DocumentCode
3153713
Title
A uniform analysis method for DSP architectures and instruction sets with a comprehensive example
Author
Owen, Robert E. ; Martin, Daniel
Author_Institution
Data/Time Int., Saratoga, CA, USA
fYear
1998
fDate
8-10 Oct 1998
Firstpage
528
Lastpage
537
Abstract
As digital signal processing finds broader areas of application, more processors are adapting to the need for DSP operations. MMX instructions have been added to the Pentium, high-performance RISC have done similar things for workstations, and microcontrollers are doing it for embedded applications. Digital signal processors too are changing as there are increased demands for higher performance. With new processors having such vastly different architectures and employing different processing strategies, it is increasingly difficult to make meaningful DSP performance comparisons between them. This paper reviews the normal representations or views of a processor: hardware architecture, programming model, instruction set architecture and benchmarks, and their role in DSP performance estimation in four critical areas. A uniform model is proposed for the first three views, which includes a new annotated form of programming model using signal-flow-graph-like techniques. Finally, one of the new types of processors, the Siemens TriCore Microcontroller-DSP, is analyzed to test and illustrate the new models and methodology
Keywords
digital signal processing chips; instruction sets; performance evaluation; signal flow graphs; DSP architectures; Siemens TriCore Microcontroller-DSP; annotated programming model; benchmarks; digital signal processing; hardware architecture; instruction sets; performance estimation; signal flow graph; Arithmetic; Benchmark testing; Data structures; Digital signal processing; Digital signal processors; Hardware; Instruction sets; Microcontrollers; Microelectronics; Reduced instruction set computing; VLIW; Workstations;
fLanguage
English
Publisher
ieee
Conference_Titel
Signal Processing Systems, 1998. SIPS 98. 1998 IEEE Workshop on
Conference_Location
Cambridge, MA
ISSN
1520-6130
Print_ISBN
0-7803-4997-0
Type
conf
DOI
10.1109/SIPS.1998.715815
Filename
715815
Link To Document