Re: [AMBER-Developers] Intel compiler optimizations for Intel MPI not working for older versions

From: Daniel Roe <daniel.r.roe.gmail.com>
Date: Wed, 23 Mar 2016 09:33:47 -0600

The version checks seem to work fine for me (caveat: only tested with
Intel 14, Intel 12, and GNU 5). If no one has objections I will cherry
pick this change over to master.

-Dan

On Wed, Mar 23, 2016 at 7:41 AM, Rai Nitin <rai.nitin.tcs.com> wrote:
> The configure script has been modified in the intel branch to address the
> issues:
>
> 1. Intel specific optimization flags are enabled only for Intel compilers
> version 13 to 16.
> 2. OpenMP in Amber build is enabled only for GNU compiler version 4.2 and
> above.
> 3. OpenMP in Amber build is enabled only for Intel compiler version 10.2
> and above.
>
> In case this works fine, these changes can be merged into master branch.
>
> Thanks,
> Nitin Rai
>
>
>
> From: Daniel Roe <daniel.r.roe.gmail.com>
> To: AMBER Developers Mailing List <amber-developers.ambermd.org>
> Date: 23-03-2016 08:53
> Subject: [AMBER-Developers] Intel compiler optimizations for Intel
> MPI not working for older versions
>
>
>
> Hi,
>
> I've been debugging a build of Amber with an older version of Intel
> compilers with Intel MPI (icc 12.0.0, impi 4.0.1). While building
> pmemd I hit this error:
>
> ifort: command line error: Unrecognized keyword 'array64byte' for
> option '-align'
>
> It seems this flag is a relatively recent introduction (hash ID
> 7b191fe02, Jan. of this year) and only gets added when intel
> compilers are used with intel MPI. If I comment out this section of
> configure2 (between lines 1481 and 1489) the compilation finishes
> successfully. After some more testing the problematic flags seem to be
> '-opt-streaming-cache-evict=0' as well as '-align array64byte'; the
> former causes object files to not be generated by mpiifort.
>
> Based on when these flags were added it seems their addition to
> configure2 is related to the experimental intel MPI/OpenMP code. Do we
> really need these flags for that or will it build without? If the
> flags are necessary, maybe some sort of test should be introduced to
> make sure the underlying compilers support them, or they are only on
> when the MPI/OpenMP build is?
>
> -Dan
>
> --
> -------------------------
> Daniel R. Roe, PhD
> Department of Medicinal Chemistry
> University of Utah
> 30 South 2000 East, Room 307
> Salt Lake City, UT 84112-5820
> http://home.chpc.utah.edu/~cheatham/
> (801) 587-9652
> (801) 585-6208 (Fax)
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
> =====-----=====-----=====
> Notice: The information contained in this e-mail
> message and/or attachments to it may contain
> confidential or privileged information. If you are
> not the intended recipient, any dissemination, use,
> review, distribution, printing or copying of the
> information contained in this e-mail message
> and/or attachments to it are strictly prohibited. If
> you have received this communication in error,
> please notify us by reply e-mail or telephone and
> immediately and permanently delete the message
> and any attachments. Thank you
>
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers



-- 
-------------------------
Daniel R. Roe, PhD
Department of Medicinal Chemistry
University of Utah
30 South 2000 East, Room 307
Salt Lake City, UT 84112-5820
http://home.chpc.utah.edu/~cheatham/
(801) 587-9652
(801) 585-6208 (Fax)
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Mar 23 2016 - 09:00:04 PDT
Custom Search