| ARPREC
(ARbitrary PRECision package) is a library implementing arbitrary
precision arithmetic with integer, real and complex datatypes,
including common
transcendental functions and quadrature routines. It is implemented in
C++, and has Fortran90 interface. It exploits IEEE Floating-Point
Standard, and improves Bailey's MPFUN, both in speed and functionality. |
|
| PARATEC
(PARAllel
Total Energy Code) is a package designed primarily for a massively
parallel computing
platform and can run on serial machines. The code performs ab-initio
quantum-mechanical total energy calculations using pseudopotentials and
a plane wave basis set. The pseudopotentials are of the standard
norm-conserving variety (typically Hamann-Schulter-Chiang or
Trouillier-Martins). Forces and stress can be easily calculated with
PARATEC and used to relax the atoms into their equilibrium positions. |
|
| PEtot
is a pseudopotential planewave electronic structure calculation package
using density functional theory. It can use norm conservation
pseudopotential and ultra-soft pseudopotential. It includes LDA, LSDA
and GGA exchange and correlation functional.
It calculates the total energy and relaxes the atoms according to the
total energy. It is written with Fortran 90 and parallelized by
MPI. It includes special subroutines for charge density mixing,
real space
implementation of the nonlocal potential, iterative eigen state solver,
and isolated system (not periodic)
calculations. |
|
| SuperLU
is
aleading scalable solver for sparse linear systems using direct
methods. It
contains a number of performance-enhancing techniques
to improve the algorithms in SuperLU, such as blocking,
asynchronous scheduling, static pivoting, and 2D matrix-processor
mapping, which are the keys for achieving high speed and scalability on
modern cache-based superscalar processors, SMP, and MPP systems.
SuperLU is widely adopted in research and
commercial use, with thousands of users. It has been integrated into
the otherpopular solvers such as PETSc and Hypre. The DOE application
codes that use SuperLU include NIMROD, Omega3P, DSpice, NIKE, Trilinos,
and
several computational plasma physicsprojects at PPPL. SuperLU is used
commercially in Mathematica, FEMLAB,Python, and the HP mathematical
library, among others. |
|
| XBLAS
(Extended
and Mixed Precision BLAS)is part of a newly emerged BLAS Standard. In
XBLAS, the input and output arguments only allow single
or double, i.e. the same as normal BLAS. But at runtime, the
users can specify an internal precision level higher than that of the
input andoutput. The highest internal precision supported is 128-bits.
We use the double-double technique by Bailey and Briggs to implement
this extra precision. XBLAS can help solve many ill-contioned problems
that arise in numerical linear algebra, with very little speed
penalty. |
