From: Joshua D. Moore (jdmoore_at_unity.ncsu.edu)
Date: Wed Jan 31 2007 - 14:47:14 CST
Hi,
I recently compiled a 32 bit version of NAMD 2.6 on an IBM blade cluster.
Most of the machines are dual core, but two are quad core. The version I
compiled seems to run fine on the dual core machines but not on the quad
core machines.
It was compiled with intel version 71 compilers for MPI, and gnu compilers
(g++ and gcc) for the native C and C++ compilers. icc and icpc compilers
didn't seem to work (as did intel 8.1 and 9.1 compilers).
Previously I had compiled version 2.6b1 on these machines and it seems to
run fine on both the dual core and the quad core machines. I used the
same compilers for this as 2.6. The only difference I can tell is that
this time I gave explicit links to the mpich libraries.
Does anyone know why I could be having this problem? (See below for files)
The charm megatests ran fine.
Thanks.
Josh
Here is my cham conv-mach.sh for mpi-linux
CMK_CPP_CHARM='/lib/cpp -P'
CMK_CPP_C='mpicc -E'
CMK_CC='mpicc '
CMK_CXX='mpicxx '
CMK_CXXPP='mpicxx -E '
CMK_CF77='mpif77'
CMK_CF90='mpif90'
CMK_RANLIB='ranlib'
CMK_LIBS='-L/usr/local/intel/mpich-1.2.6/linux86/7.1/lib -lckqt -lfmpich
-lmpichf90 -lmpich '
CMK_LD_LIBRARY_PATH="-Wl,-rpath,$CHARMLIBSO/"
CMK_NATIVE_LIBS=''
CMK_NATIVE_CC='gcc '
CMK_NATIVE_LD='gcc '
CMK_NATIVE_CXX='g++ '
CMK_NATIVE_LDXX='g++ '
CMK_NATIVE_CC='gcc '
CMK_NATIVE_CXX='g++ '
CMK_F90LIBS='-L/usr/local/intel/mkl721/lib/32 -L/usr/absoft/lib
-L/opt/absoft/lib -lf90math -lfio -lU77 -lf77math -lmkl_lapack
-lmkl_solver '
CMK_MOD_NAME_ALLCAPS=1
CMK_MOD_EXT="mod"
CMK_F90_USE_MODDIR=1
CMK_F90_MODINC="-p"
Here are my NAMD arch files
Linux-i686-MPI.arch:
include .rootdir/arch/Linux-i686-g++.arch
CHARMARCH = mpi-linux
Linux-i686-g++.arch:
NAMD_ARCH = Linux-i686
CHARMARCH = net-linux
CXX = g++
CXXOPTS = -O3 -march=pentiumpro -ffast-math -static
CC = gcc
COPTS = -O3 -march=pentiumpro -ffast-math -static
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:44:20 CST