Re: [OMPI users] openmpi tar.gz for 1.6.1 or 1.6.2

2012-07-17 Thread Anne M. Hammond
Thanks! I ran the command: mpirun --slot-list 0-3 -np 4 --report-bindings $EXECUTABLE: and this is the output of standard error: [node50.cl.corp.com:15473] [[45030,0],0] odls:default:fork binding child [[45030,1],0] to slot_list 0-3 [node50.cl.corp.com:15473] [[45030,0],0] odls:default:fork b

Re: [OMPI users] Simple Doubt about MPI_Gather in C

2012-07-17 Thread Maginot Junior
On Tue, Jul 17, 2012 at 4:14 PM, Ghobad Zarrinchian wrote: > Dear friend > > Generally, if you have an array with x elements of a given type (e.g. > MPI_Int) you should use x as your SENDCOUNT parameter. Providing the name of > the SEND buffer just determine the start address of the buffer and you

Re: [OMPI users] Simple Doubt about MPI_Gather in C

2012-07-17 Thread Ghobad Zarrinchian
Dear friend Generally, if you have an array with x elements of a given type (e.g. MPI_Int) you should use x as your SENDCOUNT parameter. Providing the name of the SEND buffer just determine the start address of the buffer and you should expilicitly specify how many elements should be sent. Tu be s

[OMPI users] Simple Doubt about MPI_Gather in C

2012-07-17 Thread Maginot Junior
Hello, I'm starting with openmpi and still have some doubts and I think I'm making some confusion around here. When using MPI_Gather, if SENDBUF is an array of x elements SENDCOUNT should be 1 if I want to send this entire array or should I use x as SENDCOUNT parameter (telling to fetch x element

Re: [OMPI users] Bad parallel scaling using Code Saturne with openmpi

2012-07-17 Thread Dugenoux Albert
Hello. As I promised, I send you results about different simulations and parameters according to the MPI options :TEST | DESCRIPTION   SHARING        MPI | WITH PBS | ELAPSE TIME 1ST ITERATION 1      Node 2