[molpro-user] Shared memory question in CCSD(T) calculations

[Sve N]

Dear Jani,
in principle it is possible to compile molpro with threaded mkl-libraries. They should perform all blas, and I think lapack routines in parallel, and only use the given amount of memory regardless of cpu-count. But I have tried this and seemingly molpro is not, or not much accelerated by this, although more than one cores are used (as shown by top). For molcas this works very well, so I wanted to try to find out, why molpro is seemingly not affected much, but I haven't had much time, the last weeks. Maybe some developer can explain further, or you give it a try...
Sven
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20130625/e290c4df/attachment.html>