[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: benchmarks
From: |
David Bateman |
Subject: |
Re: benchmarks |
Date: |
Tue, 27 Jan 2004 11:34:28 +0100 |
User-agent: |
Mutt/1.4.1i |
> So Octave faster than Matlab for complex numbers, but slower
> than Matlab for real numbers.
> It may be that Matlab is using rfftw for real input, and
> octave always assumes complex input.
> I have a vague recollection that this issue was discussed on
> octave list eons ago, but could not find this thread in
> archive...
Looking at the sources of 2.1.53, octave does use complex transforms at
all times. In fact the reason the real transform is slower, is that as
well as the transform there is an additional copy of the matrix to a
compelx matrix, before calling the tranform.
This needs seperate plans for the real transforms, and some code
rewriting. However, octave currectly using the FFTw 2.1.x API, and if
someone is going to tackle this problem, then we should probably
update, or at least offer as an alternative the newer (but
incompatiable) FFT 3.x API. There is a bonus in this in that there are
also significant sppedup in FFTW 3.x for SSE, SSE2 and 3dNow...
It doesn't seem that hard to do this, and in fact someone tried a
while back
http://www.octave.org/mailing-lists/octave-sources/2003/63
http://www.octave.org/mailing-lists/help-octave/2003/1262
Cheers
David
--
David Bateman address@hidden
Motorola CRM +33 1 69 35 48 04 (Ph)
Parc Les Algorithmes, Commune de St Aubin +33 1 69 35 77 01 (Fax)
91193 Gif-Sur-Yvette FRANCE
The information contained in this communication has been classified as:
[x] General Business Information
[ ] Motorola Internal Use Only
[ ] Motorola Confidential Proprietary
- Re: randn benchmarks, (continued)
eigenvalues 3 times speedup patch [Was: benchmarks], David Bateman, 2004/01/23
Re: benchmarks, David Bateman, 2004/01/26