BLACS error illigal grid

[ravi]

after compilation of the example code when i try to run i get the following error....
mpirun N a.out
(here N=3 no. of processors)

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'
from {-1,-1}, pnum=0, Contxt=-1, on line -1 of file 'BLACS_GRIDINIT/BLACS_GRIDMAP'.

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'
from {-1,-1}, pnum=1, Contxt=-1, on line -1 of file 'BLACS_GRIDINIT/BLACS_GRIDMAP'.

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'
from {-1,-1}, pnum=2, Contxt=-1, on line -1 of file 'BLACS_GRIDINIT/BLACS_GRIDMAP'.

what could be the reason...please some way to come out of this......
thanks in advance......
ravi

[ravi]

for the compilation i m using the intel fortran compiler --> intel_fc_80
the command for compilation is...

mpif77 example1.f \
/usr/local/include/libscalapack.a \
/usr/local/include/blacsF77init_MPI-LINUX-0.a \
/usr/local/include/blacs_MPI-LINUX-0.a \
/usr/local/include/blacsF77init_MPI-LINUX-0.a \
/usr/local/include/blas_LINUX.a

after that i give the command for runnig as
mpirun N a.out
where N=3 no. of processors
i m having P4 3.6 GHz, kernal is 2.6.9-1.667smp....after that i m getting this error...

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'
from {-1,-1}, pnum=0, Contxt=-1, on line -1 of file 'BLACS_GRIDINIT/BLACS_GRIDMAP'.

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'
from {-1,-1}, pnum=1, Contxt=-1, on line -1 of file 'BLACS_GRIDINIT/BLACS_GRIDMAP'.

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'
from {-1,-1}, pnum=2, Contxt=-1, on line -1 of file 'BLACS_GRIDINIT/BLACS_GRIDMAP'.

what could be the reason.....please bring me out of this....
tia
ravi

[craiglucas]

Dear Ravi,

BLACS ERROR 'Illegal grid (3 x 3), #procs=3'

You have 3 processes, but you process grid is 3 X 3 = 9 processes. Thus you need to request more processes in mpirun or have a 1 X 3 or 3 X 1 process grid.

Craig Lucas

[ravi]

thanks
now after i have given the command for run like
mpirun N a.out
where N=4
the result comes out to be

ScaLAPACK Example Program #1 -- May 1, 1997

Solving Ax=b where A is a 9 by 9 matrix with a block size of 2
Running on 4 processes, where the process grid is 2 by 2

INFO code returned by PDGESV = 8

According to the normalized residual the solution is incorrect.

||A*x - b|| / ( ||x||*||A||*eps*N ) = 4.64139092E+14


what are the reasons for the solution to be incorrect....that is not empasised with the example code....as i m new with this parallelization options...please help me out of this....
tia
ravi

[ravi]

i m trying to run an exmple code of scalapack for testing..but i m getting the error .....

the configuration of Pc is 3.6 GHz, 2GB RAM fedora-3 kernal 2.6.9-1.667smp each.....

for compilation of the code as

mpif77 example.f \
/usr/local/include/libscalapack.a \
/usr/local/include/blacsF77init_MPI-LINUX-0.a \
/usr/local/include/blacs_MPI-LINUX-0.a \
/usr/local/include/blacsF77init_MPI-LINUX-0.a \
/usr/local/include/blas_LINUX.a

after this i have given the command for run like
mpirun N a.out
where N=4
the result comes out to be

ScaLAPACK Example Program #1 -- May 1, 1997

Solving Ax=b where A is a 9 by 9 matrix with a block size of 2
Running on 4 processes, where the process grid is 2 by 2

INFO code returned by PDGESV = 8

According to the normalized residual the solution is incorrect.

||A*x - b|| / ( ||x||*||A||*eps*N ) = 4.64139092E+14


what are the reasons for the solution to be incorrect....that is not empasised with the example code....as i m new with this parallelization options...please help me out of this....
tia
ravi

[Stan Tomov]

Dear Ravi,
Example 1 that you are trying to run is hard coded for a
process grid of 2 by 3, so you have to request at least 6
processes. For example

Code: Select all

mpirun -np 6 example1


gives

Code: Select all

ScaLAPACK Example Program #1 -- May 1, 1997

Solving Ax=b where A is a   9 by   9 matrix with a block size of   2
Running on   6 processes, where the process grid is   2 by   3

INFO code returned by PDGESV =   0

According to the normalized residual the solution is correct.

||A*x - b|| / ( ||x||*||A||*eps*N ) =   0.00000000E+00


Regards,
Stan