Re: [AMBER] MPI with options "-ip" and "-axSTPW"

From: kurisaki <kurisaki.ncube.human.nagoya-u.ac.jp>
Date: Mon, 30 Apr 2012 19:18:22 +0900

Thank you for detailed explanation, Ben.
It can be helpful for me.
I will check it soon.

Yours sincerely,

                            IK


-----Original Message-----
From: Ben Roberts [mailto:ben.roberts.geek.nz]
Sent: Monday, April 30, 2012 7:11 PM
To: AMBER Mailing List
Subject: Re: [AMBER] MPI with options "-ip" and "-axSTPW"

Hi Ikuo,

On 30/04/2012, at 9:33 PM, kurisaki wrote:

> Dear Amber user and developers,
>
> Would you tell me any MPI programs whose mpicc can recognize the options,
"-ip" and "-axSTPW".
>
> I am now trying to compile Amber12 in My Linux machine combining intel
compiler and OPEN-MPI. The serial installation worked, but the parallel one
did not.
>
> I found that our "mpicc" did not recognize the options "-ip" and "-axSTPW"
in config.h and deleted them. Finally, I completed the compile. Compile test
went well.
>
> However, I am really interested in the effect of the options on the
computation efficiency.

These options serve to optimise the binary code for certain processor types.
Without them, as you've seen, the code will still go OK, but probably
somewhat slower. I myself haven't tested their effect on execution speed,
though.

I do know that they're version-dependent. The command-line options available
to the Intel compilers vary considerably from version to version.

A good place to start is to run "mpicc --show" (or the equivalent option,
depending on your MPI implementation). The "mpicc" is actually a wrapper,
and the "show" option will print out what the underlying command is. You can
then find out what compiler, or compiler version, mpicc is really using. (A
common mistake when using the Intel compilers is to use the default MPI
build, which is almost always a wrapper for the GNU compilers. If you're
doing this, and you wish to use MPI with the Intel compilers, you will need
to build a second MPI on your computer, using the Intel compilers to do so.)

Regards,
Ben
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber



_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Mon Apr 30 2012 - 03:30:04 PDT
Custom Search