next up previous
Next: Auxiliary Routines Up: Parallel I/O Routines For Previous: File Management Routines

I/O Routines

  The BLAPIOS provide the following blocking and nonblocking routines for reading and writing submatrices of an out-of-core file. The nonblocking routines permit the possibility of overlapping I/O to disk with computation and interprocess communication.

P_READ.
Reads a submatrix into specified location of a matrix, and leaves the file pointer for each process at the next data element for the process. This is a blocking call.

P_WRITE.
Writes a submatrix from specified location of a matrix, and leaves the file pointer for each process at the next data element for the process. This is a blocking call.

P_IREAD.
Reads a submatrix into specified location of a matrix, and leaves the file pointer for each process at the next data element for the process. This is a nonblocking call.

P_IWRITE.
Writes a submatrix from specified location of a matrix, and leaves the file pointer for each process at the next data element for the process. This is a nonblocking call.

PIOTEST.
Tests if a nonblocking parallel I/O call has completed.

PIOWAIT.
Blocks until a nonblocking parallel I/O call has completed.



Jack Dongarra
Thu Apr 18 21:51:24 EDT 1996