Abstract The mechanism of multiprocessing (MMP) has been developed and implemented as an enhancement of a standard operating system (OS ES) to support efficient execution of fine-grained parallel activities. The MMP mechanism is presented through the description of its primitives, associated data structures and its interface with the OS ES. Further efficiency enhancement of the MMP mechanism has been achieved by the hardware implementation of the MMP primitives in the specially designed processor, named the MMP processor. The details of its architecture and organisation are given in the paper. Special emphasis is put on the design of the efficient pipelined control, which resulted from the precise timing analysis of the considered design choices.
[1]
Glenford J. Myers.
Advances in computer architecture
,
1978
.
[2]
Milo Tomasevic,et al.
An operating system accelerator
,
1998,
J. Syst. Archit..
[3]
Milo Tomasevic,et al.
The Interactive Development and Testing System for a RISC-Style Processor
,
1999,
Comput. J..
[4]
Philip A. Bernstein,et al.
Principles of Transaction Processing
,
1996
.
[5]
Veljko Milutinovic,et al.
Microprocessor architecture and design for GaAs technology
,
1988
.
[6]
Devang Shah,et al.
Implementing Lightweight Threads
,
1992,
USENIX Summer.