Tpetra DefaultTypes::scalar_type is always double; should it allow float?
Created by: kddevin
It is permissible to build Tpetra with Tpetra_INST_DOUBLE=OFF and Tpetra_INST_FLOAT=ON, correct?
Currently namespace DefaultTypes says the default scalar_type is double always. If building with FLOAT but not DOUBLE is allowed, then in Tpetra_Details_DefaultTypes.hpp, we should have something like
#if defined(HAVE_TPETRA_INST_DOUBLE) typedef double scalar_type #elif defined(HAVE_TPETRA_INST_FLOAT) typedef float scalar_type #else #error "too bad, so sad" #endif
Similar logic already exists in that file for the various ordinal types.
If, instead, we always want the default scalar type to be double, we should remove the Tpetra_INST_DOUBLE CMake option, so that users do not get compilation errors by disabling DOUBLE.