rhs_maker_cpu.f90

Go to the documentation of this file.

module rhs_maker_cpu
  use rhs_maker, only : rhs_maker_bdf_t, rhs_maker_ext_t, rhs_maker_sumab_t, &
       rhs_maker_oifs_t
  use field_series, only : field_series_t
  use field, only : field_t
  use num_types, only : rp
  use scratch_registry, only : neko_scratch_registry
  implicit none
  private
 
  type, public, extends(rhs_maker_sumab_t) :: rhs_maker_sumab_cpu_t
   contains
     procedure, nopass :: compute_fluid => rhs_maker_sumab_cpu
  end type rhs_maker_sumab_cpu_t
 
  type, public, extends(rhs_maker_ext_t) :: rhs_maker_ext_cpu_t
   contains
     procedure, nopass :: compute_fluid => rhs_maker_ext_cpu
     procedure, nopass :: compute_scalar => scalar_rhs_maker_ext_cpu
  end type rhs_maker_ext_cpu_t
 
  type, public, extends(rhs_maker_bdf_t) :: rhs_maker_bdf_cpu_t
   contains
     procedure, nopass :: compute_fluid => rhs_maker_bdf_cpu
     procedure, nopass :: compute_scalar => scalar_rhs_maker_bdf_cpu
  end type rhs_maker_bdf_cpu_t
 
  type, public, extends(rhs_maker_oifs_t) :: rhs_maker_oifs_cpu_t
   contains
     procedure, nopass :: compute_fluid => rhs_maker_oifs_cpu
     procedure, nopass :: compute_scalar => scalar_rhs_maker_oifs_cpu
  end type rhs_maker_oifs_cpu_t
 
contains
 
  subroutine rhs_maker_sumab_cpu(u, v, w, uu, vv, ww, uulag, vvlag, wwlag, &
       ab, nab)
    type(field_t), intent(inout) :: u,v, w
    type(field_t), intent(inout) :: uu, vv, ww
    type(field_series_t), intent(inout) :: uulag, vvlag, wwlag
    real(kind=rp), dimension(3), intent(in) :: ab
    integer, intent(in) :: nab
    integer :: i, n
 
    n = uu%dof%size()
    !$omp parallel private(i)
    if (nab .eq. 3) then
       !OCL NORECURRENCE, NOVREC, NOALIAS
       !DIR$ CONCURRENT
       !DIR$ IVDEP
       !GCC$ ivdep
       !$omp do
       do i = 1, n
          u%x(i,1,1,1) = ab(1) * uu%x(i,1,1,1) + &
               ab(2) * uulag%lf(1)%x(i,1,1,1) + ab(3) * uulag%lf(2)%x(i,1,1,1)
          v%x(i,1,1,1) = ab(1) * vv%x(i,1,1,1) + &
               ab(2) * vvlag%lf(1)%x(i,1,1,1) + ab(3) * vvlag%lf(2)%x(i,1,1,1)
          w%x(i,1,1,1) = ab(1) * ww%x(i,1,1,1) + &
               ab(2) * wwlag%lf(1)%x(i,1,1,1) + ab(3) * wwlag%lf(2)%x(i,1,1,1)
       end do
       !$omp end do
    else
       !OCL NORECURRENCE, NOVREC, NOALIAS
       !DIR$ CONCURRENT
       !DIR$ IVDEP
       !GCC$ ivdep
       !$omp do
       do i = 1, n
          u%x(i,1,1,1) = ab(1) * uu%x(i,1,1,1) + ab(2) * uulag%lf(1)%x(i,1,1,1)
          v%x(i,1,1,1) = ab(1) * vv%x(i,1,1,1) + ab(2) * vvlag%lf(1)%x(i,1,1,1)
          w%x(i,1,1,1) = ab(1) * ww%x(i,1,1,1) + ab(2) * wwlag%lf(1)%x(i,1,1,1)
       end do
       !$omp end do
    end if
    !$omp end parallel
 
  end subroutine rhs_maker_sumab_cpu
 
  subroutine rhs_maker_ext_cpu(fx_lag, fy_lag, fz_lag, &
       fx_laglag, fy_laglag, fz_laglag, fx, fy, fz, &
       rho, ext_coeffs, n)
    type(field_t), intent(inout) :: fx_lag, fy_lag, fz_lag
    type(field_t), intent(inout) :: fx_laglag, fy_laglag, fz_laglag
    real(kind=rp), intent(in) :: rho, ext_coeffs(4)
    integer, intent(in) :: n
    real(kind=rp), intent(inout) :: fx(n), fy(n), fz(n)
    integer :: i
    type(field_t), pointer :: temp1, temp2, temp3
    integer :: temp_indices(3)
 
    call neko_scratch_registry%request_field(temp1, temp_indices(1), .false.)
    call neko_scratch_registry%request_field(temp2, temp_indices(2), .false.)
    call neko_scratch_registry%request_field(temp3, temp_indices(3), .false.)
 
    !$omp parallel private(i)
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       temp1%x(i,1,1,1) = ext_coeffs(2) * fx_lag%x(i,1,1,1) + &
            ext_coeffs(3) * fx_laglag%x(i,1,1,1)
       temp2%x(i,1,1,1) = ext_coeffs(2) * fy_lag%x(i,1,1,1) + &
            ext_coeffs(3) * fy_laglag%x(i,1,1,1)
       temp3%x(i,1,1,1) = ext_coeffs(2) * fz_lag%x(i,1,1,1) + &
            ext_coeffs(3) * fz_laglag%x(i,1,1,1)
    end do
    !$omp end do
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       fx_laglag%x(i,1,1,1) = fx_lag%x(i,1,1,1)
       fy_laglag%x(i,1,1,1) = fy_lag%x(i,1,1,1)
       fz_laglag%x(i,1,1,1) = fz_lag%x(i,1,1,1)
       fx_lag%x(i,1,1,1) = fx(i)
       fy_lag%x(i,1,1,1) = fy(i)
       fz_lag%x(i,1,1,1) = fz(i)
    end do
    !$omp end do
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       fx(i) = (ext_coeffs(1) * fx(i) + temp1%x(i,1,1,1)) * rho
       fy(i) = (ext_coeffs(1) * fy(i) + temp2%x(i,1,1,1)) * rho
       fz(i) = (ext_coeffs(1) * fz(i) + temp3%x(i,1,1,1)) * rho
    end do
    !$omp end do
    !$omp end parallel
    call neko_scratch_registry%relinquish_field(temp_indices)
 
  end subroutine rhs_maker_ext_cpu
 
  subroutine scalar_rhs_maker_ext_cpu(fs_lag, fs_laglag, fs, ext_coeffs, n)
    type(field_t), intent(inout) :: fs_lag
    type(field_t), intent(inout) :: fs_laglag
    real(kind=rp), intent(in) :: ext_coeffs(4)
    integer, intent(in) :: n
    real(kind=rp), intent(inout) :: fs(n)
    integer :: i
    type(field_t), pointer :: temp1
    integer :: temp_index
 
    call neko_scratch_registry%request_field(temp1, temp_index, .false.)
 
    !$omp parallel private(i)
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       temp1%x(i,1,1,1) = ext_coeffs(2) * fs_lag%x(i,1,1,1) + &
            ext_coeffs(3) * fs_laglag%x(i,1,1,1)
    end do
    !$omp end do
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       fs_laglag%x(i,1,1,1) = fs_lag%x(i,1,1,1)
       fs_lag%x(i,1,1,1) = fs(i)
    end do
    !$omp end do
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       fs(i) = ext_coeffs(1) * fs(i) + temp1%x(i,1,1,1)
    end do
    !$omp end do
    !$omp end parallel
 
    call neko_scratch_registry%relinquish_field(temp_index)
  end subroutine scalar_rhs_maker_ext_cpu
 
  subroutine rhs_maker_bdf_cpu(ulag, vlag, wlag, bfx, bfy, bfz, &
       u, v, w, B, Blag, Blaglag, rho, dt, bd, nbd, n)
    integer, intent(in) :: n, nbd
    type(field_t), intent(in) :: u, v, w
    type(field_series_t), intent(in) :: ulag, vlag, wlag
    real(kind=rp), intent(inout) :: bfx(n), bfy(n), bfz(n)
    real(kind=rp), intent(in) :: b(n), blag(n), blaglag(n)
    real(kind=rp), intent(in) :: dt, rho, bd(4)
    type(field_t), pointer :: tb1, tb2, tb3
    integer :: temp_indices(3)
    integer :: i, ilag
 
    call neko_scratch_registry%request_field(tb1, temp_indices(1), .false.)
    call neko_scratch_registry%request_field(tb2, temp_indices(2), .false.)
    call neko_scratch_registry%request_field(tb3, temp_indices(3), .false.)
 
    !$omp parallel private(ilag, i)
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       tb1%x(i,1,1,1) = u%x(i,1,1,1) * b(i) * bd(2)
       tb2%x(i,1,1,1) = v%x(i,1,1,1) * b(i) * bd(2)
       tb3%x(i,1,1,1) = w%x(i,1,1,1) * b(i) * bd(2)
    end do
    !$omp end do
 
    do ilag = 2, nbd
       if (ilag .eq. 2) then
          !OCL NORECURRENCE, NOVREC, NOALIAS
          !DIR$ CONCURRENT
          !DIR$ IVDEP
          !GCC$ ivdep
          !$omp do
          do i = 1, n
             tb1%x(i,1,1,1) = tb1%x(i,1,1,1) + &
                  (ulag%lf(ilag-1)%x(i,1,1,1) * blag(i) * bd(ilag+1))
             tb2%x(i,1,1,1) = tb2%x(i,1,1,1) + &
                  (vlag%lf(ilag-1)%x(i,1,1,1) * blag(i) * bd(ilag+1))
             tb3%x(i,1,1,1) = tb3%x(i,1,1,1) + &
                  (wlag%lf(ilag-1)%x(i,1,1,1) * blag(i) * bd(ilag+1))
          end do
          !$omp end do
       else if (ilag .eq. 3) then
          !OCL NORECURRENCE, NOVREC, NOALIAS
          !DIR$ CONCURRENT
          !DIR$ IVDEP
          !GCC$ ivdep
          !$omp do
          do i = 1, n
             tb1%x(i,1,1,1) = tb1%x(i,1,1,1) + &
                  (ulag%lf(ilag-1)%x(i,1,1,1) * blaglag(i) * bd(ilag+1))
             tb2%x(i,1,1,1) = tb2%x(i,1,1,1) + &
                  (vlag%lf(ilag-1)%x(i,1,1,1) * blaglag(i) * bd(ilag+1))
             tb3%x(i,1,1,1) = tb3%x(i,1,1,1) + &
                  (wlag%lf(ilag-1)%x(i,1,1,1) * blaglag(i) * bd(ilag+1))
          end do
          !$omp end do
       end if
    end do
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       bfx(i) = bfx(i) + tb1%x(i,1,1,1) * (rho / dt)
       bfy(i) = bfy(i) + tb2%x(i,1,1,1) * (rho / dt)
       bfz(i) = bfz(i) + tb3%x(i,1,1,1) * (rho / dt)
    end do
    !$omp end do
    !$omp end parallel
 
    call neko_scratch_registry%relinquish_field(temp_indices)
 
  end subroutine rhs_maker_bdf_cpu
 
  subroutine scalar_rhs_maker_bdf_cpu(s_lag, fs, s, B, rho_cp, dt, bd, nbd, n)
    integer, intent(in) :: n, nbd
    type(field_t), intent(in) :: s, rho_cp
    type(field_series_t), intent(in) :: s_lag
    real(kind=rp), intent(inout) :: fs(n)
    real(kind=rp), intent(in) :: b(n)
    real(kind=rp), intent(in) :: dt, bd(4)
    integer :: i, ilag
    type(field_t), pointer :: temp1
    integer :: temp_indices
 
    call neko_scratch_registry%request_field(temp1, temp_indices, .false.)
 
    !$omp parallel private(i, ilag)
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       temp1%x(i,1,1,1) = s%x(i,1,1,1) * b(i) * bd(2)
    end do
    !$omp end do
 
    do ilag = 2, nbd
       !OCL NORECURRENCE, NOVREC, NOALIAS
       !DIR$ CONCURRENT
       !DIR$ IVDEP
       !GCC$ ivdep
       !$omp do
       do i = 1, n
          temp1%x(i,1,1,1) = temp1%x(i,1,1,1) + &
               (s_lag%lf(ilag-1)%x(i,1,1,1) * b(i) * bd(ilag+1))
       end do
       !$omp end do
    end do
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp do
    do i = 1, n
       fs(i) = fs(i) + temp1%x(i,1,1,1) * &
            (rho_cp%x(i,1,1,1) / dt)
    end do
    !$omp end do
    !$omp end parallel
 
    call neko_scratch_registry%relinquish_field(temp_indices)
  end subroutine scalar_rhs_maker_bdf_cpu
 
  subroutine rhs_maker_oifs_cpu(phi_x, phi_y, phi_z, bf_x, bf_y, bf_z, &
       rho, dt, n)
    real(kind=rp), intent(in) :: rho, dt
    integer, intent(in) :: n
    real(kind=rp), intent(inout) :: bf_x(n), bf_y(n), bf_z(n)
    real(kind=rp), intent(inout) :: phi_x(n), phi_y(n), phi_z(n)
    integer :: i
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp parallel do
    do i = 1, n
       bf_x(i) = bf_x(i) + phi_x(i) * (rho / dt)
       bf_y(i) = bf_y(i) + phi_y(i) * (rho / dt)
       bf_z(i) = bf_z(i) + phi_z(i) * (rho / dt)
    end do
    !$omp end parallel do
 
  end subroutine rhs_maker_oifs_cpu
 
  subroutine scalar_rhs_maker_oifs_cpu(phi_s, bf_s, rho_cp, dt, n)
    type(field_t), intent(in) :: rho_cp
    real(kind=rp), intent(in) :: dt
    integer, intent(in) :: n
    real(kind=rp), intent(inout) :: bf_s(n)
    real(kind=rp), intent(inout) :: phi_s(n)
    integer :: i
 
    !OCL NORECURRENCE, NOVREC, NOALIAS
    !DIR$ CONCURRENT
    !DIR$ IVDEP
    !GCC$ ivdep
    !$omp parallel do
    do i = 1, n
       bf_s(i) = bf_s(i) + phi_s(i) * (rho_cp%x(i,1,1,1) / dt)
    end do
    !$omp end parallel do
 
  end subroutine scalar_rhs_maker_oifs_cpu
 
end module rhs_maker_cpu