[gmx-users] efficiency in HPC

[Albert]

Hello:

  I compiled 4.6.3 in HPC with following steps:


setenv OBJECT_MODE 64
/opt/cmake/2.8.9_new/bin/cmake .. 
-DCMAKE_INSTALL_PREFIX=/opt/gromacs/4.6.3 -DBUILD_SHARED_LIBS=OFF 
-DCMAKE_BUILD_TYPE=Release -DGMX_SOFTWARE_INVSQRT=OFF -DGMX_GPU=OFF 
-DGMX_MPI=ON -DGMX_THREAD_MPI
=OFF -DGMX_X11=OFF -DGMX_FLOAT_FORMAT_IEEE754=1 
-DGMX_IEEE754_BIG_ENDIAN_BYTE_ORDER=1 
-DGMX_IEEE754_BIG_ENDIAN_WORD_ORDER=1 
-DCMAKE_C_COMPILER="/usr/bin/mpcc_r" -DCMAKE_CXX_COMPILER="/usr/bin/mpCC_
r" -DCMAKE_C_FLAGS_RELEASE="-O3 -qstrict -qsimd=auto -qmaxmem=-1 
-qarch=pwr7 -qtune=pwr7" -DCMAKE_XX_FLAGS_RELEASE="-O3 -qstrict 
-qsimd=auto -qmaxmem=-1 -qarch=pwr7 -qtune=pwr7" -DFFTWF_LIBRARY="/o
pt/fftw/3.3.2/lib/libfftw3f.a" 
-DFFTWF_INCLUDE_DIR="/opt/fftw/3.3.2/include" 
-DCMAKE_LINKER="/usr/bin/mpcc_r"
gmake -j 16
gmake install

when I do the test, I found that the performance is really bad. It tooks 
more than one year for a job to be done while it only take 2 months in 
other machine. Does anybody have any idea where is the problem?

thanks a lot

Albert