Forever And Always 0.9 DEMO Mac OS
Forever And Always 0.9 DEMO Mac OS
- Choose a tag to compare
- Forever And Always 0.9 Demo Mac Os Catalina
- Forever And Always 0.9 Demo Mac Os Download
- Forever And Always 0.9 Demo Mac Os X
- Forever And Always 0.9 Demo Mac Os Downloads
- Choose a tag to compare
StuffIt Standard Edition 8.0.2 - PPC Mac OS X native. Includes StuffIt Expander, DropStuff, DropTar and DropZip. StuffIt Deluxe 8 - Full install for PPC Mac OS X. StuffIt Deluxe 10 - Full install for PPC Mac OS X. StuffIt Deluxe 11 - Full install for PPC Mac OS X. StuffIt Deluxe 2011 - Full install for PPC Mac OS X. Lnx-update-tools.cmd – retrieves latest OS X guest tools; Mac OS X: On Mac OS X you will need to be either root or use sudo to run the scripts. This is really only needed if you want to use client versions of Mac OS X. You may need to ensure the OS X scripts have execute permissions by running chmod +x against the 2 files.
bartoldeman released this
0.9.2:
Forever And Always 0.9 Demo Mac Os Catalina
- Main AUTO
- Print eigenvalues in Hopf detection function if following a
BP curve or an Rn/BT has been detected. - Don't write MAX/MIN for IPS-2, that is, IVPs (request by Sebius).
- Replace tabs with spaces in continuation lines (bug reported by Sebius).
- Fix warnings reported by newer GCCs.
- Update included BLAS from www.netlib.org/blas (LAST UPDATE:
Tuesday Apr 19th 2011) - Use 3-digit exponents for all 19.10 formats (in solution files),
to avoid numbers such as 1.1234567891-100. - Fix constant file output for strings with backslashes, using ''+str(s)+''
instead of repr(s) (report from Sebius Doedel). - Thanks to Felix Becker felix.becker@zih.tu-dresden.de: fix error detected
by -fcheck=bounds in ae.f90 and bvp.f90. - Use TRIM(ATYPE)//'0' instead of ATYPE//'0' to fix a problem detecting
spurious Torus bifurcations. Thanks to Medhi Paak for reporting this bug. - Change sorting algorithm for Floquet multipliers from selection to insertion
sort to keep complex conjugates together correctly. Solves sorting
issue reported by Medhi Paak and Sebius Doedel. - Only compute L2-NORM (or similar) for WRTBV9 if IID>=2, else it is
not used. - Better MPI parallelization:
- The code in bvp.f90 has been parallelized so that the solution
vectors are nearly always distributed and only concatenated when
Newton iterations converge (for mesh adaptation, solution
measures, and I/O). This speeds up the code significantly for higher
processor counts. - Some work space arrays used in the linear solver now have the
minimum size required (of length NTST/NP+log2(NTST) instead of
length NTST where NP is the number of processors). This significantly
reduces memory consumption. - HomCont works with MPI.
- BPCont works with MPI.
- MPI is properly ignored for algebraic problems so it can be
easily used in scripts and automatic testing, by setting e.g.
AUTO_COMMAND_PREFIX='mpirun -n 4' - AUTO now calls MPI_Init_thread instead of MPI_Init for hybrid
MPI/OpenMP configurations
- The code in bvp.f90 has been parallelized so that the solution
- Better OpenMP parallelization:
- Call STUPBV (time derivatives) in parallel for OpenMP.
- Parallellize part of fort.8 I/O (FP to string conversion) using OpenMP.
- Store UOLD vector into U past NDIM. This allows equation files to
access UOLD in FUNC. - Meshing improvements:
- Let INTERP work on two solutions in one loop, to save on recomputing weights.
- Simplify INTWTS interface/implementation to remove the need for the array X
and with less potential loss of precision. - Store 'E' (deqf/dt) from EQDF into DEQF and use it directly in NEWMSH
This avoids an addition followed by a subtraction which resulted in
rounding errors.
- Move to git repository, add .gitignore file (with Andreas Amann)
- Adapt configure script, makefiles and shell scripts to newer Mac OS and
Windows (using MSYS2).
- Print eigenvalues in Hopf detection function if following a
- Python:
- Python 3 support
- Adapt for Python3 compatibility (Andreas Amann)
All tests now run for Python3 and Python2 - Python3: only use binary mode for solutions, not constants.
Solution files use binary mode for speed since they only contain
numbers and no strings, unlike constants files, but a text mode
workaround is there now just in case the stdin pipe has no binary
buffer.
- Adapt for Python3 compatibility (Andreas Amann)
- Python packaging support
- Add setup.py/setup.cfg and move Python code to auto subdirectory
(Rike-Benjamin Schuppner) - Add init.py and use relative imports (Rike-Benjamin Schuppner)
- Add main.py to simplify testing, and allow 'python /path/to/auto'
- Python: Change relative to absolute imports and adjust commands.
Absolute imports work more nicely with existing scripts, and are
recommended by PEP8 anyway.
- Add setup.py/setup.cfg and move Python code to auto subdirectory
- IPython support
- Support IPython >= 0.11, thanks to Andreas Amann a.amann@ucc.ie.
- included support for IPython >1.x
- Prompts need to be defined differently with IPython >= 5.
- Use numpy.ndim instead of the deprecated numpy.rank.
- Remove support for long (~10 years) obsolete Numeric and numarray.
- Avoid identifying np.nonzero and np.flatnonzero (Andreas Amann)
Monkey patching np.nonzero to be np.flatnonzero had bad effects on
matplotlib, which depends on the correct np.nonzero to be available. - Avoid plain python in shebangs.
'python' may not exist, but 'python3' may exist. The auto and autox shell
scripts now take care of this.
- Python 3 support
- PyPLAUT
- For Python 3: fix handling of dashes and parser for autorc files.
- If Tk cannot be initialized, hide the plot if possible.
This allows to run the automated test without DISPLAY set. - Use NavigationToolbar2Tk for matplotlib 2.2+
- Use get/set_facecolor if available instead of deprecated functions.
- Fix problem when updating small plotting window
- PLAUT
- Fix compilation with newer versions of gfortran:
ANGLE needs 4 elements, otherwise it cannot call ARROWL with it.
- Fix compilation with newer versions of gfortran:
- PLAUT04
- Eliminate use of QT3_SUPPORT methods for pulldown menus.
Eliminate QT3_SUPPORT define. - Qt4 compat for file dialog windows.
- Replace insertItem with addItem for lists for Qt4 compatibility.
- Use toLocal8Bit() instead of implicit QString->const char * conversions.
- Call constructors for QComboBox, QSpinBox, QDialog, etc. in Qt4 style.
- Create a seperate method for every pulldown menu entry for better
QT4 compatibility. - Workaround for the menu bar in newer Mac OS X versions (Mavericks+)
- Fix set but not used warnings for PLAUT04.
- Eliminate use of QT3_SUPPORT methods for pulldown menus.
- Testing
- Setup automated testing (Continuous Integration) using Github
Actions (on Ubuntu)- Compile AUTO and manual
- Test with both python2 and python3.
- Enabled test with MPI enabled
- Properly exit with error code in test scripts.
- Upload test artifacts (logfile and verification html).
- Adjust r3b demo test reference.
- The eighth run of the fhh demo converges (since 703607e for NEWMSH)
- Fix two more results for ab demo in test reference.
- Setup automated testing (Continuous Integration) using Github
- Manual
- Fix documentation illustrating Python data access for lrz demo. Closes #3.
- Latex compile fix (Andreas Amann)
the unknown m placement specifier for figures was ignored by older
latex versions but is now an error.
- Demos
- Make tfc demo more robust.
- Tweak EPS* in fhh demo to properly detect UZ in eighth run
- Adjust EPSS in fhh demo to still detect the test point.
- Increase NTST in ab demo to make it more robust in the last run.
- Increase NTST in c.ab.3 to 70 so it does not MX unreliably.
- c2c demo: increase NTST a little to make it work correctly with GCC8
- Increase NTST in c.fhn.5 to match c.fhn.6 (elimates spurious UZ)
- Choose a tag to compare
Forever And Always 0.9 Demo Mac Os Download
- Choose a tag to compare
bartoldeman released this
Forever And Always 0.9 Demo Mac Os X
0.9.2:
- Main AUTO
- Print eigenvalues in Hopf detection function if following a
BP curve or an Rn/BT has been detected. - Don't write MAX/MIN for IPS-2, that is, IVPs (request by Sebius).
- Replace tabs with spaces in continuation lines (bug reported by Sebius).
- Fix warnings reported by newer GCCs.
- Update included BLAS from www.netlib.org/blas (LAST UPDATE:
Tuesday Apr 19th 2011) - Use 3-digit exponents for all 19.10 formats (in solution files),
to avoid numbers such as 1.1234567891-100. - Fix constant file output for strings with backslashes, using ''+str(s)+''
instead of repr(s) (report from Sebius Doedel). - Thanks to Felix Becker felix.becker@zih.tu-dresden.de: fix error detected
by -fcheck=bounds in ae.f90 and bvp.f90. - Use TRIM(ATYPE)//'0' instead of ATYPE//'0' to fix a problem detecting
spurious Torus bifurcations. Thanks to Medhi Paak for reporting this bug. - Change sorting algorithm for Floquet multipliers from selection to insertion
sort to keep complex conjugates together correctly. Solves sorting
issue reported by Medhi Paak and Sebius Doedel. - Only compute L2-NORM (or similar) for WRTBV9 if IID>=2, else it is
not used. - Better MPI parallelization:
- The code in bvp.f90 has been parallelized so that the solution
vectors are nearly always distributed and only concatenated when
Newton iterations converge (for mesh adaptation, solution
measures, and I/O). This speeds up the code significantly for higher
processor counts. - Some work space arrays used in the linear solver now have the
minimum size required (of length NTST/NP+log2(NTST) instead of
length NTST where NP is the number of processors). This significantly
reduces memory consumption. - HomCont works with MPI.
- BPCont works with MPI.
- MPI is properly ignored for algebraic problems so it can be
easily used in scripts and automatic testing, by setting e.g.
AUTO_COMMAND_PREFIX='mpirun -n 4' - AUTO now calls MPI_Init_thread instead of MPI_Init for hybrid
MPI/OpenMP configurations
- The code in bvp.f90 has been parallelized so that the solution
- Better OpenMP parallelization:
- Call STUPBV (time derivatives) in parallel for OpenMP.
- Parallellize part of fort.8 I/O (FP to string conversion) using OpenMP.
- Store UOLD vector into U past NDIM. This allows equation files to
access UOLD in FUNC. - Meshing improvements:
- Let INTERP work on two solutions in one loop, to save on recomputing weights.
- Simplify INTWTS interface/implementation to remove the need for the array X
and with less potential loss of precision. - Store 'E' (deqf/dt) from EQDF into DEQF and use it directly in NEWMSH
This avoids an addition followed by a subtraction which resulted in
rounding errors.
- Move to git repository, add .gitignore file (with Andreas Amann)
- Adapt configure script, makefiles and shell scripts to newer Mac OS and
Windows (using MSYS2).
- Print eigenvalues in Hopf detection function if following a
- Python:
- Python 3 support
- Adapt for Python3 compatibility (Andreas Amann)
All tests now run for Python3 and Python2 - Python3: only use binary mode for solutions, not constants.
Solution files use binary mode for speed since they only contain
numbers and no strings, unlike constants files, but a text mode
workaround is there now just in case the stdin pipe has no binary
buffer.
- Adapt for Python3 compatibility (Andreas Amann)
- Python packaging support
- Add setup.py/setup.cfg and move Python code to auto subdirectory
(Rike-Benjamin Schuppner) - Add init.py and use relative imports (Rike-Benjamin Schuppner)
- Add main.py to simplify testing, and allow 'python /path/to/auto'
- Python: Change relative to absolute imports and adjust commands.
Absolute imports work more nicely with existing scripts, and are
recommended by PEP8 anyway.
- Add setup.py/setup.cfg and move Python code to auto subdirectory
- IPython support
- Support IPython >= 0.11, thanks to Andreas Amann a.amann@ucc.ie.
- included support for IPython >1.x
- Prompts need to be defined differently with IPython >= 5.
- Use numpy.ndim instead of the deprecated numpy.rank.
- Remove support for long (~10 years) obsolete Numeric and numarray.
- Avoid identifying np.nonzero and np.flatnonzero (Andreas Amann)
Monkey patching np.nonzero to be np.flatnonzero had bad effects on
matplotlib, which depends on the correct np.nonzero to be available. - Avoid plain python in shebangs.
'python' may not exist, but 'python3' may exist. The auto and autox shell
scripts now take care of this.
- Python 3 support
- PyPLAUT
- For Python 3: fix handling of dashes and parser for autorc files.
- If Tk cannot be initialized, hide the plot if possible.
This allows to run the automated test without DISPLAY set. - Use NavigationToolbar2Tk for matplotlib 2.2+
- Use get/set_facecolor if available instead of deprecated functions.
- Fix problem when updating small plotting window
- PLAUT
- Fix compilation with newer versions of gfortran:
ANGLE needs 4 elements, otherwise it cannot call ARROWL with it.
- Fix compilation with newer versions of gfortran:
- PLAUT04
- Eliminate use of QT3_SUPPORT methods for pulldown menus.
Eliminate QT3_SUPPORT define. - Qt4 compat for file dialog windows.
- Replace insertItem with addItem for lists for Qt4 compatibility.
- Use toLocal8Bit() instead of implicit QString->const char * conversions.
- Call constructors for QComboBox, QSpinBox, QDialog, etc. in Qt4 style.
- Create a seperate method for every pulldown menu entry for better
QT4 compatibility. - Workaround for the menu bar in newer Mac OS X versions (Mavericks+)
- Fix set but not used warnings for PLAUT04.
- Eliminate use of QT3_SUPPORT methods for pulldown menus.
- Testing
- Setup automated testing (Continuous Integration) using Github
Actions (on Ubuntu)- Compile AUTO and manual
- Test with both python2 and python3.
- Enabled test with MPI enabled
- Properly exit with error code in test scripts.
- Upload test artifacts (logfile and verification html).
- Adjust r3b demo test reference.
- The eighth run of the fhh demo converges (since 703607e for NEWMSH)
- Fix two more results for ab demo in test reference.
- Setup automated testing (Continuous Integration) using Github
- Manual
- Fix documentation illustrating Python data access for lrz demo. Closes #3.
- Latex compile fix (Andreas Amann)
the unknown m placement specifier for figures was ignored by older
latex versions but is now an error.
- Demos
- Make tfc demo more robust.
- Tweak EPS* in fhh demo to properly detect UZ in eighth run
- Adjust EPSS in fhh demo to still detect the test point.
- Increase NTST in ab demo to make it more robust in the last run.
- Increase NTST in c.ab.3 to 70 so it does not MX unreliably.
- c2c demo: increase NTST a little to make it work correctly with GCC8
- Increase NTST in c.fhn.5 to match c.fhn.6 (elimates spurious UZ)
Forever And Always 0.9 Demo Mac Os Downloads
Forever And Always 0.9 DEMO Mac OS