m_precision_select.f90

Go to the documentation of this file.

!> @file m_precision_select.f90
!! @brief Contains module m_precision_select
 
!> @brief Working-precision kind selection (half/single/double) and corresponding MPI datatype parameters
module m_precision_select
 
    ! use, intrinsic :: iso_c_binding
 
#ifdef MFC_MPI
    use mpi  !< message passing interface (mpi) module
#endif
 
    implicit none
 
    ! Define the available precision types
    integer, parameter :: half_precision = 2  !< selected_real_kind(3, 4)
    integer, parameter :: single_precision = selected_real_kind(6, 37)
    integer, parameter :: double_precision = selected_real_kind(15, 307)
    integer, parameter :: hp = half_precision
    integer, parameter :: sp = single_precision
    integer, parameter :: dp = double_precision
 
    ! Set the working precision (wp) to single or double
#ifdef MFC_SINGLE_PRECISION
    integer, parameter :: wp = single_precision  !< Change to single_precision if needed
#else
    integer, parameter :: wp = double_precision
#endif
 
    ! Set the storage precision (stp) to half if mixed precision is requested
#ifdef MFC_MIXED_PRECISION
    integer, parameter :: stp = half_precision
#else
    integer, parameter :: stp = wp
#endif
 
#ifdef MFC_MPI
    ! Set mpi_p based on wp using the merge intrinsic function
    integer, parameter :: mpi_p = merge(mpi_double_precision, mpi_real, wp == double_precision)
    integer, parameter :: mpi_2p = merge(mpi_2double_precision, mpi_2real, wp == double_precision)
    integer, parameter :: mpi_io_p = merge(mpi_byte, mpi_p, stp == half_precision)
    ! MPI types per element. IE Real(kind=2) <=> 2 MPI_BYTE
    integer, parameter :: mpi_io_type = merge(2, 1, stp == half_precision)
#else
    integer, parameter :: mpi_p = -100  !< Default value when MPI is not used
    integer, parameter :: mpi_2p = -100
    integer, parameter :: mpi_io_p = -100
    integer, parameter :: mpi_io_type = -100
#endif
end module m_precision_select