[gmx-users] gmx4.6.1 performance problem "CPU acceleration: NONE"

[Mark Abraham]

On Thu, Apr 11, 2013 at 5:41 PM, Yorquant Wang <wangykoo at gmail.com> wrote:

> Hello:
>      I have installed GMX4.6.1 and GMX4.5.5 on the same cluster. There are
> 2 Xoen E5 2630CPU in one node on this cluster. I tested the performance of
> those two versions of GMX, found GMX4.6.1 only had 0.62 performance of
> GMX4.5.5.
>      When I checked the .log file of GMX4.6.1, I found
> ...
> CPU acceleration:   NONE
> FFT library:        FFTW 3.2 wrappers to MKL
> Large file support: enabled
> RDTSCP usage:       disabled
> Built on:           Mon Apr  8 10:46:58 CST 2013
> Built by:           root at node3 [CMAKE]
> Build OS/arch:      Linux 2.6.18-274.el5 x86_64
> Build CPU vendor:   Unknown, detect failed
> Build CPU brand:    Unknown, detect failed
>

Whatever your build system is was too weird/new for the auto-detection
system to cope with. And perhaps the assumption that your build CPU and
your mdrun CPU are the same is not valid.


> Build CPU family:   0   Model: 0   Stepping: 0
> Build CPU features:
> C compiler:         /public/software/mpi/openmpi-16-intel/bin/mpicc Intel
> icc (ICC) 12.1.0 20111011
> C compiler flags:     -std=gnu99 -Wall   -ip -funroll-all-loops  -O3
> -DNDEBUG
>
>       So I think my installed GMX4.6.1 failed to use AVX_256 kenerl.  The
> installed commend is below. Could someone give some suggestion how to solve
> this problem? Thank you very much!
>

Since the detection is failing for unknown reasons, you will need to select
an appropriate acceleration for your mdrun *execution* environment in your
call to CMake, e.g. with -DGMX_CPU_ACCELERATION=AVX_256.

Mark


> *$cmake ../ \*
>
> *-DCMAKE_INSTALL_PREFIX=/public/software/gromacs-4.6 \*
>
> *-DGMX_MPI=ON \*
>
> *-DCMAKE_C_COMPILER=/public/software/mpi/openmpi-16-intel/bin/mpicc \*
>
> *-DCMAKE_CXX_COMPILER=/public/software/mpi/openmpi-16-intel/bin/mpicxx \*
>
> * -DGMX_BINARY_SUFFIX=_mpi \*
>
> *-DFFTWF_INCLUDE_DIR=/public/software/mathlib/fftw331-float/include \*
>
> *-DFFTWF_LIBRARY=/public/software/mathlib/fftw331-float/lib/libfftw3f.so *
>
> *$make –j 6*
>
> *$make install*
>
> Cheers up
> --
> Yukun Wang
> PhD candidate
> Institute of Natural Sciences && College of Life Science, Shanghai Jiao
> Tong University
> Cell phone: 13621806236.
> China Shanghai
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>