Last Modified 26 Mar 2002

SCTB: Installation

Version 0.80

The source code for Version 0.80 of sctb is designed to be compiled into a serial version for use on single processor PCs or workstations, or a parallel version for use on any MPI capable platform.

sctb is written in FORTRAN 90. In serial mode the code has been tested on a variety of CPUs, using various operating systems and compilers. It should be possible to run this code on any computer with a valid FORTRAN 90 compiler and the LAPACK library, provided an architecture dependent Makefile is edited to reflect the appropriate compile and link options. In parallel mode the code has been tested on a cluster of Linux machines, and on the IBM SMP.

SCTB has currently been tested on:

the IBM SMP at the Aeronautical Systems Center (ASC)
an Intel Pentium 4 based RedHat Linux 7.1 machine using the Absoft compiler.

The program will be tested on other computer architectures. Porting should involve only the changes in the Makefile to reflect the system's LAPACK, FORTRAN, and linker installation. In the meantime, users are encouraged to attempt to port this code to other systems using LAPACK and FORTRAN 90. Please report both problems and successes to bernstei@dave.nrl.navy.mil.

The source code for sctb is password protected. See Obtaining the Source Code for more information.

Once you have the downloaded package, you will need to extract the files from the gzipped tar file.

On Supported Systems

On supported systems, the procedure is as simple as typing

build_scc_nrl_tb

then answer some questions. The most important is the "System Type" which is one of the following possibilities:

parallel_ibm_smp
parallel_lam_linux_absoft
ibm_smp
sgi_o2k
sgi_o3k
linux_absoft

Then answer the choices about how to do the diagonalization and where the relevant library files are installed. The defaults are good for the IBM SMP at ASC, but may or may not be optimal anyplace else.

This procedure will create several executables in the directory bin_${SYSTEM_TYPE}/ where ${SYSTEM_TYPE} is the "System Type" above. These should include the following:

eval_energy.${MODEL}.${SYSTEM_TYPE}
eval_forces.${MODEL}.${SYSTEM_TYPE}
eval_virial.${MODEL}.${SYSTEM_TYPE}
eval_second_derivs.${MODEL}.${SYSTEM_TYPE}
test_forces.${MODEL}.${SYSTEM_TYPE}
test_virial.${MODEL}.${SYSTEM_TYPE}

wher ${MODEL} is either nrl_tb for the Gamma point only code or nrl_tb_k_sum for the version that sums over k-points in the Brillouin zone.

On Other Systems

If you're not at a supported site, you'll need to do much of this work by hand. You can look at the build_scc_nrl_tb script, but this is the general procedure:

In each of the directories tb_force_routine/ and generic_forces/ , make a file called Makefile.arch.${ARCH}/ , where ${ARCH} identifies the architecture and compiler used. I suggest copying an existing Makefile.arch.? file to start with.
In the directory generic_eval/ , make a file called Makefile.sys.${SYSTEM_TYPE} where ${SYSTEM_TYPE} starts with ${ARCH}, followed by an underscore and a label identifying the diagonalization routine used.
In tb_force_routine/ , run ./build ${ARCH} TB_MODEL=NRL_TB

 In  generic_forces/  run  
./build ${ARCH} FORCE_MODEL=tb
./build ${ARCH} FORCE_MODEL=tb_k_sum

In generic_eval/ run ./build ${SYSTEM_TYPE} nrl_tb
Make a directory called bin_${SYSTEM_TYPE}/ and copy the executables (names are listed above) from build.${SYSTEM_TYPE}/ to the binaries directories.


   sctb Home Page  
  Introduction  
  About Version 0.80  
  Installation  
  List of Files  
  Usage  
  Input Files  
  Output Files  

  Trouble Shooting  
  Appendix