Re: [AMBER-Developers] nofftw3 and MKL

From: T Luchko (Lists) <"T>
Date: Mon, 13 Oct 2014 14:51:40 -0700

On Oct 13, 2014, at 1:05 PM, David A Case <case.biomaps.rutgers.edu> wrote:

> On Mon, Oct 13, 2014, Niel Henriksen wrote:
>>
>> I noticed that in the current GIT version of Amber, compiling fails if you
>> use MKL libraries and the "-nofftw3" option together. Perhaps MKL is only
>> used for fftw and thus specifying these two options is silly, but in case
>> that's not true, I thought I'd share.
>
> For Niel: what is the error?
>
> For everyone, but especially Tyler: do we need the -nofftw3 option anymore?
> How about flags like -norism, -nosse, -g95, -phenix, ....
>

-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 would suggest a minimum of GNU 4.4, which is what shipped with RHEL 6 and CentOS 6 (RHEL 7 is out now). This gives access to some new Fortran features, such as inheritance. On the Intel side, feature support for Fortran has been better. I think we could stick with 9.0 as the minimum (released in 2005). I would suggest version 11 or 12. Version 12 is from 2010 and adds a lot of Fortran 2003 and 2008 support. Other vendors, like IBM and PGI, have had good Fortran 2003 and 2008 support for a long time. g95 has poor feature support, even now.

Inheritance is the big one for me. It would simplify my code a lot.

-norism isn’t really necessary. RISM is compiled in serial automatically and I don’t think there is a need to turn it off. Parallel compilation is a problem because linking C and Fortran depends on the MPI library being used. Therefore, it is turned off by default and the user needs to specify -rism if they want it.

Tyler
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Oct 13 2014 - 15:00:02 PDT
Custom Search