On Jun 7, 2010, at 17:39 , Paolo Giannozzi wrote:
> Of course there can be mistakes in the configure.ac file
..or in some other place. It turns out that the downloaded
packages do not contain some text files that are needed
in order to link MKL libraries (a glitch in the packaging script:
they a
On Jun 4, 2010, at 21:35 , Osman Baris Malcioglu wrote:
> If there is significant demand for the older version, the configure
> script
> can be modified, with a sacrifice in performance in the configuration
> step of some select cases.
the problem is not just performance, but having a readable
Dear Baris
Thank you for the explanation. MKL are installed and managed on such cluster
by our system administrators, so I may only ask for an update...
However, I've compiled QE 4.2 by following Ricardo's advice, and it seems to
work fine...
Yours
Giuseppe
On Friday 04 June 2010 21:35:00 Osman
Dear Giuseppe,
Please notice that you are using an older (obsolete? eh, Intel certainly thinks
so) version of MKL, the new standard INTEL has for MKL compile line is
different
(http://software.intel.com/en-us/articles/intel-mkl-link-line-advisor/)
Newer versions of the espresso configure scrip
Dear Ricardo
It seems to work... But I would ask the developers why does the configure
script not found any mkl even if they are in the path...
Thank you very much
Giuseppe
On Thursday 03 June 2010 11:48:56 Ricardo Faccio wrote:
> Hi
> I faced almost the same problem but in a single QuadCore PC,
Dear all
I'm facing some problems with mkl & QE 4.2 on a linux cluster. In the case of
4.1, the configure script found (and still find...) the installed mkl
10.0.4.023 version and the make.sys looked like
BLAS_LIBS = -L/opt/intel/mkl/10.0.4.023/lib/em64t -lmkl_em64t
LAPACK_LIBS= -L/o
Hi
I faced almost the same problem but in a single QuadCore PC, I edited the
make.sys file by hand and then it looks like as following.
And it works!
Regards
Ricardo
-
Dr. Ricardo Faccio
Prof. Adjunto de F?sica
Mail: Cryssmat-Lab., C?tedra de F?