Page 1 of 1

long double precision for complex routines

PostPosted: Mon Oct 28, 2013 7:38 pm
by chientewu
Hi,

I recently need to use ZGETRF and ZGETRS routines with long double precisions for complex numbers. I have successfully installed the LAPACK with ZGETRF and ZGETRS being only double precisions. Could anyone help me to extend these two functions to long double precisions? I use ifort compiler. Thanks a bunch!

Re: long double precision for complex routines

PostPosted: Tue Nov 26, 2013 3:51 pm
by Julien Langou
Hello,

I do not know how to do the hack you want (using "long double" instead of "double") with a compiler flag. Sorry.

This said, "long double" is not much much longer than "double", so in general you would be better off using "quad". I personally am always disappointed when I promote to "long double".

To use "quad" for ZGETRF and ZGETRS, you can either edit all the routine stack so that the code use "quad". Although this seems scary this is not that bad, I did it a few times. Or you can use a multi precision LAPACK package, for example MPACK. (I never used it.)

LAPACK does not currently support "long double" nor "quad".

Cheers,
Julien.