17-03-2014, 10:35 PM
Abstract—In the Embedded Processor market, there are
various types of DSP Processors. Among them, 4 processors
are widely used: Texas Instrument TMS320C64xx families,
ARM processor families, analog devices
TigerSHARC®processors, and freescale SC3400 StarCore.
However, the performances of these 4 DSP processors have
never been compared together. This paper presents a
comparative performance study for these four popular types of
DSP processors. Different code strategies, such as TFRM,
TFBBGM, recursive & iterative structures, are incorporated
in FFT codes for four DSP processors. The experiments unveil
that freescale StarCore-Based DSP families have the best
performance for input size larger than 64 points.
I. INTRODUCTION
Digital signal processors have been widely used in the past
decade, especially during the last few years. A key
component in many industrial products uses a variety of
hardware approaches to implement DSP, ranging from the
use of microprocessors to field-programmable gate arrays
(FPGAs) and custom integrated circuits (ASICs). Different
from general purpose microprocessors, DSP processors often
demonstrate more advantage in terms of speed, cost and
energy efficiency.
There are four popular types of DSP processors in the
embedded processor market: TI TMS320C64xx families,
ARM processor families, analog devices TigerSHARC®-
processors, and freescale SC3400 StarCore. There has been
no comparative study on their performance so far.
DSP processor architectures have been molded by DSP
algorithms. Therefore, the best way to evaluate DSP
architectures is to use the typical DSP algorithms. In digital
signal processing, the Discrete Fourier Transform (DFT) is
used for analysis, design, and implementation of discretetime
signal-processing algorithms and systems [1][2]. FFTs
are efficient algorithms to compute DFT [3].
In this paper, we present a comparative performance study
for four popular types of DSP processors. Different code
strategies, such as TFRM, TFBBGM, recursive & iterative
structures, are incorporated in FFT codes for four DSPs. The
experiments unveil that freescale StarCore-Based DSP
families have the best performance for input larger than 64
pts.
The paper is organized as follows. Section 2 presents the
basic FFT techniques used in evaluation. In Section 3, the
DSP architectures are described and evaluated. Section 4
shows the evaluation results performed on the four DSP
architectures. Section 5 concludes the paper.
various types of DSP Processors. Among them, 4 processors
are widely used: Texas Instrument TMS320C64xx families,
ARM processor families, analog devices
TigerSHARC®processors, and freescale SC3400 StarCore.
However, the performances of these 4 DSP processors have
never been compared together. This paper presents a
comparative performance study for these four popular types of
DSP processors. Different code strategies, such as TFRM,
TFBBGM, recursive & iterative structures, are incorporated
in FFT codes for four DSP processors. The experiments unveil
that freescale StarCore-Based DSP families have the best
performance for input size larger than 64 points.
I. INTRODUCTION
Digital signal processors have been widely used in the past
decade, especially during the last few years. A key
component in many industrial products uses a variety of
hardware approaches to implement DSP, ranging from the
use of microprocessors to field-programmable gate arrays
(FPGAs) and custom integrated circuits (ASICs). Different
from general purpose microprocessors, DSP processors often
demonstrate more advantage in terms of speed, cost and
energy efficiency.
There are four popular types of DSP processors in the
embedded processor market: TI TMS320C64xx families,
ARM processor families, analog devices TigerSHARC®-
processors, and freescale SC3400 StarCore. There has been
no comparative study on their performance so far.
DSP processor architectures have been molded by DSP
algorithms. Therefore, the best way to evaluate DSP
architectures is to use the typical DSP algorithms. In digital
signal processing, the Discrete Fourier Transform (DFT) is
used for analysis, design, and implementation of discretetime
signal-processing algorithms and systems [1][2]. FFTs
are efficient algorithms to compute DFT [3].
In this paper, we present a comparative performance study
for four popular types of DSP processors. Different code
strategies, such as TFRM, TFBBGM, recursive & iterative
structures, are incorporated in FFT codes for four DSPs. The
experiments unveil that freescale StarCore-Based DSP
families have the best performance for input larger than 64
pts.
The paper is organized as follows. Section 2 presents the
basic FFT techniques used in evaluation. In Section 3, the
DSP architectures are described and evaluated. Section 4
shows the evaluation results performed on the four DSP
architectures. Section 5 concludes the paper.