Path: news.mathworks.com!newsfeed-00.mathworks.com!kanaga.switch.ch!switch.ch!news-fra1.dfn.de!news.f.de.plusline.net!newsfeeder.dynfx.net!news.mixmin.net!aioe.org!not-for-mail
From: dpb <none@non.net>
Newsgroups: comp.soft-sys.matlab
Subject: Re: What is the MATLAB Internal Architecture?
Date: Mon, 08 Oct 2007 07:31:56 -0500
Organization: Aioe.org NNTP Server
Lines: 54
Message-ID: <fed85k$7qm$1@aioe.org>
References: <fe7ed3$jrt$1@fred.mathworks.com> <fe92n6$rj7$1@aioe.org> <fe9ubm$586$1@fred.mathworks.com>
NNTP-Posting-Host: BiihOdOe3hehghp28DRTpQ.user.aioe.org
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
X-Complaints-To: abuse@aioe.org
User-Agent: Thunderbird 2.0.0.6 (Windows/20070728)
Xref: news.mathworks.com comp.soft-sys.matlab:431776



SmartEngineer wrote:
> dpb <none@non.net> wrote in message <fe92n6$rj7
> $1@aioe.org>...
>> SmartEngineer wrote:
>>> Hello,
>>>       Can anyone explain the internal arcitecture of 
> the 
>>> MATLAB.How it is actually working in some detail.
> 
> Hello,
>       TO start with thanks for your attention.
> 
> 
>> You type in an expression, it parses it and then if no 
> error in syntax 
>> is found, computes a result...
>       
>    I was asking for the execution of the codeThis is 
> something anybody can tell I need not come to the MATLAB 
> forum to  get such a response.
> 
>> As Dr Lara asks, what precisely, do you propose to do 
> with this 
>> information?  :)
> 
>     Well you can tell Dr.Lara that we are here to discuss 
> and gain knowledge here and implement things better using 
> MATLAB with a better understanding of the underlying 
> concepts.

Sorry you didn't seem to relate to the technique; I was trying to elicit 
some input on what, more specifically, you had in mind that knowing 
something about the internals of Matlab would help you implement.

The question as posed is so generic there's no practical way to respond 
imo.  The overall idea of efficiency in using Matlab is to try to find 
ways to express your problem such that you can take as much advantage of 
the vectorizing capabilities as possible.  There are several articles in 
the documentation that discuss efficiency and things to avoid to 
minimize overhead.  The external API section discusses the internal data 
object of the array.

Once you get through that, the best way to approach Matlab programming 
probably is then to simply implement whatever your problem is in as a 
most straightforward way you can and only if that shows up a performance 
problem to worry about it -- the prime advantage of Matlab is in being 
able to prototype in a quick manner as compared to a lower level 
language.  If one gets bogged down in the optimization question too 
early, much time may be spent on details that don't, in the end, turn 
out to make any significant difference globally although they might make 
measurable improvements locally.  This is really only a rehash of the 
well-known mantra of don't try to optimize without first profiling.

--