uBLAS is a C++ template class library that provides BLAS level 1, 2, 3 functionality for dense, packed and sparse matrices. The design and implementation unify mathematical notation via operator overloading and efficient code generation via expression templates.

uBLAS provides templated C++ classes for dense, unit and sparse vectors, dense, identity, triangular, banded, symmetric, hermitian and sparse matrices. Views into vectors and matrices can be constructed via ranges or slices and adaptor classes. The library covers the usual basic linear algebra operations on vectors and matrices: reductions like different norms, addition and subtraction of vectors and matrices and multiplication with a scalar, inner and outer products of vectors, matrix vector and matrix matrix products and triangular solver. The glue between containers, views and expression templated operations is a mostly STL conforming iterator interface.

Dense, packed and sparse matrix classes are being tested with the CLAPACK test suite.

Known limitations:

- The implementation assumes a linear memory address model.
- Tuning was focussed on dense matrices.

- Overview
- Expression Concepts
- Container Concepts
- Iterator Concepts
- Storage
- Sparse Storage
- Vector
- Sparse Vector
- Vector Proxies
- Vector Expressions
- Matrix
- Triangular Matrix
- Symmetric Matrix
- Hermitian Matrix
- Banded Matrix
- Sparse Matrix
- Matrix Proxies
- Matrix Expressions

As main development platform for uBLAS we used MSVC 6.0 with Dinkumware STL. Other compilers known to accept the library are

- MSVC 6.0 with STLPort-4.5.3
- BCC 5.5
- GCC 2.95.x, 3.0.x, 3.1
- ICC 5.0, 6.0
- Comeau 4.2.x
- MWCW

You can download the current stable release of the source code from here . Prerequisite is the latest stable release of the Boost libraries Configuration, Timer and optionally SmartPointer.

You can also check out the latest version via anonymous CVS. Here's how:

cvs -d:pserver:anonymous@cvs.boost.sourceforge.net:/cvsroot/boost login

(password is empty)

cvs -d:pserver:anonymous@cvs.boost.sourceforge.net:/cvsroot/boost checkout boost/boost/numeric/ublas

cvs -d:pserver:anonymous@cvs.boost.sourceforge.net:/cvsroot/boost checkout boost/libs/numeric/ublas

If you have gzip installed on your system, you can speed up the transfer using

cvs -d:pserver:anonymous@cvs.boost.sourceforge.net:/cvsroot/boost -z9 checkout boost/boost/numeric/ublas

cvs -d:pserver:anonymous@cvs.boost.sourceforge.net:/cvsroot/boost -z9 checkout boost/libs/numeric/ublas

You can also view the CVS archive. You may find the library here . Documentation and test programs reside here .

uBLAS has no dedicated mailing list. Feel free to use the mailing lists of Boost .

uBLAS initially was written by Joerg Walter and Mathias Koch. We would like to thank all, which supported the development of this library: David Abrahams, Ed Brey, Fernando Cacciola, Juan Jose Gomez Cadenas, Beman Dawes, Bob Fletcher, Kresimir Fresl, Joachim Kessel, Toon Knapen, John Maddock, Jens Maurer, Alexei Novakov, Gary Powell, Joachim Pyras, Peter Schmitteckert, Jeremy Siek, Markus Steffl, Michael Stevens, Benedikt Weber, Martin Weiser, Marc Zimmermann and the members of Boost

Q: I'm running the uBLAS dense vector and matrix benchmarks. Why do I
see a significant performance difference between the native C and library
implementations?

A: uBLAS distinguishes debug mode (size and type conformance checks enabled,
expression templates disabled) and release mode (size and type conformance
checks disabled, expression templates enabled). Please check, if the preprocessor
symbol `NDEBUG`

of `cassert`

is defined. `NDEBUG`

enables release mode, which in turn uses expression templates.

Q: I've written some uBLAS tests, which try to incorrectly assign different
matrix types or overrun vector and matrix dimensions. Why don't I get a
compile time or runtime diagnostic?

A: uBLAS distinguishes debug mode (size and type conformance checks enabled,
expression templates disabled) and release mode (size and type conformance
checks disabled, expression templates enabled). Please check, if the preprocessor
symbol `NDEBUG`

of `cassert`

is defined. `NDEBUG`

disables debug mode, which is needed to get size and type conformance
checks.

Q: I've written some uBLAS benchmarks to measure the performance of matrix
chain multiplications like `prod (A, prod (B, C))`

and see a
significant performance penalty due to the use of expression templates.
How can I disable expression templates?

A: You do not need to disable expression templates. Please try reintroducing
temporaries using either `prod (A, `

*matrix_type*```
(prod (B, C)))
```

or `prod (A, prod<`

*matrix_type*```
> (B, C))
```

.

If you have a problem, or have found a bug, please send us a note .

Copyright (©) 2000-2002 Joerg Walter, Mathias Koch

Permission to copy, use, modify, sell and distribute this document is
granted provided this copyright notice appears in all copies. This document
is provided ``as is'' without express or implied warranty, and with no claim
as to its suitability for any purpose.

Last revised: 1/15/2003