/[escript]/trunk/doc/user/changes.tex
ViewVC logotype

Contents of /trunk/doc/user/changes.tex

Parent Directory Parent Directory | Revision Log Revision Log


Revision 6851 - (show annotations)
Tue Jul 9 03:07:07 2019 UTC (6 weeks, 3 days ago) by aellery
File MIME type: application/x-tex
File size: 16622 byte(s)
eScript version 5.4.


1
2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
3 % Copyright (c) 2003-2018 by The University of Queensland
4 % http://www.uq.edu.au
5 %
6 % Primary Business: Queensland, Australia
7 % Licensed under the Apache License, version 2.0
8 % http://www.apache.org/licenses/LICENSE-2.0
9 %
10 % Development until 2012 by Earth Systems Science Computational Center (ESSCC)
11 % Development 2012-2013 by School of Earth Sciences
12 % Development from 2014 by Centre for Geoscience Computing (GeoComp)
13 %
14 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
15
16 \chapter{Changes from previous releases}
17 \label{app:changes}
18
19 \subsection*{5.4}
20 \begin{itemize}
21 \item New deployment options: Flatpak, Docker, Anaconda (Linux)
22 \item More efficient interpolation in the line search in the LBFGS minimizer
23 \item Changes to default settings in the Solverbuddy
24 \item \escript now uses Trilinos by default
25 \item \escript now uses direct solvers by default when solving 2-dimensional problems
26 \item Removed all references to (deprecated) matplotlib.griddata in the examples
27 \item Numerous bug fixes
28 \item All unit tests are now working
29 \end{itemize}
30
31 \subsection*{5.3}
32 \begin{itemize}
33 \item Improvements to saveDataCSV, including support for \speckley and the ability to output reference IDs
34 \item Support for the Python numpy module, including the ability to output Data objects as numpy arrays
35 \item Both \finley and \dudley now support GMSH version 4.0 MSH files
36 \item Improved inversion in downunder.
37 \item \escript no longer supports CUDA
38 \item \escript no longer supports boomerAMG
39 \item \escript can now be compiled using gcc 8 or clang++ 3.8.1
40 \item Numerous bug fixes
41 \end{itemize}
42
43 \subsection*{5.1}
44 \begin{itemize}
45 \item Improvements to maths using complex values.
46 Note: Use the \texttt{promote()} method on Data objects (added in 5.0) to convert them to complex.
47 This does not change the value itself, (eg $7 \rightarrow 7+0j$) but matters for operations like \emph{sqrt} which behave differently with complex inputs.
48 \item Added \texttt{phase} operation to return the angular part/phase/arg of values.
49 \item Improved inversion.
50 \item Corrections to $\cdot\_transpose\_product$ operations.
51 \item Improved ``integration'' testing.
52 \item NetCDF support (if enabled), can now be built with support for netcdf-4 formatted files.\\
53 See \texttt{scons/templates/README_FIRST} for the options.
54 \item Support for lazy expressions with complex values.
55 \item Solving PDEs with complex coefficients requires Trilinos support.
56 Since escript requires features which are not
57 compiled in by default, please consult the install guide if you wish to make use of
58 this feature.
59 \item Note that in the $4.2.0.1$ release, the license for the escript source code was changed to ``Apache2''.
60 This should be less restrictive than the previous license.
61 \end{itemize}
62
63 \subsection*{5.0}
64 \begin{itemize}
65 \item Minimum Python version is now 2.7.
66 \item Remaining code switched from C to C++
67 \item Compilers are now required to support c++11 at minimum.
68 \item Some support for storing and manipulating complex values in Data objects.
69 \item Some solver configurations can solve PDEs with complex values.
70 \item Experimental support for Trilinos solvers and preconditioners.
71 \item \ripley by default no longer adjusts domain lengths when number of MPI ranks is unsuitable. A new function controls this behaviour.
72 \item Some C++ exceptions are now translated to equivalent Python exceptions (e.g. ValueError).
73 \item New \escript feature test function to interrogate build features at runtime.
74 \item Support for 'long' indices for very large local domains.
75 \item \escript is now more modular. New options can be used to select what to build.
76 \end{itemize}
77
78 \subsection*{4.0 to 4.1}
79 \begin{itemize}
80 \item Added multi-resolution \ripley domains
81 \item The gmshReader now supports reading with multiple processes
82 \item Using the help() function on some domains is now more informative
83 \item User guide updated with information on use of Dirac points
84 \item Minimizer misfit now available via the callback function
85 \item Specifying use of a direct solver without a direct solver being available now raises an exception rather than silently default to a non-direct solver
86 \item Synthetic seismic examples for various wave types now included in the examples distributed
87 \item Reading NetCDF files when the default value in the file is nan no longer causes issues
88 \item Various documentation updates and fixes in user guide, install guide, and code documentation
89 \item Various compatibility fixes with Python3
90 \item HTI and VTIWave models now support setting a q value for custom boundary conditions
91 \item Using a MPI-enabled gmsh with an MPI build of \escript should now reliably wait for gmsh to end
92 \item Added an example for voxet reading with \ripley
93 \end{itemize}
94
95 \subsection*{3.4.2 to 4.0}
96 \begin{itemize}
97 \item New spectral element domain, \speckley
98 \item CUDA-based solvers and block-diagonal system matrices in \ripley
99 \item DC Resistivity forward modeling
100 \item New build options:
101 \begin{itemize}
102 \item 'domains': allows only the specified domains to be built
103 \item 'prelaunch', 'launcher', 'postlaunch': allows build-time
104 customization of escript launcher
105 \item 'cuda', 'nvcc', 'nvccflags', 'thrust_prefix': path and flags for
106 CUDA compiler and thrust headers
107 \end{itemize}
108 \item Gmsh reader now \MPI parallelised
109 \item cycle and timestamp support for saveSilo and saveVTK
110 \item PDEs using \ripley now have per-PDE assemblers to allow domains to have
111 multiple PDEs of different types without conflict
112 \item General documentation updates, both guides and code commenting
113 \item Test framework compatible with older versions of Python
114 \item Removed BASHisms from \escript launcher
115 \item Data objects now have hasNaN() and replaceNaN() methods
116 \item readBinaryGrid() now supports scaling and 64-bit floats in non-native
117 byte order
118 \item Now compiles with Cray compiler
119 \item Wave models corrected
120 \item Equality checks on Data objects now throw exceptions
121 \item General changes to help compilers with \OPENMP parallelisation
122 \item Intel-based \OPENMP flavours now have proper thread binding
123 \item \ripley dirac points no longer get lost in domains with enormous
124 elements
125 \item lots of improvements in inversion toolbox \downunder and symbolic
126 toolbox
127 \end{itemize}
128
129 \subsection*{3.4.1 to 3.4.2}
130 \begin{itemize}
131 \item Changes to SolverOptions: Most of this is under the hood at this point
132 but any code which accesses parameters from a SolverOptions object like this:
133 \begin{python}
134 variable.getSolverOptions().DIRECT
135 \end{python}
136 will need to be modified to:
137 \begin{python}
138 SolverOptions.DIRECT
139 \end{python}
140
141 \item Data objects can be created using smoothed random data.
142 \item Support for OSX10.9
143 \item Support for additional problem types in downunder.
144 \item More efficient ripley domains for large compute clusters
145 \item Added symbolic toolbox and associated documentation
146 \item Ripley now supports Dirac delta functions.
147 \item Improved reading of netCDF files
148 \item Paso is now in its own namespace
149 \item Examples have been updated to use Python 3 behaviour (with future imports for Python 2).
150 \item Ripley supports reading from gzipped files.
151 \item Fixed fault in documentation generation
152 \item Fixes to keep clang++ happy
153 \end{itemize}
154
155
156 \subsection*{3.4 to 3.4.1}
157 \begin{itemize}
158 \item Renamed design.Design to design.AbstractDesign as a more explicit/descriptive name, {\bf this will break any existing custom implementation} until changed to match.
159
160 \item Efficiency improvements in esys.downunder inversions.
161 \item Implemented more CF conventions for the netCDF datasource and added GDAL support.
162 \item Strongly coupled joint inversion added.
163 \item Preliminary work on seismic modelling.
164 \item Added support for more input data types in ER Mapper files.
165 \item Improved stability for long MPI runs.
166 \item Added timeStepFormat option to \texttt{esys.escript.saveESD}.
167 \item Document generation process more stable.
168 \item Added support for smoothed random 2D data.
169 \item Documentation updated.
170 \item Improved read support for grid-reading.
171 \item Removed last usages of C compiler.
172 \item Efficiency improvements in \texttt{esys.downunder}.
173 \item Removed support for VSL random (\texttt{boost::random} is sufficient).
174 \item Fixes related to compatibility with Intel compiler.
175 \item Simplified module structure.
176 \item Various code cleanup and bug fixes.
177 \end{itemize}
178
179 \subsection*{3.3.1 to 3.4}
180 \begin{itemize}
181 \item This release does not use the support bundles from $3.0$ so building from source will be required in more cases.
182 \item In Debian and Ubuntu, there are now two packages. One containing escript itself and the other containing documentation.
183 This fits better with the way Debian does things.
184 \item esys.weipa can now write Voxet files for data on ripley.
185 \item All support for OpenDX has been dropped.
186 \item Improved documentation HTML API documentation.
187 \item Improvements to inversion module.
188 \end{itemize}
189
190
191 \subsection*{3.3 to 3.3.1}
192 \begin{itemize}
193 \item Source packages for Debian and Ubuntu can now be made directly by dpkg-source from the source tree\footnote{For this to work, you will need to place a file called \texttt{svn_version} containing the revision number in the root of the source.}.
194 \item non-uniform 1D interpolation (see Section~\ref{sec:interpolation}).
195 \item The minimum version of python required for escript is now $2.6$.
196 \item saveVTK and saveDX (previously deprecated) have been removed from the main escript module.
197 Please see the \module{weipa} documentation for export functionality.
198 \item The \module{downunder} inversion module is included.
199 \begin{itemize}
200 \item The documentation includes an inversion cookbook to get you started.
201 \item Most operations are possible with \escript's existing dependencies but some \emph{may} require \module{pyproj} or \module{gdal}
202 but \escript will let you know if that happens.
203 \end{itemize}
204
205 \end{itemize}
206
207
208 \subsection*{3.2.1 to 3.3}
209 \begin{itemize}
210 \item Experimental support for \pythonthree.
211 \item Parameter order for the table interpolation methods has changed to be consistent.
212 Please test your scripts if you use these functions.
213 \item setX on \finley and \dudley domains will now only accept coordinates from ContinuousFunction spaces.
214 This is to avoid some potentially nasty behaviour when using periodic boundary conditions.
215 You can still use setX, just make sure that you interpolate first.
216 \end{itemize}
217
218
219 \subsection*{3.1 to 3.2}
220 \begin{itemize}
221 \item The deprecated name for the launcher has been removed. To run scripts use
222 \emph{run-escript} not \emph{escript}.
223 \item \escript is no longer automatically imported by importing \finley.
224 You will need to import escript explicitly. (All of our example scripts do this anyway.)
225 \item An experimental version of the new Dudley domain is now available.
226 \item Various bug fixes and optimisations.
227 \item New algorithms for gmsh support.
228 \item Improvements to the AMG solver. AMG is the recommended solver for symmetric problems.
229 \item Fixed compilation issues using netcdf.
230 \item Redesigned configuration files to make it easier to compile from source
231 without finding the locations of all your libraries.
232 \item Faster rendering of documentation.
233 \item Documentation is now hyperlinked.
234 \item New data export module \weipa. The \texttt{saveVTK} functionality has
235 been moved into this module, and while calling \texttt{saveVTK} from the
236 \escript module still works it is discouraged and will be removed in a
237 future release.
238 \item New \texttt{esys.escript.DataManager} class for convenient checkpointing
239 and exporting of escript data.
240 \item \VisIt simulation interface for online data access and visualization.
241 \item Simpler interpolation and support for interpolation from 3D vectors.
242 \item HRZ lumping has been added and some clarification on how to use it.
243 \item Data objects populated with ``random'' values can be created.
244 \end{itemize}
245
246 \subsection*{3.0 to 3.1}
247 \begin{itemize}
248 \item The \emph{escript} launcher has been renamed to \emph{run-escript}. The
249 old name will still work in this version but will be removed in the future.
250 \item Lazy evaluation features have been improved and documented (see \Sec{sec:lazy}).
251 \item The \emph{escript} documentation now includes a new Cookbook which
252 demonstrates how to solve sample problems using escript.
253 \item Macro elements have been introduced.
254 \item The \var{saveDataCSV} method allows one or more \Data objects to be
255 exported in CSV format (see \Sec{sec:savedatacsv}).
256 \item \Data objects can be populated by interpolating from values in a table.
257 \item The new \var{getInfLocator} and \var{getSupLocator} functions in
258 \pdetools return Locators to a minimal/maximal point over the
259 data.
260 \item There is a new class to model fault systems (\var{esys.escript.faultsystems.FaultSystem}).
261 \item A beta version of an Algebraic Multigrid (AMG) solver is included.
262 \item Inverting square matrices larger than 3x3 is now permitted if escript is
263 compiled with Lapack support.
264 \item If escript is compiled with a modern compiler, then inf/sup/Lsup will now
265 report NaN, +/-inf as appropriate if those values appear in the data.
266 \item Data.setTags will take tag names as well as tag numbers.
267 \item The Scalar, Vector, Tensor, Tensor3, Tensor4 factory methods can now take
268 arrays/nested sequence like objects as their initial values.
269 \item escript.util.mkDir can now take a list of directories to create.
270 \item Behind the scenes, \PYTHON docstrings have been rewritten from epydoc to
271 restructured text.
272 \item Various other bug fixes and performance tweaks.
273 \end{itemize}
274
275 \subsection*{2.0 to 3.0}
276 \begin{itemize}
277 \item The major change here was replacing \module{numarray} with \numpy.
278 %% I've removed the link since it is no longer valid
279 % For general instructions on converting scripts to use numpy see\\
280 % \url{http://www.stsci.edu/resources/software_hardware/numarray/numarray2numpy.pdf}.\\
281 The specific changes to \escript are:
282 \begin{itemize}
283 \item getValueOfDataPoint() which returned a \module{numarray}.array has been
284 replaced by getTupleForDataPoint() which returns a \PYTHON tuple
285 containing the components of the data point. In the case of matrices or
286 higher ranked data, the tuples will be nested.
287 Use \numpy.array(data.getTupleForDataPoint()) if a \numpyNDA object is
288 required.
289 \item getValueOfGlobalDataPoint() has similarly been replaced by
290 getTupleForGlobalDataPoint().
291 \item \code{integrate(data)} now returns a \numpyNDA instead of a
292 \module{numarray}.array.
293 \end{itemize}
294 Any python methods which previously accepted \module{numarray} objects now
295 accept \numpy objects instead.
296
297 \item The way to define solver options for \LinearPDE objects has changed.
298 There is now a \SolverOptions object attached to the \LinearPDE
299 object which handles the options of solvers used to solve the PDE.
300 The following changes apply:
301 \begin{itemize}
302 \item The \method{setTolerance} and \method{setAbsoluteTolerance} methods
303 have been removed. Instead use \method{setTolerance} and
304 \method{setAbsoluteTolerance} on the \SolverOptions object. For example:
305 \method{getSolverOptions().setTolerance(\ldots)}
306 \item The \method{setSolverPackage} and \method{setSolverMethod} methods have
307 been removed. Instead use the methods \method{setPackage},
308 \method{setSolverMethod} and \method{setPreconditioner}. For example:
309 getSolverOptions().setPackage(\ldots).
310 \item The static class variables defining packages, solvers and
311 preconditioners have been removed and are now accessed via the
312 corresponding static class variables in \SolverOptions.
313 For instance use \method{SolverOptions.PCG} instead of
314 \method{LinearPDE.PCG} to select the preconditioned conjugate gradient
315 method.
316 \item The \method{getSolution} now takes no argument. Use the corresponding
317 methods of the\\ \SolverOptions object returned by
318 \method{getSolverOptions()} to set values,\\ e.g. use
319 \method{getSolverOptions().setVerbosityOn()} instead of argument
320 \code{verbose=True} and \method{getSolverOptions().setIterMax(1000)}
321 instead of argument \code{iter_max=1000}.
322 \end{itemize}
323 \end{itemize}
324

  ViewVC Help
Powered by ViewVC 1.1.26