DESIGN OF AN EFFICIENT DATA-DRIVEN PIPELINED COMPUTER ARCHITECTURE

The combination of dataflow and von Neumann execution models is a rece nt trend in the design of high performance computers. In this paper, a data-driven hybrid computer architecture is presented. Instead of a p rogram counter, the principle of dynamic data-driven execution is used to control the execution of instructions in a von Neumann-style pipel ine architecture. One of the operands of each dyadic instruction is ex plictly stored in memory, and the memory locations of the stored opera nds are used as tags. Matching is accomplished by simply checking a pr esent flag and no special matching unit is required. No bubbles will o ccur in the pipe if sufficient parallelism exists in the program. Expe rimental results show that the proposed architecture outperforms two p reviously proposed architectures.