Peyyety, Naga Anirudh (INT)
2018-11-23 15:11:51 UTC
?Hi all,
I have been running MEEP tutorial simulations on AWS EC2 linux server on a t2.micro instance with a one vCPU. When I tried to upscale the instance to c4.4xlarge? and run the same simulation invoking mpirun -np 16 (16 vCPUs), I see now there are 16 copies of identical outputs instead of one copy.
On the meep forum, in the past similar query was posed: https://www.mail-archive.com/meep-***@ab-initio.mit.edu/msg02887.html. But the issue has not been fully resolved there, could someone give further guidance.
Additionally, I would like to know when running bulky simulations, on say c4.4xlarge instance, is it advisable to run them on 8 real cores instead of 16 vCPUs, that is using: mpirun np -8 instead of mpirun np -16?, to avoid the issues with hyperthreading?.
Thank you very much,
Best
Anirudh
I have been running MEEP tutorial simulations on AWS EC2 linux server on a t2.micro instance with a one vCPU. When I tried to upscale the instance to c4.4xlarge? and run the same simulation invoking mpirun -np 16 (16 vCPUs), I see now there are 16 copies of identical outputs instead of one copy.
On the meep forum, in the past similar query was posed: https://www.mail-archive.com/meep-***@ab-initio.mit.edu/msg02887.html. But the issue has not been fully resolved there, could someone give further guidance.
Additionally, I would like to know when running bulky simulations, on say c4.4xlarge instance, is it advisable to run them on 8 real cores instead of 16 vCPUs, that is using: mpirun np -8 instead of mpirun np -16?, to avoid the issues with hyperthreading?.
Thank you very much,
Best
Anirudh