2

I am learning MPI-IO and following a tutorial (PDF download here).

For one exercise, the correct code is:

Program MPI_IOTEST
Use MPI
Implicit None
Integer :: wsize,wrank
Integer :: ierror
Integer :: fh,offset
Call MPI_Init(ierror)
Call MPI_Comm_rank(MPI_COMM_WORLD,wrank,ierror)
Call MPI_Comm_size(MPI_COMM_WORLD,wsize,ierror)
offset=4*wrank; ! because 4 bytes is one signed int
! --- open the MPI files using a collective call
Call MPI_File_Open(MPI_COMM_WORLD,'test.dat',MPI_MODE_RDWR+MPI_MODE_CREATE,MPI_INFO_NULL,fh,ierror);

Write(*,*)'rank',wrank


Call MPI_FILE_WRITE_AT(fh, offset, wrank,1,MPI_INTEGER,mpi_status_ignore,ierror);

Call MPI_File_close(fh,ierror)
Call MPI_Finalize(ierror)
End Program MPI_IOTEST

Then you just build and run it as 24 MPI tasks. Then for validation, simply do od -i test/dat You will get the result exactly the same on the tutorial, which is given below.

0000000           0           1           2           3
0000020           4           5           6           7
0000040           8           9          10          11
0000060          12          13          14          15
0000100          16          17          18          19
0000120          20          21          22          23
0000140

But if I change 1 to num:

Call MPI_FILE_WRITE_AT(fh, offset, wrank,1,MPI_INTEGER,mpi_status_ignore,ierror);

into

Call MPI_FILE_WRITE_AT(fh, offset, wrank,num,MPI_INTEGER,mpi_status_ignore,ierror);

and before that define 

integer :: num
num=1

After rm test.dat, then re-build the file and run it, you will get:

0000000           0           0           0           0
*
ArtificiallyIntelligence
  • 887
  • 1
  • 12
  • 22

1 Answers1

4

Your error is not actually in the specification or use of num but rather in the specification of offset.

If you read the man-page of MPI_File_write_at, you have to specify the offset as MPI_Offset kind.

So if you change your program to use:

integer(kind=MPI_OFFSET_KIND) :: offset

It works fine.

Did you not notice the size of the test.dat file generated?

Timothy Brown
  • 2,220
  • 18
  • 22
  • But it is writing the test.dat in non-human readable language? seems to be binary – ATK Aug 07 '19 at 09:35