Hello
After I did lam environment setup and successfully lamboot on my four
node Rocks based cluster, I run some mpi applications under mpich
example, such as cpi. Every time I run cpi, it can get the result
properly, but I saw some warning message consistently after the result:
------------------------------------------------------------------------
-----------------------
It seems that [at least] one of the processes that was started with
mpirun did not invoke MPI_init before quitting (it is possbile that more
than one process did not invoke MPI_INIT -- mpirun was only notified of
the first one, which was no node n0).
MPIrun can *only* be used with mpi programs (i.e......................)
------------------------------------------------------------------------
------------------------
P0_25895: p4_error: semget failed for setnum:0 (This p4_error does not
always happen, only some time. )
In addition, whenever I want to run linpack coming with
'/opt/hpl/gnu/bin/xhpl' on four nodes ,
HPL.dat is like the following:
-----
-----
2 Ns
2 Ps
------
------
Then I used the following command to execute it:
# mpirun n0-3 -np 4 xhpl
I got error and no result can be achieved :
------------------------------------------------------------------------
------------------------------------
HPL error from process # 0, on line 408 of function HPL_pdinfo:
>>>NEED at least 4 processes for these tests <<<
HPL error from process # 0, on line 408 of function HPL_pdinfo:
>>> Illegal input in file HPL.dat. Exiting ... <<<
P0_25895: p4_error: semget failed for setnum:0
------------------------------------------------------------------------
-------------------------------------
It seems that [at least] one of the processes that was started with
mpirun did not invoke MPI_init before quitting (it is possbile that more
than one process did not invoke MPI_INIT -- mpirun was only notified of
the first one, which was no node n0).
MPIrun can *only* be used with mpi programs (i.e......................)
------------------------------------------------------------------------
-------------------------------------
>From the command I run, it did included '-np 4'. Why did program still
asked for: >>>NEED at least 4 processes for these tests <<<
But I can run Linpack on one node successfully under LAM with one
process configuration in HPL.dat.
Also, mpirun under MPICH can run 4 nodes configuration Linpack without
any issues.
Why did these two error happen? Can anybody give any suggestions?
Thanks
|
