[quarkz]

Hi,

I am now trying to read a data file which has about 2000x2000 pts. In the past, when file size is smaller, the file was read by 1 processor and then distributed to the rest of the processors.

However, now the file size has gotten too big for the current memory. I am able to write to the file in parallel using:

...
DO irank = 0, nprocs - 1
CALL MPI_BARRIER(MPI_COMM_WORLD, ierr)
IF (irank == myrank) THEN
IF (myrank == 0) THEN
OPEN(10, FILE=’output’)
ELSE
OPEN(10, FILE=’output’, POSITION=’APPEND’)
ENDIF
WRITE(10) (outdata(i), i=ista,iend)
CLOSE(10)
ENDIF
ENDDO
...

I tried to do the same for the reading case but it can't work. Is it possible to read in fortran files by skipping e.g. 20000 values? If it's possible, then I should be able to solve my problem. Btw, I am using fortran binary files.

Thanks