Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Choosing a DSP for porting a Matlab code to Real-time application

Status
Not open for further replies.

metalgarri

Junior Member level 3
Joined
Jan 3, 2011
Messages
25
Helped
3
Reputation
6
Reaction score
3
Trophy points
1,283
Activity points
1,469
Hi everybody!
I have a problem:
I have a Matlab code that perform signal processing (machine learning algorithms and stuff like that...) on a signal from a custom sensor in my system. I need to port this code on a DSP based platform, and I also have some real-time specification to achieve. I have already translated the code from Matlab to C.

If, for example, I have this matlab code that runs on an I7 PC and it takes "N" ms to run, how can I choose a DSP to be sure that the code will run in "2N" ms?
Is it always a matter of trying with a platform, and upgrading to a more powerful one if it does not fit the real-time requirements, or there is a formal way to "calculate" which DSP is the right one?

PS: the objective of this work is to build a custom board with a DSP and all the circuits needed for this application.

Thank you for your answer! :grin:

Carmine
 

DSP performance is usually quantified in a MIPS metric. This should give a first guess about the required DSP power.
 

So, for example if I know that the processor on which I ran the Matlab code has approx. 70 GFLOPS, to run the code twice slower I should choose a DSP with approx. 35 GFLOPS...Is it right?
 

A DSP solution won't necessarily use floating point arihmetic. The other question is if the Matlab implementation uses compiled code or just the regular Matlab interpreter.
 

It includes regular matlab code and at a certain point a C funzion is called. About the use of floating point, it has been required at the beginning, to maintain the same data format and the same results of matlab code...
 

The procedure that seems to be suitable for an estimation would actually compiling and simulating the code at the IDE for a specific core platform.
 

i had try test my image to demo of Matlab , same on type of file of demo and has same size but there are various error such as the error on calling file missing , no response occur , Matlab program close suddenly . Because I'm like to survey the nude websites then I has not finish my thesis .

So, I want to hints everybody to use C language .
 

Status
Not open for further replies.

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top