DocumentCode
3374185
Title
Architectural simulation system for M.f.a.s.t
Author
Moller, C.H.L. ; Pechanek, G.G.
Author_Institution
Microelectron. Div., IBM Corp., Research Triangle Park, NC, USA
fYear
1996
fDate
8-11 Apr 1996
Firstpage
221
Lastpage
232
Abstract
The paper discusses the simulation system used to verify the architecture of the Mwave folded array signal transform (M.f.a.s.t.) processor, a single chip scalable very long instruction word (VLIW) processor array being developed by IBM Microelectronics. M.f.a.s.t. simulation, at this stage of development of the architecture, is at a very high abstract level and is intended to meet the requirements of being implementation independent and permit investigation of the architecture itself, as opposed to simply verifying the congruence of the architecture and an implementation of it. The M.f.a.s.t. simulator uses a collection of functional models, implemented in C, to represent the M.f.a.s.t. architecture. The models, each reflecting the behaviour of a component of the architecture, run as independent processes and communicate among themselves using a socket mechanism. This paper discusses the merits and mechanics of this approach, and the reasons it was adopted. Two large applications and many small test cases have been written for the M.f.a.s.t. architecture and run on the M.f.a.s.t simulator. While it is difficult, for reasons that are discussed, to make easily interpretable statements about the performance of the simulator, data from these examples indicate that the simulator will emulate the execution, under average conditions, of approximately 3000 execution-unit operations per second
Keywords
array signal processing; digital signal processing chips; instruction sets; parallel architectures; pulse transformers; reconfigurable architectures; virtual machines; M.f.a.s.t. processor; Mwave folded array signal transform processor; architectural simulation system; architecture verification; execution emulation; execution-unit operations; functional models; independent processes; simulator performance; single chip scalable very long instruction word processor array; socket mechanism; Computational modeling; Computer aided instruction; Computer architecture; Concurrent computing; Digital signal processing; Microelectronics; Parallel processing; Testing; VLIW; Video compression;
fLanguage
English
Publisher
ieee
Conference_Titel
Simulation Symposium, 1996., Proceedings of the 29th Annual
Conference_Location
New Orleans, LA
ISSN
1080-241X
Print_ISBN
0-8186-7432-6
Type
conf
DOI
10.1109/SIMSYM.1996.492170
Filename
492170
Link To Document