Regarding FFTW routines usage in NAMD

From: Puneet Singh (singh.punit1990_at_gmail.com)
Date: Tue Oct 20 2015 - 07:39:50 CDT

Dear Sir,
I am trying to build namd with fftw , i have multiple version of fftw at
different locations . So to check which version of fftw my namd2 binary
refers i did *ldd namd2* :
    linux-vdso.so.1 => (0x00007fff7f2ff000)
    libdl.so.2 => /lib64/libdl.so.2 (0x0000003a1cc00000)
    libtcl8.5.so => /home/puneet/NAMD_BUILD/common_dependencies/lib/
libtcl8.5.so (0x00002b1785241000)
    libm.so.6 => /lib64/libm.so.6 (0x0000003a1c800000)
    libmpicxx.so.12 => /opt/intel//impi/
5.0.1.035/intel64/lib/libmpicxx.so.12 (0x00002b17855c4000)
    libmpifort.so.12 => /opt/intel//impi/
5.0.1.035/intel64/lib/libmpifort.so.12 (0x00002b17857e4000)
    libmpi.so.12 => /opt/intel//impi/5.0.1.035/intel64/lib/libmpi.so.12
(0x00002b1785a70000)
    librt.so.1 => /lib64/librt.so.1 (0x0000003a1d400000)
    libpthread.so.0 => /lib64/libpthread.so.0 (0x0000003a1c400000)
    libstdc++.so.6 => /usr/lib64/libstdc++.so.6 (0x0000003a28400000)
    libgcc_s.so.1 => /lib64/libgcc_s.so.1 (0x0000003a27000000)
    libc.so.6 => /lib64/libc.so.6 (0x0000003a1c000000)
    /lib64/ld-linux-x86-64.so.2 (0x0000003a1bc00000)
    libimf.so =>
/opt/intel/composer_xe_2015.0.090/compiler/lib/intel64/libimf.so
(0x00002b17861a3000)
    libsvml.so =>
/opt/intel/composer_xe_2015.0.090/compiler/lib/intel64/libsvml.so
(0x00002b178665e000)
    libirng.so =>
/opt/intel/composer_xe_2015.0.090/compiler/lib/intel64/libirng.so
(0x00002b17872ad000)
    libintlc.so.5 =>
/opt/intel/composer_xe_2015.0.090/compiler/lib/intel64/libintlc.so.5
(0x00002b17874b4000)

fftw is not there in the dependency list & running this namd2 binary
against apoa1 example with 500 steps (23 processes on 24 cores) finishes in
36 seconds without fftw routines!

So i wish to know weather the namd2 binary needs fftw library ? I have
doubts that i may not have configured properly ,The following command i am
using to configure namd2:

./*config* mpi-linux-x86_64-MPI --charm-arch mpi-linux-x86_64-ifort-mpicxx
--with-fftw --fftw-prefix $ROOT_DIR/common_dependencies --cxx icpc --cc
icc --cc-opts --with-tcl --tcl-prefix $ROOT_DIR/common_dependencies

I will be benchmarking against apoa1 example ,so, does solving against this
input file needs fftw routines or can fftw routines improve namd
performance (36 seconds) ?

Eagerly Awaiting your reply,

Thanks & Regards,
Puneet Singh

This archive was generated by hypermail 2.1.6 : Tue Dec 27 2016 - 23:21:25 CST