Title :
Improvements to the ETS dynamic dataflow architecture
Author :
Patadia, Phenil ; Karani, Vijay ; Kavi, Krishna ; Shanmugam, Ponnarasu ; Shirazi, Behrooz ; Hurson, Ali R.
Author_Institution :
Texas Univ., Arlington, TX, USA
Abstract :
This paper discusses an enhancement to on Explicit Token Store (ETS) dataflow architecture by a variation in token format. As a hypothetical machine model, the experiment shows that the proposed variation achieved better performance by reducing the number of instruction references while maintaining same fine grained parallism of ETS architecture. The ability of the token formation unit to form two tokens for every matched token is exploited by using multiple fetch units. The results of the simulation demonstrate the ability of multiple fetch units in increasing the throughput and reducing pipeline bubbles through the ALU. Our results indicate that it is possible to implement dataflow model of execution and achieve RISC-like "one instruction per cycle" performance.<>
Keywords :
parallel processing; performance evaluation; reduced instruction set computing; ALU; Explicit Token Store dataflow architecture; RISC; dataflow model; fine grained parallism; hypothetical machine model; instruction references; multiple fetch units; performance; pipeline bubbles; token format;
Conference_Titel :
System Sciences, 1994. Proceedings of the Twenty-Seventh Hawaii International Conference on
Conference_Location :
Wailea, HI, USA
Print_ISBN :
0-8186-5090-7
DOI :
10.1109/HICSS.1994.323146