? what's wrong with this n dim fft code 
Author Message
 ? what's wrong with this n dim fft code


Quote:
>   I D/L an n dim FFT F77 code, modified from F66, and add the declaration
> on it. But the code just does not work. What is wrong?
>   The original code is fft.f, and I changed it to fft.f90 and add
> kind_var.f90

The original code had statements beginning in line 6 on most
numbered lines, so it's hardly surprising that the fixed-format
file didn't compile.  The most likely problem you have with the
modified code is that you are not invoking it correctly.  Post
your driver program that invokes it incorrectly so that we can
make a diagnosis.

Quote:
>   Also, where can I find n-dim FFT code that does NOT restricted to have

the

Quote:
> number of data points to be power of 2?

The code you posted seems to cover powers of 3 and 5, although
the code for DFT on 5 elements is a bit lacking in efficiency.


Thu, 14 Apr 2005 04:30:32 GMT  
 ? what's wrong with this n dim fft code


Quote:
> Hi:

>   I D/L an n dim FFT F77 code, modified from F66, and add the declaration

> on it. But the code just does not work. What is wrong?

Without seeing it, it is hard to say.  N dimensional FFT is just FFT
on each dimension.  That is, separation of variables can be applied.

Quote:
>   The original code is fft.f, and I changed it to fft.f90 and add
> kind_var.f90

>   Also, where can I find n-dim FFT code that does NOT restricted to have
the

> number of data points to be power of 2?

The fast part of FFT requires the length to have small factors, and
2 is the smallest.  Also, the 2 algorithm is easier to write.   If the
length doesn't have small factors, like 2, 3, or even 5, it won't be
fast.

-- glen



Thu, 14 Apr 2005 04:53:49 GMT  
 ? what's wrong with this n dim fft code

Quote:
> >   Also, where can I find n-dim FFT code that does NOT restricted to have
> > the number of data points to be power of 2?

One usually constructs n-D FFTs out of 1-D FFTs.

FFTPACK at netlib supports 2,3,4,5-primitive passes, IIRC. FFTW adds 8 and
16, I think, and will (dynamically) select the best mix for your application.

        Jan



Fri, 15 Apr 2005 18:58:19 GMT  
 
 [ 3 post ] 

 Relevant Pages 

1. geturl: what's wrong (file dim changes)

2. FFT asm code in Dr. Dobb's Journal

3. What's wrong with this code?

4. What's wrong with this code?

5. What's wrong with this code?

6. Accessing 'DIM'ed group variables

7. What's wrong with my code?

8. What's wrong with my Win32forth code?

9. What's wrong with this code?

10. What's wrong with this awk code ???

11. Compiler (people Retired) API's ;FFT's;FHT's

12. what's wrong with this code?

 

 
Powered by phpBB® Forum Software