The recommendation of Intel is to use -O2 until they find out what it is. They are investigating the case, its not related to a similar bug that was reported earlier.
Ciao Gerhard ==================================== Dr. Gerhard H. Fecher Institut of Inorganic and Analytical Chemistry Johannes Gutenberg - University 55099 Mainz ________________________________________ Von: laurence.marks at gmail.com [laurence.marks at gmail.com]" im Auftrag von "Laurence Marks [L-marks at northwestern.edu] Gesendet: Montag, 14. Juni 2010 17:58 An: Gerhard Fecher Cc: pblaha at theochem.tuwien.ac.at Betreff: Re: [Wien] IFORT 11.1 and symmetry (2x missing) Thanks for the analysis. I will be interested to find out what Intel says; I have seen some strange losses of accuracy with 11.1 using -O3 and this might be a contributor. On Mon, Jun 14, 2010 at 10:47 AM, Gerhard Fecher <fecher at uni-mainz.de> wrote: > Its even worth, > as it depends where you set aext(1:5,1:5)=0.d0 > see the attached version that is just calling inva, > if the aext(1:5,1:5)=0.d0 is not directly in front of the loops, > then it crashes, at least in the Windows Version of IFORT 11.1 > > I guess using -O3 is presently rather unsafe to be used in 11.1. > > I will go to report it to Intel. > > > Ciao > Gerhard > > ==================================== > Dr. Gerhard H. Fecher > Institut of Inorganic and Analytical Chemistry > Johannes Gutenberg - University > 55099 Mainz