1.1.0 [2013-01-08]
- Add support for solving singular problems with Krylov solvers (PETSc only)
- Add new typedef dolfin::la_index for consistent indexing with linear algebra backends.
- Change default unsigned integer type to std::size_t
- Add support to attaching operator null space to preconditioner
(required for smoothed aggregation AMG)
- Add basic interface to the PETSc AMG preconditioner
- Make SCOTCH default graph partitioner (GNU compatible free license, unlike ParMETIS)
- Add scalable construction of mesh dual graph for mesh partitioning
- Improve performance of mesh building in parallel
- Add mesh output to SVG
- Add support for Facet and cell markers to mesh converted from Diffpack
- Add support for Facet and cell markers/attributes to mesh converted from Triangle
- Change interface for auto-adaptive solvers: these now take the goal
functional as a constructor argument
- Add memory usage monitor: monitor_memory_usage()
- Compare mesh hash in interpolate_vertex_values
- Add hash() for Mesh and MeshTopology
- Expose GenericVector::operator{+=,-=,+,-}(double) to Python
- Add function Function::compute_vertex_values not needing a mesh argument
- Add support for XDMF and HDF5
- Add new interface LinearOperator for matrix-free linear systems
- Remove MTL4 linear algebra backend
- Rename down_cast --> as_type in C++ / as_backend_type in Python
- Remove KrylovMatrix interface
- Remove quadrature classes
- JIT compiled C++ code can now include a dolfin namespace
- Expression string parsing now understand C++ namespace such as std::cosh
- Fix bug in Expression so one can pass min, max
- Fix bug in SystemAssembler, where mesh.init(D-1, D) was not called before assemble
- Fix bug where the reference count of Py_None was not increased
- Fix bug in reading TimeSeries of size smaller than 3
- Improve code design for Mesh FooIterators to avoid dubious down cast
- Bug fix in destruction of PETSc user preconditioners
- Add CellVolume(mesh) convenience wrapper to Python interface for UFL function
- Fix bug in producing outward pointing normals of BoundaryMesh
- Fix bug introduced by SWIG 2.0.5, where typemaps of templated typedefs
are not handled correctly
- Fix bug introduced by SWIG 2.0.5, which treated uint as Python long
- Add check that sample points for TimeSeries are monotone
- Fix handling of parameter "report" in Krylov solvers
- Add new linear algebra backend "PETScCusp" for GPU-accelerated linear algebra
- Add sparray method in the Python interface of GenericMatrix, requires scipy.sparse
- Make methods that return a view of contiguous c-arrays, via a NumPy array, keep a reference from the object so it wont get out of scope
- Add parameter: "use_petsc_signal_handler", which enables/disable PETSc system signals
- Avoid unnecessary resize of result vector for A*b
- MPI functionality for distributing values between neighbours
- SystemAssembler now works in parallel with topological/geometric boundary search
- New symmetric assembler with ability for stand-alone RHS assemble
- Major speed-up of DirichletBC computation and mesh marking
- Major speed-up of assembly of functions and expressions
- Major speed-up of mesh topology computation
- Add simple 2D and 3D mesh generation (via CGAL)
- Add creation of mesh from triangulations of points (via CGAL)
- Split the SWIG interface into six combined modules instead of one
- Add has_foo to easy check what solver and preconditioners are available
- Add convenience functions for listing available linear_algebra_backends
- Change naming convention for cpp unit tests test.cpp -> Foo.cpp
- Added cpp unit test for GenericVector::operator{-,+,*,/}= for all la backends
- Add functionality for rotating meshes
- Add mesh generation based on NETGEN constructive solid geometry
- Generalize SparsityPattern and STLMatrix to support column-wise storage
- Add interfaces to wrap PaStiX and MUMPS direct solvers
- Add CoordinateMatrix class
- Make STLMatrix work in parallel
- Remove all tr1::tuple and use boost::tuple
- Fix wrong link in Python quick reference.
