Title :
A Configurable Processor Synthesis System
Author :
Gay, W. ; Gloster, C.
Author_Institution :
Howard Univ., USA
Abstract :
In this paper we present a configurable digital signal processor synthesis (CPS) System that produces a library of high-performance processors wherein each processor executes a specific digital signal processing (DSP) algorithm. Each processor contains a small instruction set and implements a particular application. These algorithm specific DSPs (ASDSPs) are used to alleviate bottlenecks in software by replacing computationally intense portions of a high level DSP algorithm with custom hardware. Each ASDSP generated by the CPS system is individually loaded into a commercially available configurable system. A library of algorithm specific DSPs for a sample application set of fundamental complex arithmetic modules is generated using the CPS system. The resulting library of floating point processors is implemented on a Firebird FPGA board containing a Virtex XCV2000E FPGA part. The results are compared with a comparable software algorithm implemented on a 2.79 GHz Pentium IV general purpose processor. This comparison shows that, despite having a clock speed that is an order of magnitude slower than the microprocessor (54 MHZ versus 2.79 GHZ), the proposed ASDSP ran an order of magnitude (13X) faster than the microprocessor implementation in the best case.
Keywords :
digital arithmetic; field programmable gate arrays; microprocessor chips; signal processing; Firebird FPGA board; Virtex XCV2000E FPGA; complex arithmetic modules; configurable digital signal processor synthesis; configurable processor synthesis system; digital signal processing algorithm; floating point processors; high-performance processors; Application software; Digital signal processing; Digital signal processors; Field programmable gate arrays; Microprocessors; Signal processing algorithms; Signal synthesis; Software algorithms; Software libraries; Variable speed drives;
Conference_Titel :
Field-Programmable Custom Computing Machines, 2007. FCCM 2007. 15th Annual IEEE Symposium on
Conference_Location :
Napa, CA
Print_ISBN :
978-0-7695-2940-0
DOI :
10.1109/FCCM.2007.43