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
Thread-safe and more performant parser based on ExprTK (#387)
* Update to install instructions...
Point release 3.2.1. The version bump is due to using new manylinux2010 images.
* - Added ExprTK based parser and removed muParser. EXPRTK is faster than
muParser (see labnotes).
- All tests have been passing.
- Tweaks are made to exprtk operators/functions to make sure old
expressions used scripts remains valid.
* Added function ln(double v) to EXPRTK. Compatibility with muParser. All tests are passing.
* It is working fine now with Vinu model. Needs to fix the sbmlParserReader which uses numVars_.
* Some more cleanup. Testing on travis. Removed moose.Func(); user should
use `moose.Function`. moose.Func was disabled long back.
* GSolve also honors MOOSE_NUM_THREADS ...
* Local tests are passing. Fixed memory leak issue. A little refactoring and memory segv at the end of the test is now clean.
Removed unneeded/outdated tests and other scripts.
Added new tests based on Vinu model. It is very comprehensive which
tests multithreaded Ksolve/Gsolve, SBML, and moose.Function
Update to install instructions... (#386)
Point release 3.2.1. The version bump is due to using new manylinux2010 images.
* Thanks to clang++; caught another memory related bug. If this passes, do
a memory profiling.
* Minor cleanup. Removed memory allocation and copying at couple of places.
Still memory leaks in exprtk. I think I can't get rid of them.
Parser is a member object rather than member pointer; still some leak but better version than before.
* Renamed test folder for regular tests. Added custom test which uses 4 threads to run Ksolve.
Modified atol and rtol in assertion in one of the tests. Using parallel version of Ksolve/Gsolve everywhere by setting MOOSE_NUM_THREADS=$(nproc)
Increased `rtol` value to make sure tests passes with BOOST solver as well. When the boost's ode solvers are used, the number generated are slightly different.
* - Added a docker file to run travis pipeline locally.
- Fixed python2 seg-fault in new tests. The segv was because / means int division in python2.
* Converting some rdesigneur scripts to tests; since we need to be able to
plot in nonblocking mode, I've added `block` option in `rdes.display`
function.
Multiple tests using pydoc. They are based on rdesigneur tutorials.
install requires matplotlib for rdesigneur.
* - The seg-fault in rdesigneur tests are due to #388.
* In the ksolve, we partition the interval beforehand now. The
single-threaded and multi-threaded numbers are not the same. Fixed
tests to accomodate slightly different values resulted n multi-threaded
mode.
* Categorised tests. Also turned all doctest to normal assertion tests.
Moved alpha tests to fixme tests.
* All tests passing locally with both GSL and BOOST.
* Compilation fixes for the OSX build.
One test marked as python3 only. The pyrun expression uses python3
specific division and print function. Causes segfault with python2
(wontfix)
0 commit comments