Re: fftw: Usage of SSE in 64bit?

To: Charles Montgomery <cgm@physics.utoledo.edu>
Cc: debian-science@lists.debian.org
Subject: Re: fftw: Usage of SSE in 64bit?
From: "Steven G. Johnson" <stevenj.mit@gmail.com>
Date: Tue, 21 Jun 2011 20:55:42 -0400
Message-id: <[🔎] 180F1825-1B51-41F4-A643-6C277003E963@fftw.org>
In-reply-to: <[🔎] F0787835-FEC5-46CA-95D4-EA472D9C5546@physics.utoledo.edu>
References: <201103231452.27320.carsten.aulbert@aei.mpg.de> <201103231027.05275.cgm@physics.utoledo.edu> <[🔎] itp3vn$7vr$1@dough.gmane.org> <[🔎] F0787835-FEC5-46CA-95D4-EA472D9C5546@physics.utoledo.edu>


On Jun 21, 2011, at 9:22 AM, Charles Montgomery wrote:

I haven't had a chance yet to check, but that sounds like the likelyexplanation. I missed that in the documentation. It isn'tmentioned again in the section on 'FFTW Constants in Fortran', whichrecommends just using the file /usr/include/fftw3.f, which has onlyINTEGER rather than (non-standard) INTEGER*8.

The constants are indeed INTEGERs, as are the array lengths etcetera.Only the plan is INTEGER*8.

We've never heard of a modern Fortran compiler that didn't supportINTEGER*8. If you have an ancient system that only has INTEGER, thenprobably it is a 32-bit system and INTEGER is fine.

Reply to:

References:
- Re: fftw: Usage of SSE in 64bit?
  - From: "Steven G. Johnson" <stevenj@alum.mit.edu>
- Re: fftw: Usage of SSE in 64bit?
  - From: Charles Montgomery <cgm@physics.utoledo.edu>

Prev by Date: Re: Proposal: Debian Science mailing lists
Next by Date: Re: Proposal: Debian Science mailing lists
Previous by thread: Re: fftw: Usage of SSE in 64bit?
Next by thread: Re: fftw: Usage of SSE in 64bit?
Index(es):
- Date
- Thread