[Need-Attention-to]About func "interp_pv_upwind" in file "/src/dynamics/operators_mod.F90"

[ShoiLyaung]

Funcion name: subroutine interp_pv_upwind(block, dstate, dt, substep)

This function costs a lot in swm_xxx.

Mainly because these loops (Line 910-967):

    call perf_start('interp_pv_upwind')

    associate (mesh   => block%mesh      , &
               un     => dstate%u_lon    , & ! in
               vn     => dstate%v_lat    , & ! in
               ut     => block%aux%u_lat , & ! in
               vt     => block%aux%v_lon , & ! in
               pv     => block%aux%pv    , & ! in
               pv_lon => block%aux%pv_lon, & ! out
               pv_lat => block%aux%pv_lat)   ! out
    select case (upwind_order_pv)
    case (1)
      do k = mesh%full_kds, mesh%full_kde
        do j = mesh%full_jds_no_pole, mesh%full_jde_no_pole
          do i = mesh%half_ids, mesh%half_ide
            b = abs(vt%d(i,j,k)) / (sqrt(un%d(i,j,k)**2 + vt%d(i,j,k)**2) + eps)
            pv_lon%d(i,j,k) = b * upwind1(sign(1.0_r8, vt%d(i,j,k)), upwind_wgt_pv, pv%d(i,j-1:j,k)) + &
                              (1 - b) * 0.5_r8 * (pv%d(i,j-1,k) + pv%d(i,j,k))
          end do
        end do
        do j = mesh%half_jds, mesh%half_jde
          do i = mesh%full_ids, mesh%full_ide
            b = abs(ut%d(i,j,k)) / (sqrt(ut%d(i,j,k)**2 + vn%d(i,j,k)**2) + eps)
            pv_lat%d(i,j,k) = b * upwind1(sign(1.0_r8, ut%d(i,j,k)), upwind_wgt_pv, pv%d(i-1:i,j,k)) + &
                              (1 - b) * 0.5_r8 * (pv%d(i-1,j,k) + pv%d(i,j,k))
          end do
        end do
      end do
    case (3)
      do k = mesh%full_kds, mesh%full_kde
        do j = mesh%full_jds_no_pole, mesh%full_jde_no_pole
          do i = mesh%half_ids, mesh%half_ide
            b = abs(vt%d(i,j,k)) / (sqrt(un%d(i,j,k)**2 + vt%d(i,j,k)**2) + eps)
            pv_lon%d(i,j,k) = b * upwind3(sign(1.0_r8, vt%d(i,j,k)), upwind_wgt_pv, pv%d(i,j-2:j+1,k)) + &
                              (1 - b) * 0.5_r8 * (pv%d(i,j-1,k) + pv%d(i,j,k))
          end do
        end do
        do j = mesh%half_jds, mesh%half_jde
          do i = mesh%full_ids, mesh%full_ide
            b  = abs(ut%d(i,j,k)) / (sqrt(ut%d(i,j,k)**2 + vn%d(i,j,k)**2) + eps)
            pv_lat%d(i,j,k) = b * upwind3(sign(1.0_r8, ut%d(i,j,k)), upwind_wgt_pv, pv%d(i-2:i+1,j,k)) + &
                              (1 - b) * 0.5_r8 * (pv%d(i-1,j,k) + pv%d(i,j,k))
          end do
        end do
      end do
    case (5)
      do k = mesh%full_kds, mesh%full_kde
        do j = mesh%full_jds_no_pole, mesh%full_jde_no_pole
          do i = mesh%half_ids, mesh%half_ide
            b = abs(vt%d(i,j,k)) / (sqrt(un%d(i,j,k)**2 + vt%d(i,j,k)**2) + eps)
            pv_lon%d(i,j,k) = b * upwind5(sign(1.0_r8, vt%d(i,j,k)), upwind_wgt_pv, pv%d(i,j-3:j+2,k)) + &
                              (1 - b) * 0.5_r8 * (pv%d(i,j-1,k) + pv%d(i,j,k))
          end do
        end do
        do j = mesh%half_jds, mesh%half_jde
          do i = mesh%full_ids, mesh%full_ide
            b = abs(ut%d(i,j,k)) / (sqrt(ut%d(i,j,k)**2 + vn%d(i,j,k)**2) + eps)
            pv_lat%d(i,j,k) = b * upwind5(sign(1.0_r8, ut%d(i,j,k)), upwind_wgt_pv, pv%d(i-3:i+2,j,k)) + &
                              (1 - b) * 0.5_r8 * (pv%d(i-1,j,k) + pv%d(i,j,k))
          end do
        end do
      end do
    end select
    call fill_halo(pv_lon, east_halo=.false., south_halo=.false.)
    call fill_halo(pv_lat, west_halo=.false., north_halo=.false.)
    end associate

    call perf_stop('interp_pv_upwind')