Prev: How do you make an unavailabe COM port available?
Next: Returning a specified iteration from iterative solvers such as bicgstab
From: Joseph on 14 May 2010 17:08
I understand that typically "vectorizing" and removing for-loops makes Matlab code (within Matlab, I'm not talking about compiling/hand-converting to C, C++, etc.)
faster inside the Matlab application.

However, I have successfully "vectorized" some code involving "cell arrays" and it appears not to be significantly faster, it even may be somewhat slower.

Is this to be anticipated?

Is there a general paradigm for speeding up cell array operations?
Is a double-for loop the best way to do this?

Thanks,
Sincerely,
Joseph
PhD and Matlab "expert"
From: Bruno Luong on 14 May 2010 17:15
"Joseph " <josephamarks(a)yahoo.com> wrote in message <hske3k$ea8$1(a)fred.mathworks.com>...
> I understand that typically "vectorizing" and removing for-loops makes Matlab code (within Matlab, I'm not talking about compiling/hand-converting to C, C++, etc.)
> faster inside the Matlab application.
>
> However, I have successfully "vectorized" some code involving "cell arrays" and it appears not to be significantly faster, it even may be somewhat slower.
>
> Is this to be anticipated?
>

Yes, cell array (and structure) operation is always slow, no matter what you do.

Bruno
From: Jan Simon on 14 May 2010 17:19
Dear Joseph!

> I understand that typically "vectorizing" and removing for-loops makes Matlab code (within Matlab, I'm not talking about compiling/hand-converting to C, C++, etc.)
> faster inside the Matlab application.
>
> However, I have successfully "vectorized" some code involving "cell arrays" and it appears not to be significantly faster, it even may be somewhat slower.
>
> Is this to be anticipated?
>
> Is there a general paradigm for speeding up cell array operations?
> Is a double-for loop the best way to do this?

This is a very general question. The answer depends on the special case.
If you search for a string in a cell, a vectorized call to STRCMP will be much faster than two for loop. A vectorized call with CELLFUN(@strmatch, ...) might be much slower than two for loops.

Jan
From: Joseph on 14 May 2010 17:25
Just saying thanks for replying.

If anyone does know a way to speed up cell arrays, please post.

Thanks, Bruno for honesty. Sometimes in a situation like this, people try to "hide" the data ....
From: Joseph on 14 May 2010 17:32
Dear Jan!

Thanks.

In my specific example, I'm talking about simple storing/fetching data to/from a cell array.

Consider storing data in a cell array with a double-for loop, thusly:

count = 0
for j = 1: n
for i = 1:m
count = count+1;
B{count,1} = {d{j}, ei}, F(i,j) };
end
end

To be honest, I still don't quite naturally "get" the difference between {} and () in these cell operations, but I digress. From someone else's blog, I understand that curly means content ...

In any event, I vectorized the problem successfully, but it really didn't speed things up.
My question is, is this to be expected? {I think Bruno would say yes, Jan maybe would say maybe}

Thanks Again!


"Jan Simon" <matlab.THIS_YEAR(a)nMINUSsimon.de> wrote in message <hskeo9$rsd$1(a)fred.mathworks.com>...
> Dear Joseph!
>
> > I understand that typically "vectorizing" and removing for-loops makes Matlab code (within Matlab, I'm not talking about compiling/hand-converting to C, C++, etc.)
> > faster inside the Matlab application.
> >
> > However, I have successfully "vectorized" some code involving "cell arrays" and it appears not to be significantly faster, it even may be somewhat slower.
> >
> > Is this to be anticipated?
> >
> > Is there a general paradigm for speeding up cell array operations?
> > Is a double-for loop the best way to do this?
>
> This is a very general question. The answer depends on the special case.
> If you search for a string in a cell, a vectorized call to STRCMP will be much faster than two for loop. A vectorized call with CELLFUN(@strmatch, ...) might be much slower than two for loops.
>
> Jan
 |  Next  |  Last
Pages: 1 2
Prev: How do you make an unavailabe COM port available?
Next: Returning a specified iteration from iterative solvers such as bicgstab