Re: [AMBER-Developers] nofftw3 and MKL

From: Daniel Roe <>
Date: Mon, 13 Oct 2014 20:42:18 -0600


I think that overall it's better to have the '-nofftw3' option rather
than not having it. Personally, I favor keeping it just for
convenience. On several systems where I maintain Amber distributions
for our lab we don't (yet) need any fftw3-related stuff, and not
having to configure fftw3 does speed up the build time quite a bit (I
love '-nomtkpp' for similar reasons). Also there is always the
possibility of wanting to build Amber with compilers that for some
reason choke on fftw3 (due to e.g. a compiler bug that only hits the
fftw3 build), and having the whole Amber build process fail just
because of that seems bad.

Actually, what I'd really be in favor of is having the fftw3 (actually
any 'external' libraries) configure/build process be like NetCDF where
it is only configured and built when needed (first time, whenever
compiler changes etc). Also, being able to specify an already-compiled
fftw3 would be great.

I've also found the '-nosse' option crucial when building with intel
compilers when running on systems that are not very homogeneous (like
here at CHPC). The default optimizations for intel compilers that we
use are so aggressive that the resulting binaries are only able to run
on some of the machines, and on others it fails with a 'instruction
not supported' type error. With '-nosse' this is not a problem.


On Mon, Oct 13, 2014 at 6:41 PM, David A Case <> wrote:
> On Mon, Oct 13, 2014, T Luchko (Lists) wrote:
>> -nofftw3 is there because fftw3 requires a minimum of GNU 4.3 and Intel
>> 9. It used to be that the minimum compiler version we supported was GNU
>> 4.1.2. I’d be happy to get rid of this restriction.
> I think we should assume compilers able to handle fftw3. I also note that
> pmemd uses -DPUBFFT even though fftw3 has been configured and is available,
> say with gnu compilers chosen. Does anyone know why this is?
>> Inheritance is the big one for me. It would simplify my code a lot.
> Are you think of the "extends" keyword, or something else here? Goggling
> on "inheritance Fortran" yields a mishmash of hits, with not an obvious single
> thing being "inheritance".
> ....dac
> _______________________________________________
> AMBER-Developers mailing list

Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 307
Salt Lake City, UT 84112-5820
(801) 587-9652
(801) 585-6208 (Fax)
AMBER-Developers mailing list
Received on Mon Oct 13 2014 - 20:00:02 PDT
Custom Search