they
have release an SYGV like routine.

Cheers,
Julien.


On Jan 28, 2013, at 9:05 AM, Kevin Galiano 
<kgaliano@Domain.Removed<mailto:kgaliano@Domain.Removed>> wrote:

Good afternoon,

I am interested in using Lapack to solve the generalized eigenvalue problem: 
A*x = lambda * B * x, where a parameter N is the order of matrices A and B.

For this, I am interested in using the subroutine DSYGV.

I would like to ask, what is the largest order, N,  of matrices A and B, that 
this subroutine can handle to properly calculate eigenvalues. I appreciate your 
information.

Cordially,

Kevin Galiano
_______________________________________________
Lapack mailing list
Lapack@Domain.Removed<mailto:Lapack@Domain.Removed>
http://lists.eecs.utk.edu/mailman/listinfo/lapack

-------------- next part --------------
An HTML attachment was scrubbed...
URL: 
http://lists.eecs.utk.edu/mailman/private/lapack/attachments/20130128/54a3746c/attachment.html
 

<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