[mohammad]

Dear all.
I'm using a supercomputer. I use Ansys121 workbench and 16 CPUs. When I want to start my run I must choose a "Run Mode". There are 7 ,odes available on the list.

1-Serial
2-PVM Local
3-PVM parallel distributed
4-HP MPI Local Parallel
5-HP MPI Distributed Parallel
6-MPICH Local Parallel
7-MPICH Distributed Parallel

Which option is fastest for CFX121. this facility it is not already installed on the computer center. what command line should I use for LINUX

(for Window , for example the command line is like this:
Win 32x: cfx5parallel -install-mpich-service
Win 64x: cfx5parallel -install-mpich2-service)

I'm looking forward to hearing from u

[cfd_newbie]

Originally Posted by mohammad

Dear all.
I'm using a supercomputer. I use Ansys121 workbench and 16 CPUs. When I want to start my run I must choose a "Run Mode". There are 7 ,odes available on the list.

1-Serial
2-PVM Local
3-PVM parallel distributed
4-HP MPI Local Parallel
5-HP MPI Distributed Parallel
6-MPICH Local Parallel
7-MPICH Distributed Parallel

Which option is fastest for CFX121. this facility it is not already installed on the computer center. what command line should I use for LINUX

(for Window , for example the command line is like this:
Win 32x: cfx5parallel -install-mpich-service
Win 64x: cfx5parallel -install-mpich2-service)

I'm looking forward to hearing from u





So, when I want to use it's slowww 1 Iteration per hour( in serial mode its about 12 Iterations per hour.

For example you have 2 nodes - 1. Node1 and 2. Node2.
Each of these nodes will have 8 cores (most probably).
Than you should select each Node1 and Node2 eight time, in total you will have 16 cores for your computations.

In this case you should use either HP MPI Distributed Parallel or MPICH Distributed Parallel.

Doesn't CFX help says anything about this ?

Raashid

[mohammad]

Originally Posted by cfd_newbie

For example you have 2 nodes - 1. Node1 and 2. Node2.
Each of these nodes will have 8 cores (most probably).
Than you should select each Node1 and Node2 eight time, in total you will have 16 cores for your computations.

In this case you should use either HP MPI Distributed Parallel or MPICH Distributed Parallel.

Doesn't CFX help says anything about this ?

Raashid

Hi Dear Raashid...long time no see
you mean I must great PBS file evrey time including these command for nodes?

Cuz when I use my single PC with 4 cpu I just activate it once from the CFX Launcher--> Tools--->command line [command: cfx5parallel -install -mpich2 -service]

Iwnt to know how to activate it like that in the supercomputer. I'm using GUI by the way...

again; if I work with commands in PBS, do you think I will get rid of this problem to just introduce the CPUs with command?!!! do you have a sample of such PBS file??

[mohammad]

Originally Posted by mohammad

Dear all.
I'm using a supercomputer. I use Ansys121 workbench and 16 CPUs. When I want to start my run I must choose a "Run Mode". There are 7 ,odes available on the list.

1-Serial
2-PVM Local
3-PVM parallel distributed
4-HP MPI Local Parallel
5-HP MPI Distributed Parallel
6-MPICH Local Parallel
7-MPICH Distributed Parallel

Which option is fastest for CFX121. this facility it is not already installed on the computer center. what command line should I use for LINUX

(for Window , for example the command line is like this:
Win 32x: cfx5parallel -install-mpich-service
Win 64x: cfx5parallel -install-mpich2-service)

I'm looking forward to hearing from u

Is is a contradiction...???
Hey guys,
I paid attention to one thing.

the Linux on the supercomputer is Linux-64 bit....then why on the list of "Run mode" all "MPICH" options aren't "MPICH2" . As i know in windows 64 its MPICH2 and in 32 is MPICH...

What does this contradiction mean?



regards