If you look at how the pmemd config.h files are written, there is actually
the capability to set optimization level to something less aggressive - this
is in the files under pmemd/config_data. I think currently the two biggest
configuration problems for pmemd are mpi and mkl. For mpi, I think an
argument could be made for moving to mpif77 (we really don't need mpif90) in
place of specifying the compiler - I originally was worried about trusting
any old mpif77 script, but in general this has not been nearly as big a
problem as unpredictability of mpi linkage requirements for a given
installation. For mkl, I think we are stuck having to pray that intel stops
changing things every minor release, and maybe even simplifies things; it
may be too much to hope for. With pmemd, unless you are using generalized
Born, MKL really does not help, so I have taken to mostly not using it
myself.
Regards - Bob
----- Original Message -----
From: "Volodymyr Babin" <vbabin.ncsu.edu>
To: "AMBER Developers Mailing List" <amber-developers.ambermd.org>
Sent: Monday, February 01, 2010 10:12 PM
Subject: Re: [AMBER-Developers] questions about compiler flags
>> 1. Is anyone still familiar with the Large File support in Linux? My
>> questions are about the flags
>>
>> -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE
>>
>> added by configure. Specifically:
>>
>> a. Aren't these only needed for x86 (32-bit) compiles? Do we need them
>> for x86-64 (em64t)?
>> b. My Googling seems to indicate that only the first flag is required,
>> not
>> the second one; might this be true? Does anyone have a test of this
>> functionality?
>
> I am not familiar, but googling/looking into /usr/include/features.h
> makes me think that both flags may have effect on libc on both 32 and
> 64 bit platforms. These #defines seem to activate different libc features,
> and, assuming that both are needed, I would keep both flags set.
> (64-bit offset and functions that take off_t argument [fseeko, ftello]).
> It is possible that in some cases one flag implies the other, but I
> did not see this clearly stated anywhere.
>
>> 2. Do we really need the "-ip -O3 -xHost" flags for optimized ifort
>> compiles?
>> Is this better/worse/the same as "-fast" (used for pmemd)? Does sander
>> need
>> anything more than -O3? Is there a good compromise between "fastest of
>> all
>> 2^N possible combinations of flags on this particular machine" and
>> "something
>> that sacrifices a few percent in performance but which is likely to both
>> work
>> now and continue working after Intel puts out its next release"?
>
> Why not just have -O2 there and a README file containing more
> aggressive combinations for the curious.
>
> Best,
>
> Volodymyr
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Feb 02 2010 - 05:30:04 PST
