MPI question

2 messages 2 people Latest: Jan 10, 2013

MPI question

From: Leonid Gibiansky Date: January 08, 2013 technical
Dear All, I have a problem with MPI on the Windows 7 computer: parallel execution fails (not on all models; not always; and not even always reproducible on the same run) when Nonmem starts $COV step, or starts new $EST record in the multi-stage estimation run. Bob Bauer was not able to reproduce it on his side, but he helped to pin-point the blame on the MPI since FPI runs fine on the same runs. Has anybody else seen this problem, and if yes, were you able to fix it? Thanks! Leonid -------------------------------------- Leonid Gibiansky, Ph.D. President, QuantPharm LLC web: www.quantpharm.com e-mail: LGibiansky at quantpharm.com tel: (301) 767 5566

RE: MPI question

From: Sam Liao Date: January 10, 2013 technical
Dear Leonid: I also have the same problem with MPI 64-bit on Windows 7, using gfortran 64-bit compiler. They converged/terminated with different ofv and parameter estimates. Another problem I have when I used PsN parallel_retries, I was not able to get the same results when I used the same initial estimates from the best run from parallel retries. I believe the MPI was not involved in the parallel_retries. If anyone know how to fix this problem, please let me know. Best regards, Sam Liao PharMax Research Inc. PO Box 15661 Newport Beach, CA 92659 Phone: (201)-988-2043 Fax: (720)-294-6783
Quoted reply history
-----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Leonid Gibiansky Sent: Monday, January 07, 2013 6:43 PM To: nmusers Subject: [NMusers] MPI question Dear All, I have a problem with MPI on the Windows 7 computer: parallel execution fails (not on all models; not always; and not even always reproducible on the same run) when Nonmem starts $COV step, or starts new $EST record in the multi-stage estimation run. Bob Bauer was not able to reproduce it on his side, but he helped to pin-point the blame on the MPI since FPI runs fine on the same runs. Has anybody else seen this problem, and if yes, were you able to fix it? Thanks! Leonid -------------------------------------- Leonid Gibiansky, Ph.D. President, QuantPharm LLC web: www.quantpharm.com e-mail: LGibiansky at quantpharm.com tel: (301) 767 5566