Lenngren

Programming; Languages, Libraries and Tools

A note about compilers

The Intel compilers are the recommended compiler suite for Lenngren. In general the Intel compilers generates faster code than the GNU compilers. The GNU compilers are available for compatibility reasons and as a backup in case the Intel compilers has problems with a particular code.

The Intel compilers optimize very aggressively which may cause problems with certain codes. Be sure to verify your results when compiling with high levels of optimization. In rare cases reducing the optimization level may actually make the code run faster.

C

Three C compilers are available. The standard GNU gcc compiler, the Portland Group pgcc compiler and the Intel icc C compiler

Gnu GCC

On-line help:

GCC Homepage

> man gcc

> info gcc

PGI pgcc

Environment:

module add pgi

On-line help:

PGI Product Documentation

> man pgcc

Intel icc

Environment:

module add i-compilers

On-line help:

Intel C/C++ Compiler Information page

> icc -help

C++

Three C++ compilers are available. The standard GNU g++ compiler, the Portland Group pgCC compiler and the Intel icpc C/C++ compiler

GNU g++

On-line help:

GCC Homepage > man g++

> info g++

PGI pgCC

On-line help:

PGI Product Documentation

> man pgCC

Intel icpc

See information about icpc above.

Fortran77

Three Fortran77 compilers are available. The GNU g77 compiler, the Portland Group pgf77 compiler and the Intel ifort F77/F90 compiler

GNU g77

On-line help:

GCC/G77 Homepage

> man g77

> info g77

PGI pgf77

Environment:

module add pgi

On-line help:

PGI Product Documentation

> man pgf77

Intel ifort

Environment:

module add i-compilers

On-line help:

Intel Fortran Compiler Information page

> ifort -help

Known bug:

Sometimes ifort fails with a Segmentation fault. This is actually the license handler flexlm that segfaults. This bug can be circumvented by doing (in tcsh) unsetenv LM_LICENSE_FILE.

Fortran90

Two Fortran90 compilers are available. The Portland Group pgf90 compiler and the Intel ifort F77/F90 compiler

PGI pgf90

Environment:

module add pgi

On-line help:

PGI Product Documentation > man pgf90

Intel ifc

See information about ifort above.

MPI

MPI is a library specification for message-passing, proposed as a standard by a broadly based committee of vendors, implementors, and users.

The currently provided MPI implementation is Scali MPI Connect from Scali. More detailed information can be found in the Introduction to SCALI MPI .

Scali MPI

Environment:

module add scampi

module add i-compilers (to use the Intel compilers)

On-line help:

ScaMPI Users Guide

Performance Tools

mpiP - Lightweight MPI Performance Monitoring

mpiP is a quick and easy way to analyze the MPI bottlenecks in your application. Just link with the library as specified in the module. Here is an example of output from an 8 processor run of the MILC quantum chemistry code.

Environment:

module add mpiP

On-line help:

mpiP Home Page

PAPI - Performance Application Programming Interface

PAPI is a library to gather hardware performance information directly from the processor. It is used by performance tools as well as directly by application engineers.

Environment:

module add papi

On-line help:

PAPI Home Page

module show papi

PapiEx - Execute arbitrary application and measure hardware performance counters with PAPI

PapiEx is a performance analysis tool designed to transparently and passively measure the hardware performance counters of an application using PAPI. Note: To measure 32-bit floating-point operations you need to set the environment variable PAPI_PENTIUM4_FP to SSE_SP. To go back to measuring 64-bit floating-point operations set it to SSE_DP.

Environment:

module add papi monitor papiex

On-line help:

module show papiex

Numerical Libraries

Intel Math Kernel Libraries

The Intel Math Kernel Library (MKL) contains implementations of BLAS (level 1,2,3 and extended) and LAPACK, the CBLAS interface to BLAS and FFT and vector math functions. There are versions of the library tuned for Pentium III, Pentium 4, Xeon with EM64T and Itanium processors.

Environment:

module add mkl

On-line help:

module show mkl

Math Kernel Library Documentation

Intel MKL homepage

Chemistry

Gaussian 03

Gaussian is a connected system of programs for performing a variety of semi-empirical and ab initio molecular orbital calculations. Present revision of Gaussian is 03.D01.

Gaussian is associated with two types of parallelism:

• Shared memory parallelism on a single multi-cpu node. The default is to run on both processors of the EM64T node
• Distributed memory parallelism (using Linda) over multiple nodes.

This latest release supports both type of parallelism. The number of nodes should be set in the gaussian input file with the %nproclinda option

Environment:

module add gaussian/latest

Sample input:

$g03root/g03/tests/com

Batch script:

esubmit -n <nodes> -t <minutes> $g03root/g03/g03.easy jobname

On-line help:

Gaussian 03 Online Manual

> ghelp

Gamess

General Atomic and Molecular Electronic Structure System PRESENTLY, THE INSTALLATION OF GAMESS DOES NOT WORK!

Environment:

module add gamess

On-line help:

gmshelp

Sample input:

$GMSPATH/../../tests

Batch script:

esubmit -n <nodes> -t <minutes> $GMSPATH/gamess.easy jobname

NWChem

Highly parallelized Quantum chemistry/molecular dynamics code from Pacific Northwest National Laboratories (PNNL)

Environment:

module add nwchem

Batch script:

/pdc/vol/nwchem/4.7/bin/nwchem.easy

Examples:

/pdc/vol/nwchem/4.7/examples

Documentation:

NWChem home page

NAMD

A parallel, object-oriented molecular dynamics code designed for high-performance simulation of large biomolecular systems

Environment:

module add namd

Batch script:

$NAMDROOT/namd.easy

Documentation:

NAMD home page

Gromacs

A molecular dynamics package

Environment:

module add gromacs

Batch script:

/pdc/vol/gromacs/3.3.1/bin/gromacs.easy

Documentation:

Gromacs home page"

Dalton

An ab initio electronic structure program

Environment:

module add dalton

Batch script

/pdc/vol/dalton/2.0/bin/dalton.easy

MOPAC

A simple semi-empirical serial program

Environment:

module add mopac

Execution:

mopac input.mop

Home:

http://openmopac.net/

Multiphysics

Comsol

A finite-element based program for simulating multiphysics and single-physics applications.

Note that this code is not parallel.

Environment:

module add comsol

Documention:

Comsol home page

CEM

Efield

Package for solving the Maxwell equations.

Environment:

module add efield

Documention:

Efield home page