You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* The default values changed in the commit (ba1065c) breaks example such
as
https://github.com/BhallaLab/moose-examples/blob/09ca09e68f163a0920653a9400214aff10b6ba16/snippets/rxdFuncDiffusion.py
. The sparse matrix has upper limit of number of rows and columns to
20,000. In the example linked, new values makes it impossible to run the
example as is.
To overcome this, we notify the user whenever this limit is reached and
ask them to set `diffLength` before setting up the `x1`.
On top of this, we can also revert default values to original larger
values. Not sure why these values were changed to the scale of 1e-6?
Unless the reason is known, I am not going to change current values. The
above linked example now works with this commit.
* Script which switches solver is now working fine. Forgot to uncomment `unzombifyModel` in `~Stoich` in previous PR #387.
* switch solver script is working fine now. Added to the test suite.
* During zombiefication, deepcopy of parser is expected. Added more tests
and fixed the deepcopy of parser issue. In Zombiefication, the parser
was not deep copied resulting in erronous values in couple of tests. There
is some memory leak now.
* Prepare a PR now. Test on travis. Some tests might fail with BOOST.
Memory leak is still there.
* 1. Removed copy constructor from moose::Function. It was not used anywhere.
Increased rtol value for a test. Multithreaded solvers gives different
results.
2. Fixed all memory leaks. Probably I don't understand how EXPRTK really
works. using smart pointers with EXPRTK caused memory leak.
3. Log memory leak on travis using valgrind. Install valgrind in docker images as well.
* During zombiefication, deepcopy of parser is expected. Added fix and a
test case.
* Added more tests and fixed the deepcopy of parser issue.
In Zombiefication, the parser was not deep coped resulting in erronous
values in couple of tets. There is some memory leak now (not very large,
8 bytes per variable per expression).
* Prepare a PR now. Test on travis. Some tests might fail with BOOST.
Memory leak is still there.
* removed copy constructor; it was not used anywhere. [skip ci]
Increased rtol value for a test. Multithreaded solvers gives different
results.
* Fixed all memory leaks. Probably I don't understand how EXPRTK really
works. using smart pointers with EXPRTK caused memory leak. Log memory
leak on travis using valgrind. Fixed tests. Let the EXPRK object manage
its memory.
* BOOST/GSL are only needed in ksolve directory: tweaked cmake to change
compilation flags on ksolve directory only. Build will be faster for various configurations.
Moved one test to fixme (Vinu's model).
* Updated gitlab pipeline. On a single threaded virtual machine, numThreads was set to 0 in couple
of tests. Fixed that. Install pyneuroml to the pipeline. Small leeway in a test.
* std can be close to zero. np.allclose may not handle it properly all the
time.
* boost solver with GSSA may have large variation.
* boost solver with GSSA may have large variation. Test for the sum of all errors.
* Increased the tolerance a bit more.
0 commit comments