QUESO for Bayesian Calibration
Compiling Dakota's QUESO sub-package requires:
- A Unix-like platform (not supported on Windows)
- Enabling MPI (see above) Note: For Dakota versions 6.3 and later, MPI is no longer required to enable QUESO.
- The Boost program_options compiled library
- GSL 1.15 or newer, which means that resulting Dakota binaries will include GPL licensed components
If the program gsl-config is not on your PATH, you'll either need to put it there, or set the environment variable export GSL_CONFIG=/usr/local/gsl/1.15/bin/gsl-config or similar. You may also need to set export BOOST_ROOT=/usr/local/boost/1.49.0 or similar to make sure that QUESO and Dakota are built with the same Boost.
The following CMake settings are necessary to enable QUESO:
HAVE_QUESO:BOOL=ON DAKOTA_HAVE_GSL:BOOL=ON # Optionally, to work around a Dakota 6.1 bug: HAVE_QUESO_GPMSA:BOOL=OFF # Optionally, specify a CBLAS to avoid numerical differences induced by using GSL's # For example, on RHEL6: -DDAKOTA_CBLAS_LIBS:FILEPATH=/usr/lib64/atlas/libcblas.so.3
HDF5 Results Output
Building Dakota with HDF5 support requires HDF5 version 1.10.2 or higher. Parallel HDF5 is not needed at this time, but will be in future Dakota releases. In addition, the Python h5py module is needed to run some of the HDF5-related tests and for the HDF5 examples.
The following CMake setting is necessary to enable HDF5:
In addition, the environment variable HDF5_ROOT should be set to the top-level HDF5 directory (i.e. that contains the include and lib folders).
Building Dakota documentation build requires appropriate versions of Perl, Doxygen, and LaTeX and only works on Unix platforms: set ENABLE_DAKOTA_DOCS:BOOL=TRUE
Dakota X11/Motif-based graphics default to ON, but are turned off if necessary libraries aren't found. This feature can be force disabled by setting HAVE_X_GRAPHICS:BOOL=FALSE
This table lists the most common CMake options for enabling or disabling specific Dakota sub-packages.
|HAVE_AMPL||ON||May need to be OFF if compiling with C99 support|
|HAVE_DOT||ON||Turns OFF if packages/DOT missing|
|HAVE_FFTW||OFF||Off due to dependence on GPL-licensed FFTW|
|HAVE_NLPQL||ON||Turns OFF if packages/NLPQL missing|
|HAVE_NPSOL||ON||Turns OFF if packages/NPSOL missing|
|HAVE_QUESO||OFF||Off due to dependence on GPL-licensed GSL|
|DAKOTA_HAVE_GSL||OFF||Required when HAVE_QUESO=ON|
The following options affect tests
|DAKOTA_ENABLE_TESTS||ON||Whether to enable Dakota system-level tests|
|DAKOTA_EXPAND_SUBTESTS||OFF||Whether to register on CTest test per system-level subtest|
|DAKOTA_ENABLE_TPL_TESTS||OFF||Whether to enable the following TPL tests|
|ENABLE_TESTS||OFF||Whether to enable tests in Acro and Utilib|
Dakota specification maintenance mode is used by developers when modifying Dakota XML input specification. It generates nidrgen and Dt binaries in packages/nidr and when dakota.input.nspec changes, will use them to update relevant generated files in DAKOTA_SOURCE/src since currently they need to be checked in. To enable specification maintenance, use the option ENABLE_SPEC_MAINT:BOOL=TRUE
Experimental (Morse-Smale) Adaptive Sampling
(Experimental capability not recommended for general users.) To enable basic adaptive sampling, set HAVE_ADAPTIVE_SAMPLING:BOOL=ON. To enable topologically-based metrics via the Approximate Nearest Neighbor and Dionysus libraries, install ANN and Dionysus (see prerequisites) and set HAVE_MORSE_SMALE:BOOL=ON.