Add comprehensive checks for problem cells

docs/documentation/case.md

@@ -90,11 +90,18 @@ Definition of the parameters is described in the following subsections.
 | ---:             |    :----:      |          :---                             |
 | `run_time_info`  | Logical        | Output run-time information               |
 | `rdma_mpi`       | Logical        | (GPUs) Enable RDMA for MPI communication. |
+| `comp_debug`     | Logical        | Comprehensive variable checking           |
 
 - `run_time_info` generates a text file that includes run-time information including the CFL number(s) at each time-step.
 - `rdma_mpi` optimizes data transfers between GPUs using Remote Direct Memory Access (RDMA).
 The underlying MPI implementation and communication infrastructure must support this
 feature, detecting GPU pointers and performing RDMA accordingly.
+- `comp_debug` enables comprehensive error checking.
+At each Runge-Kutta sub-step, all conservative variables are checked for NaNs.
+The volume fractions are checked to ensure they are in the range [0, 1].
+Negative densities are also checked for.
+If any of these checks find problems, the file `comp_debug.txt` will be written to the case directory with information
+about what problems were found and the simulation state will be saved for visualization.
 
 ### 2. Computational Domain
 

examples/3D_performance_test/comp_debug.txt

@@ -0,0 +1 @@
+ -1

src/post_process/m_global_parameters.fpp

@@ -109,6 +109,7 @@ module m_global_parameters
     logical :: mixture_err     !< Mixture error limiter
     logical :: alt_soundspeed  !< Alternate sound speed
     logical :: hypoelasticity  !< Turn hypoelasticity on
+    logical :: comp_debug      !< Turn on comprehensive debug
     logical, parameter :: chemistry = .${chemistry}$. !< Chemistry modeling
     !> @}
 
@@ -317,6 +318,7 @@ contains
         relax = .false.
         relax_model = dflt_int
         hypoelasticity = .false.
+        comp_debug = .false.
 
         bc_x%beg = dflt_int; bc_x%end = dflt_int
         bc_y%beg = dflt_int; bc_y%end = dflt_int

src/post_process/m_mpi_proxy.fpp

@@ -169,7 +169,7 @@ contains
             & 'heat_ratio_wrt', 'pi_inf_wrt', 'pres_inf_wrt', 'cons_vars_wrt', &
             & 'prim_vars_wrt', 'c_wrt', 'qm_wrt','schlieren_wrt', 'bubbles', 'qbmm',   &
             & 'polytropic', 'polydisperse', 'file_per_process', 'relax', 'cf_wrt',     &
-            & 'adv_n', 'ib', 'cfl_adap_dt', 'cfl_const_dt', 'cfl_dt' ]
+            & 'adv_n', 'ib', 'cfl_adap_dt', 'cfl_const_dt', 'cfl_dt', 'comp_debug' ]
             call MPI_BCAST(${VAR}$, 1, MPI_LOGICAL, 0, MPI_COMM_WORLD, ierr)
         #:endfor
 

src/post_process/m_start_up.f90

@@ -83,7 +83,7 @@
             polydisperse, poly_sigma, file_per_process, relax, &
             relax_model, cf_wrt, sigma, adv_n, ib, &
             cfl_adap_dt, cfl_const_dt, t_save, t_stop, n_start, &
-            cfl_target
+            cfl_target, comp_debug
 
         ! Inquiring the status of the post_process.inp file
         file_loc = 'post_process.inp'
@@ -184,6 +184,56 @@
 
     end subroutine s_perform_time_step
 
+    subroutine s_perform_comprehensive_debug(varname, pres, c, H)
+
+        real(kind(0d0)) :: pres
+        real(kind(0d0)) :: c
+        real(kind(0d0)) :: H
+        integer :: t_fail, ios
+        logical :: exists
+        character(LEN=name_len) :: varname !<
+        character(LEN=name_len) :: file_name = 'comp_debug.txt'
+        character(LEN=path_len + name_len) :: file_path
+
+        ! Opening the run-time information file
+        file_path = trim(case_dir)//'/'//trim(file_name)
+
+        inquire (file=file_path, exist=exists)
+        if (exists) then
+            open (12, file=file_path)
+            ! Read the file line by line
+            do
+                read (12, '(I9)', iostat=ios) t_fail
+                if (ios /= 0) exit  ! Exit loop on error or end-of-file
+            end do
+            if (t_fail == -1) return
+        else
+            return
+        end if
+
+        if (proc_rank == 0) then
+            print("(A, I6)"), " Post Processing suspicious time-step: ", t_fail
+        end if
+
+        ! Populating the grid and conservative variables
+        call s_read_data_files(t_fail)
+        ! Populating the buffer regions of the grid variables
+        if (buff_size > 0) then
+            call s_populate_grid_variables_buffer_regions()
+        end if
+
+        ! Populating the buffer regions of the conservative variables
+        if (buff_size > 0) then
+            call s_populate_conservative_variables_buffer_regions()
+        end if
+
+        ! Converting the conservative variables to the primitive ones
+        call s_convert_conservative_to_primitive_variables(q_cons_vf, q_prim_vf)
+
+        call s_save_data(t_fail, varname, pres, c, H)
+
+    end subroutine s_perform_comprehensive_debug
+
     subroutine s_save_data(t_step, varname, pres, c, H)
 
         integer, intent(inout) :: t_step

src/post_process/p_main.fpp

@@ -34,6 +34,8 @@ program p_main
 
     call s_initialize_modules()
 
+    if (comp_debug) call s_perform_comprehensive_debug(varname, pres, c, H)
+
     if (cfl_dt) then
         t_step = n_start
         n_save = int(t_stop/t_save) + 1

src/simulation/m_global_parameters.fpp

@@ -159,6 +159,7 @@ module m_global_parameters
     logical :: hypoelasticity !< hypoelasticity modeling
     logical, parameter :: chemistry = .${chemistry}$. !< Chemistry modeling
     logical :: cu_tensor
+    logical :: comp_debug     !< Variable checking at every RK step
 
     logical :: bodyForces
     logical :: bf_x, bf_y, bf_z !< body force toggle in three directions
@@ -637,6 +638,9 @@ contains
         ! Cuda aware MPI
         cu_tensor = .false.
 
+        ! Comprehensive debugging
+        comp_debug = .false.
+
         bodyForces = .false.
         bf_x = .false.; bf_y = .false.; bf_z = .false.
         !< amplitude, frequency, and phase shift sinusoid in each direction

src/simulation/m_mpi_proxy.fpp

@@ -203,7 +203,7 @@ contains
             & 'polydisperse', 'qbmm', 'acoustic_source', 'probe_wrt', 'integral_wrt',   &
             & 'prim_vars_wrt', 'weno_avg', 'file_per_process', 'relax',          &
             & 'adv_n', 'adap_dt', 'ib', 'bodyForces', 'bf_x', 'bf_y', 'bf_z',    &
-            & 'cfl_adap_dt', 'cfl_const_dt', 'cfl_dt' ]
+            & 'cfl_adap_dt', 'cfl_const_dt', 'cfl_dt', 'comp_debug' ]
             call MPI_BCAST(${VAR}$, 1, MPI_LOGICAL, 0, MPI_COMM_WORLD, ierr)
         #:endfor
 

src/simulation/m_sim_helpers.f90

@@ -6,13 +6,16 @@
     use m_global_parameters
 
     use m_variables_conversion
+
+    use m_mpi_proxy
     ! ==========================================================================
 
     implicit none
 
     private; public :: s_compute_enthalpy, &
  s_compute_stability_from_dt, &
- s_compute_dt_from_cfl
+ s_compute_dt_from_cfl, &
+ s_check_cells
 
 contains
 
@@ -245,4 +248,84 @@
 
     end subroutine s_compute_dt_from_cfl
 
+    subroutine s_check_cells(q_cons_Vf, t_step, stage, errors)
+
+        type(scalar_field), dimension(sys_size) :: q_cons_vf
+        integer, intent(in) :: t_step, stage
+        integer :: j, k, l, i
+        integer errors
+        logical :: exists
+
+        character(LEN=name_len) :: file_name = 'comp_debug.txt'
+        character(LEN=path_len + name_len) :: file_path
+        character(100) :: str_format
+
+        ! Opening the run-time information file
+        file_path = trim(case_dir)//'/'//trim(file_name)
+
+        str_format = "(I9, A, I3, A, I4, I4, I4, A, I2, A, I5, A, I5, I5, I5)"
+
+        open (12, FILE=trim(file_path), &
+              STATUS='replace')
+
+        errors = 0
+
+        ! Check all variables for NaNs
+        do i = 1, sys_size
+            !$acc update host(q_cons_vf(i)%sf)
+            do l = 0, p
+                do k = 0, n
+                    do j = 0, m
+                        if (ieee_is_nan(q_cons_vf(i)%sf(j, k, l))) then
+                            write (12, str_format) t_step, " NaN(s) in conservative variables after RK stage ", &
+                                stage, " at (j,k,l) ", j, k, l, " equation", i, " proc", proc_rank, &
+                                " (m, n, p)", m, n, p
+                            errors = errors + 1
+                        end if
+                    end do
+                end do
+            end do
+        end do
+
+        ! Check for invalid volume fractions
+        do i = advxb, advxe
+            do l = 0, p
+                do k = 0, n
+                    do j = 0, m
+                        if (q_cons_vf(i)%sf(j, k, l) < 0d0) then
+                            write (12, str_format) t_step, " Volume fraction < 0 after RK stage ", &
+                                stage, " at (j,k,l) ", j, k, l, " equation", i, " proc", proc_rank, &
+                                " (m, n, p)", m, n, p
+                            errors = errors + 1
+                        elseif (q_cons_vf(i)%sf(j, k, l) > 1d0 + verysmall) then
+                            write (12, str_format) t_step, " Volume fraction > 1 after RK stage ", &
+                                stage, " at (j,k,l) ", j, k, l, " equation", i, " proc", proc_rank, &
+                                " (m, n, p)", m, n, p
+                            errors = errors + 1
+                        end if
+                    end do
+                end do
+            end do
+        end do
+
+        ! Check for invalid densities
+        do i = contxb, contxe
+            do l = 0, p
+                do k = 0, n
+                    do j = 0, m
+                        if (q_cons_vf(advxb + i - 1)%sf(j, k, l) < 0d0 .and. q_cons_vf(i)%sf(j, k, l) < 0d0 .or. &
+                            q_cons_vf(advxb + i - 1)%sf(j, k, l) > 0d0 .and. q_cons_Vf(i)%sf(j, k, l) < 0d0) then
+                            print *, q_cons_vf(advxb + i - 1)%sf(j, k, l), q_cons_vf(i)%sf(j, k, l)
+                            write (12, str_format) t_step, " Density is negative after RK stage ", &
+                                stage, " at (j,k,l) ", j, k, l, " equation", i, " proc", proc_rank, &
+                                " (m, n, p)", m, n, p
+                            errors = errors + 1
+                        end if
+                    end do
+                end do
+            end do
+        end do
+
+    end subroutine
+
 end module m_sim_helpers

src/simulation/m_start_up.fpp

@@ -162,7 +162,8 @@ contains
             pi_fac, adv_n, adap_dt, bf_x, bf_y, bf_z, &
             k_x, k_y, k_z, w_x, w_y, w_z, p_x, p_y, p_z, &
             g_x, g_y, g_z, n_start, t_save, t_stop, &
-            cfl_adap_dt, cfl_const_dt, cfl_target
+            cfl_adap_dt, cfl_const_dt, cfl_target, &
+            comp_debug
 
         ! Checking that an input file has been provided by the user. If it
         ! has, then the input file is read in, otherwise, simulation exits.