Skip to content

Commit 8e5faaa

Browse files
wilfonbawilfonba
committed
precheck and remove debug prints
1 parent f5bd013 commit 8e5faaa

3 files changed

Lines changed: 7 additions & 120 deletions

File tree

‎src/simulation/m_bubbles_EL.fpp‎

Lines changed: 2 additions & 115 deletions
Original file line numberDiff line numberDiff line change
@@ -619,7 +619,7 @@ contains
619619
!! @param stage Current stage in the time-stepper algorithm
620620
subroutine s_compute_bubble_EL_dynamics(q_prim_vf, bc_type, stage)
621621
#ifdef MFC_OpenMP
622-
!DIR$ OPTIMIZE (-O1)
622+
!DIR$ OPTIMIZE (-O1)
623623
#endif
624624
type(scalar_field), dimension(sys_size), intent(inout) :: q_prim_vf
625625
type(integer_field), dimension(1:num_dims, 1:2), intent(in) :: bc_type
@@ -829,7 +829,7 @@ contains
829829
do l = 1, E_idx
830830
if (q_beta(1)%sf(i, j, k) > (1._wp - lag_params%valmaxvoid)) then
831831
rhs_vf(l)%sf(i, j, k) = rhs_vf(l)%sf(i, j, k) + &
832-
(q_cons_vf(l)%sf(i, j, k)/q_beta(1)%sf(i, j, k)) * &
832+
(q_cons_vf(l)%sf(i, j, k)/q_beta(1)%sf(i, j, k))* &
833833
q_beta(2)%sf(i, j, k)
834834
end if
835835
end do
@@ -932,37 +932,6 @@ contains
932932

933933
type(integer_field), dimension(1:num_dims, 1:2), intent(in) :: bc_type
934934
integer :: i, j, k, l
935-
integer, save :: smear_call_count = 0
936-
937-
smear_call_count = smear_call_count + 1
938-
939-
! DEBUG: bubble state checksum before smearing
940-
$:GPU_UPDATE(host='[intfc_rad, intfc_vel, mtn_pos, n_el_bubs_loc]')
941-
block
942-
real(wp) :: rad_loc, rad_glb, vel_loc, vel_glb
943-
real(wp) :: posx_loc, posx_glb, posy_loc, posy_glb, posz_loc, posz_glb
944-
integer :: kk, ierr2, nbubs_glb_chk
945-
rad_loc = 0._wp; vel_loc = 0._wp
946-
posx_loc = 0._wp; posy_loc = 0._wp; posz_loc = 0._wp
947-
do kk = 1, n_el_bubs_loc
948-
rad_loc = rad_loc + intfc_rad(kk, 2)
949-
vel_loc = vel_loc + intfc_vel(kk, 2)
950-
posx_loc = posx_loc + mtn_pos(kk, 1, 2)
951-
posy_loc = posy_loc + mtn_pos(kk, 2, 2)
952-
posz_loc = posz_loc + mtn_pos(kk, 3, 2)
953-
end do
954-
call MPI_Allreduce(rad_loc, rad_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
955-
call MPI_Allreduce(vel_loc, vel_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
956-
call MPI_Allreduce(posx_loc, posx_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
957-
call MPI_Allreduce(posy_loc, posy_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
958-
call MPI_Allreduce(posz_loc, posz_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
959-
call MPI_Allreduce(n_el_bubs_loc, nbubs_glb_chk, 1, MPI_INTEGER, MPI_SUM, MPI_COMM_WORLD, ierr2)
960-
if (proc_rank == 0) print *, "DEBUG BUBSTATE call=", smear_call_count, &
961-
" n_bubs=", nbubs_glb_chk, &
962-
" sum_R=", rad_glb, " sum_Rdot=", vel_glb, &
963-
" sum_x=", posx_glb, " sum_y=", posy_glb, " sum_z=", posz_glb
964-
end block
965-
$:GPU_UPDATE(device='[intfc_rad, intfc_vel, mtn_pos, n_el_bubs_loc]')
966935

967936
call nvtxStartRange("BUBBLES-LAGRANGE-SMEARING")
968937
$:GPU_PARALLEL_LOOP(private='[i,j,k,l]', collapse=4)
@@ -984,46 +953,6 @@ contains
984953
call s_smoothfunction(n_el_bubs_loc, intfc_rad, intfc_vel, &
985954
mtn_s, mtn_pos, q_beta, kahan_comp)
986955

987-
! DEBUG: checksum after Gaussian smearing (before communication)
988-
! all_cells = sum over entire grid (interior + buffer) per rank, then MPI_SUM.
989-
! This is the total Gaussian integral and MUST be decomposition-invariant.
990-
! If it differs between 1-rank and multi-rank, the smearing kernel itself is wrong.
991-
! interior = sum over interior cells only (0:m, 0:n, 0:p).
992-
$:GPU_UPDATE(host='[q_beta(1)%sf]')
993-
block
994-
real(wp) :: all_loc, all_glb
995-
real(wp) :: int_loc, int_glb
996-
real(wp) :: max_loc, max_glb
997-
integer :: ierr2
998-
999-
all_loc = 0._wp
1000-
int_loc = 0._wp
1001-
max_loc = 0._wp
1002-
1003-
do l = idwbuff(3)%beg, idwbuff(3)%end
1004-
do k = idwbuff(2)%beg, idwbuff(2)%end
1005-
do j = idwbuff(1)%beg, idwbuff(1)%end
1006-
all_loc = all_loc + real(q_beta(1)%sf(j, k, l), kind=wp)
1007-
if (j >= 0 .and. j <= m .and. k >= 0 .and. k <= n .and. &
1008-
l >= 0 .and. l <= p) then
1009-
int_loc = int_loc + real(q_beta(1)%sf(j, k, l), kind=wp)
1010-
max_loc = max(max_loc, abs(real(q_beta(1)%sf(j, k, l), kind=wp)))
1011-
end if
1012-
end do
1013-
end do
1014-
end do
1015-
1016-
call MPI_Allreduce(all_loc, all_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
1017-
call MPI_Allreduce(int_loc, int_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
1018-
call MPI_Allreduce(max_loc, max_glb, 1, MPI_DOUBLE_PRECISION, MPI_MAX, MPI_COMM_WORLD, ierr2)
1019-
1020-
if (proc_rank == 0) print *, "DEBUG PRE_COMM call=", smear_call_count, &
1021-
" all_cells=", all_glb, " interior=", int_glb, &
1022-
" buf=", all_glb - int_glb, " max=", max_glb, &
1023-
" n_bubs=", n_el_bubs_glb
1024-
end block
1025-
$:GPU_UPDATE(device='[q_beta(1)%sf]')
1026-
1027956
call nvtxStartRange("BUBBLES-LAGRANGE-BETA-COMM")
1028957
if (lag_params%cluster_type >= 4) then
1029958
call s_populate_beta_buffers(q_beta, bc_type, 3, kahan_comp)
@@ -1032,48 +961,6 @@ contains
1032961
end if
1033962
call nvtxEndRange
1034963

1035-
! DEBUG: checksum after communication (before 1-beta conversion)
1036-
! interior = sum over interior cells — this is what the source term uses.
1037-
! weighted = position-weighted checksum to detect spatial redistribution errors.
1038-
$:GPU_UPDATE(host='[q_beta(1)%sf, q_beta(2)%sf]')
1039-
block
1040-
real(wp) :: int_loc, int_glb
1041-
real(wp) :: weighted_loc, weighted_glb
1042-
real(wp) :: max_loc, max_glb
1043-
real(wp) :: dbeta_loc, dbeta_glb
1044-
integer :: gj, gk, gl, ierr2
1045-
1046-
int_loc = 0._wp
1047-
weighted_loc = 0._wp
1048-
max_loc = 0._wp
1049-
dbeta_loc = 0._wp
1050-
1051-
do l = 0, p
1052-
do k = 0, n
1053-
do j = 0, m
1054-
int_loc = int_loc + real(q_beta(1)%sf(j, k, l), kind=wp)
1055-
max_loc = max(max_loc, abs(real(q_beta(1)%sf(j, k, l), kind=wp)))
1056-
dbeta_loc = dbeta_loc + real(q_beta(2)%sf(j, k, l), kind=wp)
1057-
gj = j + start_idx(1)
1058-
gk = k + merge(start_idx(2), 0, num_dims >= 2)
1059-
gl = l + merge(start_idx(num_dims), 0, num_dims >= 3)
1060-
weighted_loc = weighted_loc + &
1061-
real(q_beta(1)%sf(j, k, l), kind=wp) * real(gj + gk*1000 + gl*1000000, kind=wp)
1062-
end do
1063-
end do
1064-
end do
1065-
1066-
call MPI_Allreduce(int_loc, int_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
1067-
call MPI_Allreduce(max_loc, max_glb, 1, MPI_DOUBLE_PRECISION, MPI_MAX, MPI_COMM_WORLD, ierr2)
1068-
call MPI_Allreduce(weighted_loc, weighted_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
1069-
call MPI_Allreduce(dbeta_loc, dbeta_glb, 1, MPI_DOUBLE_PRECISION, MPI_SUM, MPI_COMM_WORLD, ierr2)
1070-
1071-
if (proc_rank == 0) print *, "DEBUG POST_COMM call=", smear_call_count, &
1072-
" interior=", int_glb, " max=", max_glb, " weighted=", weighted_glb, &
1073-
" dbeta_dt=", dbeta_glb
1074-
end block
1075-
$:GPU_UPDATE(device='[q_beta(1)%sf, q_beta(2)%sf]')
1076-
1077964
!Store 1-beta
1078965
$:GPU_PARALLEL_LOOP(private='[j,k,l]', collapse=3)
1079966
do l = idwbuff(3)%beg, idwbuff(3)%end

‎src/simulation/m_bubbles_EL_kernels.fpp‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -147,7 +147,7 @@ contains
147147
! Loop over bubbles in this cell
148148
$:GPU_LOOP(parallelism='[seq]')
149149
do lb = cell_list_start(i, j, k), &
150-
cell_list_start(i, j, k) + cell_list_count(i, j, k) - 1
150+
cell_list_start(i, j, k) + cell_list_count(i, j, k) - 1
151151

152152
bub_idx = cell_list_idx(lb)
153153

@@ -246,7 +246,7 @@ contains
246246
do di = di_beg, di_end
247247
$:GPU_LOOP(parallelism='[seq]')
248248
do lb = cell_list_start(di, dj, dk), &
249-
cell_list_start(di, dj, dk) + cell_list_count(di, dj, dk) - 1
249+
cell_list_start(di, dj, dk) + cell_list_count(di, dj, dk) - 1
250250

251251
bub_idx = cell_list_idx(lb)
252252

@@ -344,11 +344,11 @@ contains
344344
else
345345
!< 2D cartesian function: Equation (48) from Madea and Colonius 2018
346346
! We smear particles considering a virtual depth (lag_params%charwidth) with lag_params%charNz cells
347-
dzp = (lag_params%charwidth / (lag_params%charNz + 1._wp))
347+
dzp = (lag_params%charwidth/(lag_params%charNz + 1._wp))
348348

349349
func = 0._wp
350350
do i = 0, lag_params%charNz
351-
zc = (-lag_params%charwidth/2._wp + dzp * (0.5_wp + i)) ! Center of virtual cell i in z-direction
351+
zc = (-lag_params%charwidth/2._wp + dzp*(0.5_wp + i)) ! Center of virtual cell i in z-direction
352352
Lz2 = (center(3) - zc)**2._wp
353353
distance = sqrt((center(1) - nodecoord(1))**2._wp + (center(2) - nodecoord(2))**2._wp + Lz2)
354354
func = func + dzp/lag_params%charwidth*exp(-0.5_wp*(distance/stddsv)**2._wp)/(sqrt(2._wp*pi)*stddsv)**3._wp

‎src/simulation/m_start_up.fpp‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1203,7 +1203,7 @@ contains
12031203
! carried out if the simulation is in fact not truly executed in parallel.
12041204

12051205
call s_mpi_bcast_user_inputs()
1206-
print*, proc_rank, lag_params%charwidth, lag_params%charNz
1206+
print *, proc_rank, lag_params%charwidth, lag_params%charNz
12071207
call s_initialize_parallel_io()
12081208

12091209
call s_mpi_decompose_computational_domain()

0 commit comments

Comments
 (0)