This thesis describes a high-performance implementation of the Message
Passing Interface (MPI) library based on a shared memory communication
mechanism.
Our implementation, called MPI/MBCF, combines two protocols to utilize
the shared memory communication mechanism: the write protocol and
the eager protocol.
In the write protocol, Remote Write is used for communication
with no buffering.
In the eager protocol, Memory-Based FIFO is used for buffering
by the library.
These two protocols are switched autonomously according to the precedence
of send and receive functions.

The performance of the MPI/MBCF was evaluated on a cluster of workstations.
We measured the round-trip time and the peak bandwidth, and executed
the NAS Parallel Benchmarks.
The results show that a message passing library achieves high performance
by using a shared memory communication mechanism.