May 2, 2014. mutils-0.4-2
------------------------
- Voronoi diagram output in triangle mex file. triangle returns -1
for 'infinite ray' edges. In the triangle mex file this value is
converted to 0.
- Exact predicates by Shewtchuk added to point location routines.
In this release exact predicates are used by default. The code is
routhly twice slower, but robust in cases when points lie very close
to triangle edges/tetrahedron sides. NOTE: next release will
not have exact predicates turned on by default, since the native
mutils code will be made more robust in the hard cases.
- Point in tetrahedron implemented for tsearch2 - see
ex_tsearch_3d in examples directory.
- bug fixes
The MEX files provided with the release are compiled as follows:
- Linux: Ubuntu 14.04, gcc 4.8, with OpenMP
- MacOS: OS X Maverics, Xcode 5.0.2, no Openmp
- Windows: Windows 7, with OpenMP
Sep 6, 2013. mutils-0.4-1
------------------------
This is a minor bugfix release:
- fix compilation on MacOS with XCode only (no gcc from macports)
- the default compiled mex files for Mac are compiled using Xcode 4.6 on
Mountain Lion 10.8.2 WITH NO OPENMP SUPPORT. There is a better change
they will run out of the box. Previously, mex files compiled with macports
lacked libraries on some systems.
Mex files are compiled WITHOUT OpenMP support on Windows and Mac,
and WITH OpenMP on Linux. To change compile options, edit mutils_config.m.
Jun 20, 2013. mutils-0.4
------------------------
This release contains several new features and significant code changes:
- cpu affinity in parallelized mex files if off by default. It can be
turned on by an option passed to the mex function.
- number of threads in OpenMP parallelized mex files can now be set using
nthreads field of opts structure
- FEM interpolation implemented using SSE2/AVX on supporting CPUs, for
triangular and quad elements, which gives significant speedups.
- sorted lists implemented using SSE/AVX: considerable speed
improvements
- einterp returns local coordinates of markers in containing elements
- sparse_create is parallelized (see ex_sparse)
- METIS is patched to use 64-bit integers, hence it works for large
matrices with more than 2^32 non-zero entries. This applies both to
metis_part, and to SuiteSparse metis interface. To disable 64-bit
metis in SuiteSparse, in suitesparse_install.m comment out copyfile
lines below "% apply 64-bit patch for METIS 4" comment
The mex files distributed by default are compiled with SSE2 only.
If your compiler and CPU support it, you might want to recompile
with appropriate compiler options. You can change them in mutils_config.m.
Oct 23, 2012. mutils-0.3
------------------------
This release contains several new features and bug fixes:
- parallel SpMV (look at ex_spmv example)
- METIS graph partitioning MEX function (metis_part)
- fast RCM reordering MEX function (mrcm)
- einterp MEX function now supports a range of different elements:
tri3, tri7, quad4, quad9 for 1 and 2 degrees of freedom per node
(number of interpolated values per node)
- compiles on Windows (VS), Linux (gcc) and Mac
(needs Xcode + additionally installed gcc, see INSTALL for details)
Also improved documentation
- extensive documentation in the MATLAB scripts
- HTML reports generated for all scripts in examples/ directory.
Can be found in Documentation.
Aug 31, 2012. mutils-0.2
------------------------
This is a quick bugfix release. There were some problems with
quadtree('locate') that caused segfaults in certain cases.
In addition, this distribution contains binary compiled mex files
for Windows (using the Windows SDK compiler), and for Linux using
gcc.
Note that the Windows binaries are NOT compiled with OpenMP support,
so there are only sequential MEX files.
Aug 8, 2012. mutils-0.1
------------------------
MUTILS is a package that provides a set of utilities we find useful
for high-resolution modeling. The tools are written in C and made
available in MATLAB through the MEX interface, hence installation
requires a working MEX compiler. The code is optimized for modern
multi-core CPUs. Some functions are parallelized using OpenMP.
Initial release contains the following functions:
sparse_create a routine that assembles sparse matrices resulting
from the Finite Element Method. It is significantly
faster than MATLABs sparse, and uses a lot less
memory.
quadtree a 2D quad-tree implementation with point-in-triangle
location and spatial reordering functionality.
tsearch2 a quad-tree based point-in-triangle location routine,
similar to MATLAB’s tsearch, but significantly
faster and parallelized with OpenMP.
einterp a FEM interpolation routine using SSE2 instructions and
OpenMP parallelization. Currently only the 7-node
triangular element is supported, with more elements
coming at users request.
A short documentation of the MEX functions is available in the *.m
files and can be accessed from MATLAB by typing e.g., help quadtree.
mutils come with a set of examples - have a look in the examples/
directory.
In addition, mutils provide scripts to download and install external
software packages:
triangle by Jonathan Richard Shewchuk. The mutils package provides
MEX interface for triangle
SuiteSparse by Tim Davis.