Debug ibm stability

docs/documentation/case.md

@@ -441,13 +441,14 @@ A particle cloud is a compact specification of a bed of identical circular (2D)
 | `cloud_geometry`  | Integer | Shape of the cloud region. |
 | `shell_inner_radius` | Real | Inner radius for hemisphere-shell clouds (`cloud_geometry = 2`). |
 | `shell_outer_radius` | Real | Outer radius for hemisphere-shell clouds (`cloud_geometry = 2`). |
+| `shell_axis`      | Integer | Axis the hemisphere-shell cloud opens toward (`cloud_geometry = 2`). |
 | `moving_ibm`      | Integer | Motion flag applied to every particle (see `patch_ib(j)%%moving_ibm`). |
 | `seed`            | Integer | Random seed for reproducible placement (used by `packing_method = 1`). |
 | `packing_method`  | Integer | Algorithm used to place the particles. |
 
 - `cloud_geometry` selects the cloud region:
   - `1` (box) uses `x[y,z]_centroid` and `length_x[y,z]` to define the region.
-  - `2` uses `x[y,z]_centroid`, `shell_inner_radius`, and `shell_outer_radius` to define a half-annulus in 2D and a hemisphere shell in 3D. Particle centres are sampled between `shell_inner_radius + radius` and `shell_outer_radius - radius`, and the flat plane is kept clear by one particle radius. The flat face is fixed at `y_centroid` in 2D and `z_centroid` in 3D; the filled region opens toward positive `y` in 2D and positive `z` in 3D. The full shell extent (`x[y,z]_centroid +/- shell_outer_radius` on the open side, and one particle radius of clearance on the flat-face side) must lie inside the computational domain; a hemisphere shell also requires at least two dimensions (`n > 0`).
+  - `2` uses `x[y,z]_centroid`, `shell_inner_radius`, `shell_outer_radius`, and `shell_axis` to define a half-annulus in 2D and a hemisphere shell in 3D. Particle centres are sampled between `shell_inner_radius + radius` and `shell_outer_radius - radius`, and the flat plane is kept clear by one particle radius. `shell_axis` (`1`=x, `2`=y, `3`=z; default `3`) selects which axis the shell opens toward from its flat face at that axis's centroid; in 2D there is no z-axis, so any value other than `1` opens toward `+y` (matching the fixed behavior before `shell_axis` existed). The open axis needs one particle radius of clearance on its flat-face side and the full `shell_outer_radius` on its open side; the other axis (2D) or two axes (3D) need the full shell extent (`centroid +/- shell_outer_radius`) inside the domain. A hemisphere shell also requires at least two dimensions (`n > 0`).
 - `packing_method` selects how the `num_particles` are positioned within the cloud region:
   - `1` (rejection sampling) draws random positions and rejects any that violate `min_spacing`, producing a disordered bed. `seed` makes the placement reproducible.
   - `2` (lattice) places the particles on the optimally dense lattice for the geometry — a triangular lattice in 2D and a face-centered cubic lattice in 3D. The lattice spacing is derived from the particle density (`num_particles` over the region area/volume); if that spacing is below the required `2*radius + min_spacing`, the region is too dense and the run aborts.

src/common/m_constants.fpp

@@ -28,8 +28,8 @@ module m_constants
     integer, parameter  :: num_stl_models_max = 10
     !> Maximum number of immersed boundary patches (legacy, not used for patch_ib sizing)
     !> Fixed capacity of patch_ib (namelist patches + local particle bed subset after reduction)
-    integer, parameter  :: num_local_ibs_max = 2000            !< Maximum number of immersed boundary patches (patch_ib)
-    integer, parameter  :: num_ib_patches_max_namelist = 54000
+    integer, parameter  :: num_local_ibs_max = 8000            !< Maximum number of immersed boundary patches (patch_ib)
+    integer, parameter  :: num_ib_patches_max_namelist = 216000
     integer, parameter  :: num_particle_clouds_max = 10        !< Maximum number of particle bed patch specifications
     integer, parameter  :: num_bc_patches_max = 10             !< Maximum number of boundary condition patches
     integer, parameter  :: max_2d_fourier_modes = 10           !< Max Fourier mode index for 2D modal patch (geometry 13)

src/common/m_derived_types.fpp

@@ -379,6 +379,7 @@ module m_derived_types
         integer  :: moving_ibm  !< Motion flag: 0=static, 1=moving (forces), 2=forced path
         integer  :: seed  !< Random seed for reproducible placement
         integer  :: cloud_geometry  !< Cloud region geometry: 1=box, 2=hemisphere shell
+        integer  :: shell_axis  !< Axis the hemisphere shell opens toward: 1=x, 2=y, 3=z (2D ignores 3)
         integer  :: packing_method  !< Packing algorithm: 1=rejection sampling, 2=lattice
         integer  :: periodic  !< Periodic overlap flag for box rejection packing: 0=off, 1=on
     end type particle_cloud_parameters

src/common/m_finite_differences.fpp

@@ -30,12 +30,17 @@ contains
         real(wp), dimension(-local_buff_size:q + local_buff_size), intent(in) :: s_cc
         integer                                                               :: i       !< Generic loop iterator
 
+        ! Coefficients always extend at least fd_number_in beyond the interior on each side, so a stencil centered on a
+        ! ghost-adjacent cell (e.g. an immersed boundary near a domain boundary) has a real coefficient to read instead of
+        ! reading past the caller's allocation. offset_s, when given, widens this further (never narrows it) for callers
+        ! that need more than fd_number_in of margin.
+
         if (present(offset_s)) then
-            lB = -offset_s%beg
-            lE = q + offset_s%end
+            lB = -max(fd_number_in, offset_s%beg)
+            lE = q + max(fd_number_in, offset_s%end)
         else
-            lB = 0
-            lE = q
+            lB = -fd_number_in
+            lE = q + fd_number_in
         end if
 
         ! Computing the 1st order finite-difference coefficients

src/post_process/m_derived_variables.fpp

@@ -32,18 +32,20 @@ contains
             allocate (fd%gm_rho_sf(-offset_x%beg:m + offset_x%end,-offset_y%beg:n + offset_y%end,-offset_z%beg:p + offset_z%end))
         end if
 
-        ! Allocate FD coefficients (up to 4th order; higher orders need extension)
+        ! Allocate FD coefficients (up to 4th order; higher orders need extension). s_compute_finite_difference_coefficients
+        ! always extends at least fd_number beyond the interior on each side, widened further by offset_x/y/z when those
+        ! are larger (multi-block Silo ghost zones); the allocation must cover whichever bound ends up wider.
 
         if (omega_wrt(2) .or. omega_wrt(3) .or. qm_wrt .or. schlieren_wrt .or. liutex_wrt) then
-            allocate (fd%fd_coeff_x(-fd_number:fd_number,-offset_x%beg:m + offset_x%end))
+            allocate (fd%fd_coeff_x(-fd_number:fd_number,-max(fd_number, offset_x%beg):m + max(fd_number, offset_x%end)))
         end if
 
         if (omega_wrt(1) .or. omega_wrt(3) .or. qm_wrt .or. liutex_wrt .or. (n > 0 .and. schlieren_wrt)) then
-            allocate (fd%fd_coeff_y(-fd_number:fd_number,-offset_y%beg:n + offset_y%end))
+            allocate (fd%fd_coeff_y(-fd_number:fd_number,-max(fd_number, offset_y%beg):n + max(fd_number, offset_y%end)))
         end if
 
         if (omega_wrt(1) .or. omega_wrt(2) .or. qm_wrt .or. liutex_wrt .or. (p > 0 .and. schlieren_wrt)) then
-            allocate (fd%fd_coeff_z(-fd_number:fd_number,-offset_z%beg:p + offset_z%end))
+            allocate (fd%fd_coeff_z(-fd_number:fd_number,-max(fd_number, offset_z%beg):p + max(fd_number, offset_z%end)))
         end if
 
     end subroutine s_initialize_derived_variables_module

src/simulation/m_bubbles_EL.fpp

@@ -185,18 +185,19 @@ contains
 
         if (lag_params%vel_model > 0 .and. lag_params%pressure_force) then
             @:ALLOCATE(grad_p_x(0:m, 0:n, 0:p))
-            @:ALLOCATE(fd_coeff_x_pgrad(-fd_number:fd_number, 0:m))
+            ! s_compute_finite_difference_coefficients always extends fd_number beyond the interior on each side
+            @:ALLOCATE(fd_coeff_x_pgrad(-fd_number:fd_number,-fd_number:m + fd_number))
             call s_compute_finite_difference_coefficients(m, x_cc, fd_coeff_x_pgrad, buff_size, fd_number, fd_order)
             $:GPU_UPDATE(device='[fd_coeff_x_pgrad]')
             if (n > 0) then
                 @:ALLOCATE(grad_p_y(0:m, 0:n, 0:p))
-                @:ALLOCATE(fd_coeff_y_pgrad(-fd_number:fd_number, 0:n))
+                @:ALLOCATE(fd_coeff_y_pgrad(-fd_number:fd_number,-fd_number:n + fd_number))
                 call s_compute_finite_difference_coefficients(n, y_cc, fd_coeff_y_pgrad, buff_size, fd_number, fd_order)
                 $:GPU_UPDATE(device='[fd_coeff_y_pgrad]')
             end if
             if (p > 0) then
                 @:ALLOCATE(grad_p_z(0:m, 0:n, 0:p))
-                @:ALLOCATE(fd_coeff_z_pgrad(-fd_number:fd_number, 0:p))
+                @:ALLOCATE(fd_coeff_z_pgrad(-fd_number:fd_number,-fd_number:p + fd_number))
                 call s_compute_finite_difference_coefficients(p, z_cc, fd_coeff_z_pgrad, buff_size, fd_number, fd_order)
                 $:GPU_UPDATE(device='[fd_coeff_z_pgrad]')
             end if