/[escript]/trunk/scons/meerkat_options.py
ViewVC logotype

Log of /trunk/scons/meerkat_options.py

Parent Directory Parent Directory | Revision Log Revision Log


Links to HEAD: (view) (annotate)
Sticky Revision:

Revision 6799 - (view) (annotate) - [select for diffs]
Modified Mon Mar 25 05:53:58 2019 UTC (3 weeks, 3 days ago) by aellery
File length: 3207 byte(s)
Diff to previous 6797
I have rewritten the solverbuddy. Briefly:

1. The remaining AMG code has been removed from PASO.
2. If Trilinos is available, escript will now use it by default.
3. eScript will use a direct solver by default, (if one is available,) when solving 2 dimensional problems and an iterative solver, by default, when solving 3 dimensional problems. This can be changed by a user by manually specifying which solver to use.
4. There is a new option available, setHermitian(), that allows a user to specify when a coefficient matrix is Hermitian.
5. Symmetry information is always passed onto the Trilinos solver when this information is relevant.
6. All tests have been updated, when relevant, to reflect these changes.
7. I fixed a couple of undocumented bugs.



Revision 6797 - (view) (annotate) - [select for diffs]
Modified Tue Mar 5 04:51:59 2019 UTC (6 weeks, 2 days ago) by aellery
File length: 3379 byte(s)
Diff to previous 6782
Improved documentation for the escript subworld feature.



Revision 6782 - (view) (annotate) - [select for diffs]
Modified Wed Jan 30 06:38:11 2019 UTC (2 months, 2 weeks ago) by aellery
File length: 3387 byte(s)
Diff to previous 6771
Temporary fix to bug 433. Trilinos will now solve for the Hessian matrix used by the LBFGS algorithm in inversions.py using GMRES instead of CG.



Revision 6771 - (view) (annotate) - [select for diffs]
Modified Wed Jan 9 05:31:41 2019 UTC (3 months, 1 week ago) by aellery
File length: 3430 byte(s)
Diff to previous 6767
Updated the list of changes in the userguide

Revision 6767 - (view) (annotate) - [select for diffs]
Modified Fri Jan 4 02:48:53 2019 UTC (3 months, 1 week ago) by aellery
File length: 3430 byte(s)
Diff to previous 6766
Fixed a minor typo in inversions.py that was causing gravmag_netcdf.py to break.



Revision 6766 - (view) (annotate) - [select for diffs]
Modified Fri Jan 4 01:31:39 2019 UTC (3 months, 1 week ago) by aellery
File length: 3428 byte(s)
Diff to previous 6757
I have fixed the bug that was causing the tests run_gravity.py and run_inversion_gravmag_2d.py to fail on the buildbot gnu-omp-py3-paso-long.

The interpolation function I coded for minimizers.py finds the minimum of a parabola that goes through the points phi(alpha_lo) and phi(alpha_hi) and whose gradient matches phi(alpha) at alpha_lo. Occasionally the minimum of this function lies outside of the domain [alpha_lo, alpha_hi] and, when this occurs, the line search function breaks down. The interpolation function now checks that the minimum is within the interval and if it is not it falls back on the bisection method.



Revision 6757 - (view) (annotate) - [select for diffs]
Modified Fri Dec 14 05:53:07 2018 UTC (4 months ago) by aellery
File length: 3430 byte(s)
Diff to previous 6756
eScripts has a new feature!

Finley can now read gmsh 4.0 msh files!



Revision 6756 - (view) (annotate) - [select for diffs]
Modified Thu Nov 29 07:23:43 2018 UTC (4 months, 2 weeks ago) by aellery
File length: 3423 byte(s)
Diff to previous 6755
Temporarily undoing last commit.

Revision 6755 - (view) (annotate) - [select for diffs]
Modified Thu Nov 29 01:40:36 2018 UTC (4 months, 2 weeks ago) by aellery
File length: 3431 byte(s)
Diff to previous 6754
A large number of the tests in doc/examples were creating files during the testing phase and then not deleting them afterwards. I have gone through these tests and added some code to do a clean up once each test is done.



Revision 6754 - (view) (annotate) - [select for diffs]
Modified Wed Nov 21 06:54:51 2018 UTC (4 months, 3 weeks ago) by aellery
File length: 3423 byte(s)
Diff to previous 6751
Lutz - I have finished removing the algebraic multigrid stuff from PASO. 



Revision 6751 - (view) (annotate) - [select for diffs]
Modified Wed Nov 21 02:35:29 2018 UTC (4 months, 3 weeks ago) by aellery
File length: 3423 byte(s)
Diff to previous 6748
Fixed support bug 443 (escripts would not compile on openSUSE).



Revision 6748 - (view) (annotate) - [select for diffs]
Modified Thu Nov 15 23:50:51 2018 UTC (5 months ago) by aellery
File length: 3292 byte(s)
Diff to previous 6740
Fixed bug 450 (saveSilo tests were failing).

The tests were failing on meerkat because Python could not locate the example mesh files when WEIPA_WORKDIR was undefined. This is similar to the problem that was causing the dudley test run_inputOutput.py to fail.

While debugging, I noticed that the saveSilo test script was creating a lot of files and then not deleting them after the tests had run. I have changed a couple of lines to fix this.



Revision 6740 - (view) (annotate) - [select for diffs]
Modified Fri Nov 9 00:31:37 2018 UTC (5 months, 1 week ago) by aellery
File length: 3296 byte(s)
Diff to previous 6739
Updated the trilinos section of the install guide.



Revision 6739 - (view) (annotate) - [select for diffs]
Modified Wed Nov 7 23:58:41 2018 UTC (5 months, 1 week ago) by aellery
File length: 3263 byte(s)
Diff to previous 6738
I have removed support for boomerAMG from eScript. Specific changes are:

i) I have removed the definition of the variables "boomeramg", "boomeramg_prefix" and "boomeramg_libs" from SConstruct as well as the boomerAMG section of dependencies.py. Similarly, I have removed the boomerAMG section of paso/src/SConscript. Also, I have removed boomeramg from the report that scons prints out after compilation.
ii) I have removed the enum SO_PRECONDITIONER_BOOMERAMG from escriptcpp.cpp 
iii) I have removed the enum SO_PRECONDITIONER_BOOMERAMG from escriptcore/src/SolverOptions.h and have updated a few comments.
iv) I have removed the case for SO_PRECONDITIONER_BOOMERAMG from the switches in SolverBuddy::getSummary(), SolverBuddy::getName() and SolverBuddy::setPreconditioner() in SolverOptions.cpp
v) In paso/src/Preconditioner.h, I have redefined the structure Preconditioner_AMG_Root so that it does not include a pointer to a Preconditioner_BoomerAMG.
vi) I have updated a couple of files in paso (AMG.cpp, AMG_Root.cpp) to reflect the changes in this structure (specifically, the fact that prec->boomeramg is now meaningless).
vii) I have removed a preprocessor directive in paso/src/Options.h (#define PASO_BOOMERAMG 60) and have updated paso/src/Options.cpp to reflect this change by removing this case from several switches.
viii) The files paso/src/BOOMERAMG.h and paso/src/BOOMERAMG.cpp have been removed from the latest commit.

On meerkat, with these changes, escripts compiles without any problems.




Revision 6738 - (view) (annotate) - [select for diffs]
Modified Fri Nov 2 04:49:38 2018 UTC (5 months, 2 weeks ago) by aellery
File length: 2938 byte(s)
Diff to previous 6737
Fixed another minor testing bug in dudley and finley.

Revision 6737 - (view) (annotate) - [select for diffs]
Modified Fri Nov 2 04:05:47 2018 UTC (5 months, 2 weeks ago) by aellery
File length: 2942 byte(s)
Diff to previous 6736
Fixed bug in downunder test run_dcforward.py

Revision 6736 - (view) (annotate) - [select for diffs]
Modified Thu Nov 1 23:50:29 2018 UTC (5 months, 2 weeks ago) by aellery
File length: 2938 byte(s)
Diff to previous 6735
Minor bug fixes.

I fixed several errors that were preventing eScripts from compiling (on Debian Stretch) when CUDA was enabled.

A set of problems arose due to a preprocessor directive starting on line 869 of RipleyDomain.cpp that the compiler only includes in the specific case where CUDA is enabled but MPI is not. Specifically:
(i) the switch statement inside this directive did not explicitly handle several values of the enum variables in escript/SolverOptions.h and had no default case. As a result the compiler was bouncing back an error. I have resolved this issue by adding a default case to the statement.
(ii) the if statement following the switch in this preprocessor directive set the variable "package" to a non-existent enum "SO_PACKAGE_DEFAULT" which I believe is a typo for "SO_DEFAULT".

I have also fixed:
(iii) The file ripley_UnitTest.cpp was trying to include the nonexistent file "escript/Esys_MPI.h" instead of "escript/EsysMPI.h", causing the compiler to bounce back an error.




Revision 6735 - (view) (annotate) - [select for diffs]
Modified Wed Oct 31 23:59:03 2018 UTC (5 months, 2 weeks ago) by aellery
File length: 2580 byte(s)
Diff to previous 6726
Minor bug fix.

eScripts was not compiling correctly when PAPI was enabled because performance.cpp in paso does not contain a declaration of PasoException. I have added a single line to the top of this file that points the compiler towards PasoException.h. This has resolved the error.



Revision 6726 - (view) (annotate) - [select for diffs]
Modified Mon Oct 8 04:44:29 2018 UTC (6 months, 1 week ago) by aellery
File length: 2568 byte(s)
Diff to previous 6719
Fixed bug #435 (compile warnings under clang++).



Revision 6719 - (view) (annotate) - [select for diffs]
Modified Tue Sep 25 02:53:15 2018 UTC (6 months, 3 weeks ago) by aellery
File length: 2568 byte(s)
Diff to previous 6717
I have modified getNumpy so that it no longer accepts Data objects that contain complex numbers, due to a bug.



Revision 6717 - (view) (annotate) - [select for diffs]
Modified Fri Sep 21 04:32:52 2018 UTC (6 months, 3 weeks ago) by aellery
File length: 2500 byte(s)
Diff to previous 6715
Fixed the boost numpy linker bug.



Revision 6715 - (view) (annotate) - [select for diffs]
Modified Tue Sep 4 23:38:18 2018 UTC (7 months, 1 week ago) by aellery
File length: 3081 byte(s)
Diff to previous 6712
Minor bug fix

Revision 6712 - (view) (annotate) - [select for diffs]
Modified Wed Aug 29 00:02:21 2018 UTC (7 months, 2 weeks ago) by aellery
File length: 2636 byte(s)
Diff to previous 6709

eScripts has a new feature!!!


When compiled with boost version 1.63 or higher, the command getNumpy( [data] ) will return a numpy ndarray containing the values in the data object [data]. The code is optimised to use MPI. Please note that at the moment the code does not support masks or complex numbers.





Revision 6709 - (view) (annotate) - [select for diffs]
Modified Fri Aug 17 04:56:51 2018 UTC (8 months ago) by aellery
File length: 2483 byte(s)
Diff to previous 6707
Escript now has a new feature!

saveDataCSV will now write the reference id of the node or element as the first column of the output CSV file when the user calls on it using the flag "refid=True".



Revision 6707 - (view) (annotate) - [select for diffs]
Added Mon Aug 13 22:51:29 2018 UTC (8 months ago) by aellery
File length: 2375 byte(s)
saveDataCSV now works correctly.




This form allows you to request diffs between any two revisions of this file. For each of the two "sides" of the diff, enter a numeric revision.

  Diffs between and
  Type of Diff should be a

  ViewVC Help
Powered by ViewVC 1.1.26