The ABINIT Users Mailing List ( CLOSED )

[paul.boulanger@umontreal.ca]

Hi Marc,
         I cannot attach files to the Abinit forum; instead I have sent the 
files directly to you in an email message. I have begun testing the 
alternatives you have mentioned for the convergence of the forces and I am 
waiting for the results.

         For the memory leak, I would love to contribute. I have tried using 
memory profiling software available for Mac Os X, but I can't bind them to 
Abinis. The problem comes from the convention here in our group. The 
executables are shared trough the servers and I don't have the authority to 
bind the programs. Maybe you have a better idea how to monitor the memory 
between SCF steps??

          I believe the error must be the classical memory leak: a dynamical 
array that has memory allocated with the statement ALLOCATE which is not 
deallocated. Then, the memory remains in use, even though it has no 
associated variable name. This allocation error must be in a procedure used 
by the datasets in a convergence study of Ecut, where it first happened to 
me. I believe it is in each SCF cycle, but I cannot verify that supposition. 

Regards

paul