The Community for Technology Leaders
Green Image
ISSN: 1556-6056
Yuval Nacson , , Haifa
Yoav Etsion , , Haifa
Eby G. Friedman , University of Rochester, Rochester
avinoam kolodny , Technion, Haifa
Uri Weiser , Technion Office, Haifa Tel Aviv
Switch on Event Multithreading (SoE MT) processors run multiple threads on a pipeline machine, while the pipeline switches threads on stall events. The thread switch penalty is determined by the number of stages in the pipeline that are flushed of in-flight instructions. In this paper, CFMT, a new architecture of SoE MT, is introduced. In CFMT, a multistate pipeline register (MPR) holds the microarchitectural state of multiple different threads within the execution pipeline stages, where only one thread is active at a time. The MPRs eliminate the need to flush in-flight instructions and therefore significantly improve performance. In recent years, novel memory technologies such as RRAM and STT-MRAM, have been developed. All of these technologies are nonvolatile, store data as resistance, and can be described as "memristors." Memristors are power efficient, dense, and fast as compared to standard memory technologies. Memristors therefore provide the opportunity to place the MPRs physically within the pipeline stages. A performance analysis of CFMT is compared to conventional SoE MT processors, demonstrating up to a 2X performance improvement, while the operational mechanism, due to the use of memristors, is low power and low complexity as compared to conventional SoE MT processors.
C.0.d Modeling of computer architecture, B Hardware, B.3 Memory Structures, B Hardware, B.7 Integrated Circuits, B.7.1 Types and Design Styles, B.7.1.e Memory technologies, C Computer Systems Organization, C.0 General, C.0.a Emerging technologies, C Computer Systems Organization, C.0 General
shahar kvatinsky, Yuval Nacson, Yoav Etsion, Eby G. Friedman, avinoam kolodny, Uri Weiser, "Memristor-Based Multithreading", IEEE Computer Architecture Letters, vol. , no. , pp. 0, 5555, doi:10.1109/L-CA.2013.3
96 ms
(Ver 3.3 (11022016))