Parallel Colt / News: Recent posts

- Fixed bug in solve() methods in linear algebra causing to return incorrect results for rectangular matrices.
- Fixed bug in HyBR causing incorrect results when a preconditioner was used.

- Added cern.colt.map.tlong (hashmaps holding (long -> long)).
- Added cern.colt.matrix.tlong (matrices holding long elements).
- Performance improvements in sparse matrix-vector product (column-compressed and row-compressed matrices).
- Some refactoring and cosmetic changes.

- Added JPlasma 1.0.
- New, multithreaded versions of dense Cholesky and QR decompositions (JPlasma is used).
- Added CSparseJ 1.0.
- Added sparse Cholesky, LU and QR decompositions (CSparseJ is used).
- Added SparseDoubleAlgebra and SparseFloatAlgebra.
- Added 2D and 3D dense floating point matrices with elements stored in multidimensional arrays (to avoid 2^31 size limitation).
- Removed JCublas.
- Removed corejava (replaced by String.format()).
- Fixed few bugs in MRNSD and CGLS.
- Some refactoring and many cosmetic changes.

- Added new aggregate() method in 1D dense matrices.
- Added Modified Residual Norm Steepest Descent (MRNSD) solver.
- Added Conjugate Gradient for Least Squares (CGLS) solver.
- Added new solve() method in DoubleQRDecomposition and FloatQRDecomposition.
- Fixed few bugs in HyBR.
- Performance improvements in row-compressed sparse matrices.

- Fixed bug in build.xml causing generation of incomplete source code distribution.
- Fixed bug in DelegateDoubleMatrix1D and DelagateFloatMatrix1D causing NullPointerException when toString() method was called.
- Fixed bug in JCublas.cublasInit() causing NullPointerException in some cases.
- Added DiagonalDoubleMatrix2D and DiagonalFloatMatrix2D.
- Some optimizations in complex matrices.
- JTransforms updated to version 2.3.
- Cosmetic changes.

- Added iterative solvers and preconditioners (adapted from Matrix Toolkits for Java).
- Added HyBR solver - a Hybrid Bidiagonalization Regularization.
- Added nonlinear optimization package.
- Added JCublas (currently used only for matrix-matrix multiplication).
- Added CCDoubleMatrix2D and CCFloatMatrix2D.
- Added CCMDoubleMatrix2D and CCMFloatMatrix2D.
- Added DenseColDoubleMatrix2D and DenseColFloatMatrix2D.
- Added OpenLongDoubleHashMap and OpenLongFloatHashMap.
- Added cern.colt.matrix.io (matrix reader/writer, adapted from Matrix Toolkits for Java).
- Added cern.colt.matrix.tint (matrices holding int elements).
- Added conversion methods in SparseDoubleMatrix2D and SparseFloatMatrix2D.
- Added unit tests for iterative solvers.
- Rewritten unit tests for matrices.
- JLAPACK replaced by netlib-java.
- Internal storage changed in SparseDoubleMatrix1D, SparseDoubleMatrix2D, SparseDoubleMatrix3D, RCMDoubleMatrix2D and their single precision equivalents.
- From now on, the trigonometric transforms are only supported by DenseDoubleMatrix1D, DenseDoubleMatri2D, DenseDoubleMatrix3D, DenseColDoubleMatrix1D, DenseColDoubleMatri2D, DenseColDoubleMatrix3D and their single precision equivalents.
- JTransforms updated to version 2.2.
- Some refactoring and many cosmetic changes.

- JTransforms updated to version 2.1

- A lot of refactoring (now each object of a given type has its own package).
- JTransforms updated to version 2.0 (no more power-of-two size limitation).
- Multithreaded implementation of Object matrices.
- Fixed bug in DComplex.inv() and FComplex.inv().

Initial release of Parallel Colt at SourceForge.

http://sourceforge.net/projects/parallelcolt/