Prev: Fortran Ex. (Outputs 2 vectors to Matlab) with Mex file
Next: Seasonal adjustment in ML (Census X-12)
From: David Romero-Antequera on 4 Aug 2010 14:50
"Andy " <myfakeemailaddress(a)gmail.com> wrote in message <i3c9jl$2c7$1(a)fred.mathworks.com>...
> > % from
> > psi=zeros(size(y));
> > % to
> > psi=zeros(size(y),'double');
>
> % Slightly faster:
>
> m = 50000;
> n = 5000;
> tic
> a=zeros(m,n);
> timea = toc
> clear a
>
> tic
> b=zeros([m,n],'double');
> timeb = toc
> clear b
>
> tic
> c(m,n) = 0;
> timec = toc
>
> % displays:
> timea =
> 0.873374775
> timeb =
> 9.3843e-005
> timec =
> 4.9347e-005

This helped a little. Thanks!
From: David Romero-Antequera on 4 Aug 2010 14:52
Walter Roberson <roberson(a)hushmail.com> wrote in message <i3cce8$cbf$2(a)canopus.cc.umanitoba.ca>...
> David Romero-Antequera wrote:
>
> > That's something that it is not in my hands. On the other hand, the user
> > provided function, can be something that can't be optimized anymore, like
> >
> > @(t,f) abs(f).^2;
> >
> > I just want to make sure that there's nothing else that I can do.
>
> Would f be real or complex? If it would be real, then
>
> @(t,f) f.^2
>
> would be more efficient.
>
>
> Profile first, _then_ worry.

Probably that is something that I did not mention. Y is a complex valued function. Even if it is not, I do some Fourier transforms to it, so there is always some complex components around.
From: David Romero-Antequera on 4 Aug 2010 16:02
Walter Roberson <roberson(a)hushmail.com> wrote in message <i3cce8$cbf$2(a)canopus.cc.umanitoba.ca>...
> David Romero-Antequera wrote:
>
> > That's something that it is not in my hands. On the other hand, the user
> > provided function, can be something that can't be optimized anymore, like
> >
> > @(t,f) abs(f).^2;
> >
> > I just want to make sure that there's nothing else that I can do.
>
> Would f be real or complex? If it would be real, then
>
> @(t,f) f.^2
>
> would be more efficient.
>
>
> Profile first, _then_ worry.

Thanks for the suggestion (actually, double suggestion) of profile. Here are a piece of the results:

SpectralBPM>NLF 139 4.905 s 0.850 s
ifft2 140 1.979 s 1.979 s
fft2 140 1.586 s 1.586 s
@(t,f)V.*f 139 0.584 s 0.584 s

Most of the time is spent in computing the fft2 and ifft2. So I guess there is not much that I can do here :(

On the other hand, there's another part where I compute something like:

exp(1i*t*M);

where t is an scalar and M, in the worst case, is a complex matrix (relatively big). So, is there any way to speed this computing? Can I compute the value of the matrix faster based on some previous state? (because t will vary monotonically).

Thanks, again
 | 
Pages: 1
Prev: Fortran Ex. (Outputs 2 vectors to Matlab) with Mex file
Next: Seasonal adjustment in ML (Census X-12)