The M-Machine is an experimental multicomputer being developed to test
architectural concepts motivated by the constraints of modern semicon
ductor technology and the demands of programming systems. The M-Machin
e computing nodes are connected with a 3-D mesh network; each node is
a multithreaded processor incorporating 9 function units, on-chip cach
e, and local memory. The multiple function units are used to exploit b
oth instruction-level and thread-level parallelism. A user accessible
message passing system yields fast communication and synchronization b
etween nodes. Rapid access to remote memory is provided transparently
to the user with a combination of hardware and software mechanisms. Th
is paper presents the architecture of the M-Machine and describes how
its mechanisms attempt to maximize both single thread performance and
overall system throughput. The architecture is complete and the MAP ch
ip, which will serve as the M-Machine processing node, is currently be
ing implemented.