Signal Processing on Simple/Scalar Processors vs DSP Processors – Speed and performance comparison

Keywords: DSP, ARM, Signal Processing, GPUs, Vector Processors, ILP, SISD, FFT, IFFT

DSP Processors are specialized processors with dedicated hardware used in baseband signal processing. On the other hand, Scalar processors are general purpose processing units found on most of Personal PCs. The purpose of this study is to analyze and compare speed performance of Scalar processors vs DSP Processors when faced with programs containing Rich mathematical operations.


Scalar Processors are general purpose processor known as a “single instruction stream single data stream” (SISD) CPU’s. These are meant for general purpose computing and the hardware is not optimized for specific application. Most  of the general purpose computers are SISD. These processors have very simplified hardware and contains state-of-the-art execution units like ALU, FPU etc. These cores are built around simple code execution logic i.e. Fetch instruction, decode instruction, load data and execute instruction.

Digital Signal Processors (DSP) on the other hand are spe- cialized microprocessor chip, with its architecture optimized for the operational needs of digital signal processing. DSP processors contains specialized hardware that with both data and instruction level parallelism (ILP) along with multiple execution units to reduce structure hazards. Such cores are extensively used in applications requiring rich mathematical operations in general and Digital signal processing in specific. As early mentioned, DSP processors are specialized hard- ware that outperforms Scalar processor where rich mathematical operations are involved. The purpose of this paper is to let the two hardware face signal operations involving extensive computation and compare the performance difference.


In order to compare the performance of two architecture, we will use ARM Cortex-M processor as Scalar processor and Texas Instruments C6000 family of DSP (TMS320C67xx). The Test program used will consist of three parts.

1)    Fast Fourier Transform

2)    Inverse Fast Fourier Transform

3)    Image Processing Data

The first two operations are applied to uniformly distributed signal data while later is applied to pixel manipulation of dummy images.

Time taken by each of the aforementioned operations is sampled and the average time of the operations on the two architectures is compared and presented in the subsequent section.


1. Fast Fourier Transform

A sample data of hundred of bytes are processed on ARM Cortex-M and TSM320C6713 Processor. The average time for the operations on the two architectures is shown in Fig-1.

Fig. 1. Average response time to FFT Operation

2)  Inverse Fast Fourier Transform

Similar to FFT transform operation shown in Fig-1, Inverse Fast Fourier Transform was applied to uniformly distributed data set. The average time of the operation was noted for the two architecture and is shown in Fig-2.

Fig. 2. Average response time to IFFT Operation

3)  Image Processing Data

In order to compare the performance of two architecture, two images each of 1024-pixels were processed on two architecture. The main operation performed was the addition of two images pixel by pixel and assigning the calculated pixel value to third image. The average operation time is shown in Fig-3.

Fig. 3. Image processing operation average time


The overall performance of the two architecture to mathematics enrich operation like the once aforementioned is summarized in Fig-4. Clearly (From Fig-4) the DSP processors outperform Scalar processors where signal processing is involved. It is therefore concluded that the claim that the DSP hardware is optimized for signal operations hold true.

Fig. 4. Average time of Operations


1)    Texas Instruments, TMS320C6713B Digital Signal Pro- cessor Datasheet

2)    STM32F4xx ARM Cortex-M 32b MCU+FPU 210DMIPS

3)    ARM Thumb Instruction manual

4)    ARM Cortex-M Processors Reference Manual

49 thoughts on “Signal Processing on Simple/Scalar Processors vs DSP Processors – Speed and performance comparison”

Leave a Reply

Your email address will not be published. Required fields are marked *