Frontier fix ups

Author: wilfonba

CMakeLists.txt

@@ -164,7 +164,6 @@ elseif (CMAKE_Fortran_COMPILER_ID STREQUAL "Cray")
         "SHELL: -h acc_model=no_fast_addr"
         "SHELL: -h list=adm"
         "SHELL: -munsafe-fp-atomics" # Not unsafe for operations we do
-        "SHELL: -O3 -hacc"
     )
 
     add_link_options("SHELL:-hkeepfiles")

src/common/m_mpi_common.fpp

@@ -1055,6 +1055,8 @@ contains
             !! Remaining number of cells, in a particular coordinate direction,
             !! after the majority is divided up among the available processors
 
+        integer :: recon_order !< reconstruction order
+
         integer :: i, j !< Generic loop iterators
 
         if (num_procs == 1 .and. parallel_io) then
@@ -1064,6 +1066,12 @@ contains
             return
         end if
 
+        if (igr) then
+            recon_order = igr_order
+        else
+            recon_order = weno_order
+        end if
+
         ! 3D Cartesian Processor Topology
         if (n > 0) then
 
@@ -1092,7 +1100,7 @@ contains
 
                         if (mod(num_procs, i) == 0 &
                             .and. &
-                            (m + 1)/i >= num_stcls_min*weno_order) then
+                            (m + 1)/i >= num_stcls_min*recon_order) then
 
                             tmp_num_procs_x = i
                             tmp_num_procs_y = num_procs/i
@@ -1102,7 +1110,7 @@ contains
                                 .and. &
                                 (n + 1)/tmp_num_procs_y &
                                 >= &
-                                num_stcls_min*weno_order) then
+                                num_stcls_min*recon_order) then
 
                                 num_procs_x = i
                                 num_procs_y = num_procs/i
@@ -1138,13 +1146,13 @@ contains
 
                         if (mod(num_procs, i) == 0 &
                             .and. &
-                            (m + 1)/i >= num_stcls_min*weno_order) then
+                            (m + 1)/i >= num_stcls_min*recon_order) then
 
                             do j = 1, num_procs/i
 
                                 if (mod(num_procs/i, j) == 0 &
                                     .and. &
-                                    (n + 1)/j >= num_stcls_min*weno_order) then
+                                    (n + 1)/j >= num_stcls_min*recon_order) then
 
                                     tmp_num_procs_x = i
                                     tmp_num_procs_y = j
@@ -1157,7 +1165,7 @@ contains
                                         .and. &
                                         (p + 1)/tmp_num_procs_z &
                                         >= &
-                                        num_stcls_min*weno_order) &
+                                        num_stcls_min*recon_order) &
                                         then
 
                                         num_procs_x = i
@@ -1186,7 +1194,7 @@ contains
                 if (proc_rank == 0 .and. ierr == -1) then
                     call s_mpi_abort('Unsupported combination of values '// &
                                      'of num_procs, m, n, p and '// &
-                                     'weno_order. Exiting.')
+                                     'weno/igr_order. Exiting.')
                 end if
 
                 ! Creating new communicator using the Cartesian topology
@@ -1294,7 +1302,7 @@ contains
 
                     if (mod(num_procs, i) == 0 &
                         .and. &
-                        (m + 1)/i >= num_stcls_min*weno_order) then
+                        (m + 1)/i >= num_stcls_min*recon_order) then
 
                         tmp_num_procs_x = i
                         tmp_num_procs_y = num_procs/i
@@ -1304,7 +1312,7 @@ contains
                             .and. &
                             (n + 1)/tmp_num_procs_y &
                             >= &
-                            num_stcls_min*weno_order) then
+                            num_stcls_min*recon_order) then
 
                             num_procs_x = i
                             num_procs_y = num_procs/i
@@ -1323,7 +1331,7 @@ contains
                 if (proc_rank == 0 .and. ierr == -1) then
                     call s_mpi_abort('Unsupported combination of values '// &
                                      'of num_procs, m, n and '// &
-                                     'weno_order. Exiting.')
+                                     'weno/igr_order. Exiting.')
                 end if
 
                 ! Creating new communicator using the Cartesian topology

src/simulation/m_global_parameters.fpp

@@ -195,7 +195,7 @@ module m_global_parameters
     $:GPU_DECLARE(create='[mpp_lim,model_eqns,mixture_err,alt_soundspeed]')
     $:GPU_DECLARE(create='[avg_state,mp_weno,weno_eps,teno_CT,hypoelasticity]')
     $:GPU_DECLARE(create='[hyperelasticity,hyper_model,elasticity,low_Mach]')
-    $:GPU_DECLARE(create='[viscous,shear_stress,bulk_stress,cont_damage]')
+    $:GPU_DECLARE(create='[shear_stress,bulk_stress,cont_damage]')
 
     logical :: relax          !< activate phase change
     integer :: relax_model    !< Relaxation model

src/simulation/m_igr.fpp

@@ -163,7 +163,7 @@ contains
     end subroutine s_initialize_igr_module
 
     subroutine s_igr_iterative_solve(q_cons_vf, bc_type, t_step)
-
+        !DIR$ OPTIMIZE (-haggress)
         type(scalar_field), dimension(sys_size), intent(inout) :: q_cons_vf
         type(integer_field), dimension(1:num_dims, -1:1), intent(in) :: bc_type
         integer, intent(in) :: t_step
@@ -261,7 +261,7 @@ contains
     end subroutine s_igr_iterative_solve
 
     subroutine s_igr_sigma_x(q_cons_vf, rhs_vf)
-
+        !DIR$ OPTIMIZE (-haggress)
         type(scalar_field), &
             dimension(sys_size), &
             intent(inout) :: rhs_vf
@@ -328,7 +328,7 @@ contains
     end subroutine s_igr_sigma_x
 
     subroutine s_igr_riemann_solver(q_cons_vf, rhs_vf, idir)
-
+        !DIR$ OPTIMIZE (-haggress)
         type(scalar_field), &
             dimension(sys_size), &
             intent(inout) :: rhs_vf

src/simulation/m_start_up.fpp

@@ -1143,7 +1143,9 @@ contains
                 io_time_final = maxval(io_proc_time)
             end if
 
-            grind_time = time_final*1.0e9_wp/(sys_size*maxval((/1,m_glb/))*maxval((/1,n_glb/))*maxval((/1,p_glb/)))
+            grind_time = time_final * 1.0e9_wp / &
+                (real(sys_size, wp) * real(maxval((/1, m_glb/)), wp) * &
+                 real(maxval((/1, n_glb/)), wp) * real(maxval((/1, p_glb/)), wp))
 
             print *, "Performance:", grind_time, "ns/gp/eq/rhs"
             inquire (FILE='time_data.dat', EXIST=file_exists)