TY - GEN
T1 - Performance model for a reconfigurable coprocessor
AU - Rizvi, Syed S.
AU - Hyder, Syed N.
AU - Riasat, Aasia
PY - 2008/12/1
Y1 - 2008/12/1
N2 - This paper presents an analytical model for the performance of a generic reconfigurable coprocessor (RC) system. The system is characterized by a standard processor with a portion that is reconfigurable. We describe a general performance model for the speedup of a generic RC system. We demonstrate how different parameters of speedup model can affect the performance of reconfigurable system (RS). In addition, we implement our pre-developed speedup model for a system that allows preloading of the functional blocks (FB) into the reconfigurable hardware (RH). The redevelopment of speedup model with the consideration of preloading demonstrates some interesting results that can be used to improve the performance of RH with a coprocessor. Finally, we develop a performance model for a specific application. The application is characterized by a main iterative loop in which a core operation is to be defined in a FB. Our experiments show that the minimum and maximum speedup mainly depends on the probabilities of miss and hit for the FB that resides in the RH of a coprocessor. In addition, our simulation results for application specific model demonstrate how the probability of dependency degrades the achievable speedup.
AB - This paper presents an analytical model for the performance of a generic reconfigurable coprocessor (RC) system. The system is characterized by a standard processor with a portion that is reconfigurable. We describe a general performance model for the speedup of a generic RC system. We demonstrate how different parameters of speedup model can affect the performance of reconfigurable system (RS). In addition, we implement our pre-developed speedup model for a system that allows preloading of the functional blocks (FB) into the reconfigurable hardware (RH). The redevelopment of speedup model with the consideration of preloading demonstrates some interesting results that can be used to improve the performance of RH with a coprocessor. Finally, we develop a performance model for a specific application. The application is characterized by a main iterative loop in which a core operation is to be defined in a FB. Our experiments show that the minimum and maximum speedup mainly depends on the probabilities of miss and hit for the FB that resides in the RH of a coprocessor. In addition, our simulation results for application specific model demonstrate how the probability of dependency degrades the achievable speedup.
UR - http://www.scopus.com/inward/record.url?scp=84878585159&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84878585159&partnerID=8YFLogxK
U2 - 10.1007/978-1-4020-8741-7_91
DO - 10.1007/978-1-4020-8741-7_91
M3 - Conference contribution
AN - SCOPUS:84878585159
SN - 9781402087400
T3 - Advances in Computer and Information Sciences and Engineering
SP - 515
EP - 520
BT - Advances in Computer and Information Sciences and Engineering
T2 - 2007 International Conference on Systems, Computing Sciences and Software Engineering, SCSS 2007, Part of the International Joint Conferences on Computer, Information, and Systems Sciences, and Engineering, CISSE 2007
Y2 - 3 December 2007 through 12 December 2007
ER -