some workflows may benefit from wider integer or floating point types. these should be configurable through options passed to cmake, which should correctly force-disable things currently incompatible with certain types (such as some of the hip and probably cuda code). additionally, they should set the corresponding type aliases (real_type/index_type) in resolve/Common.hpp and resolve/lusol/lusol_precision.f90
additional notes
the hip/cuda code could be modified to be compatible with wider integer types (in the event native support is not possible) by partitioning the job, probably
an instance in which another similar project has done this sort of thing can be found here^1
overview
some workflows may benefit from wider integer or floating point types. these should be configurable through options passed to cmake, which should correctly force-disable things currently incompatible with certain types (such as some of the hip and probably cuda code). additionally, they should set the corresponding type aliases (
real_type
/index_type
) inresolve/Common.hpp
andresolve/lusol/lusol_precision.f90
additional notes