OP2 profile

.github/workflows/local_ci.yml

@@ -32,7 +32,7 @@ jobs:
       - name: Checkout repo
         uses: actions/checkout@v3
         with:
-          ref: OP2_refactor
+          token: ${{ github.token }}
 
       # 2. Debug info ---------------------------------------
       - name: Print runner info

CHANGELOG.md

@@ -43,7 +43,7 @@ Major changes since `v1.1.0` (high-level).
 - `op_get_global_set_offset` return type changed from `int` to `idx_g_t`.
 - `SafeLong` debug wrapper (`op2/include/SafeLong.h`, `op2/src/core/SafeLong.cpp`): optional arithmetic overflow/underflow checker for `idx_g_t`, enabled via `-DUSE_SAFELONG`.
 - `op_arg_idx` / `op_arg_info` support in C (previously Fortran only); 2-dim map variant added for Fortran.
-- `op_timing2`: improved timing and instrumentation API (`op2/include/op_timing2.h`).
+- **`op_profile`**: tree-based timing and instrumentation API (`op2/include/op_profile.h`), replacing the interim `op_timing2` name. Functions: `op_profile_start`, `op_profile_enter`, `op_profile_enter_kernel`, `op_profile_next`, `op_profile_exit`, `op_profile_end`, `op_profile_output`, `op_profile_output_json`. Controlled by `OP_PROFILE_LEVEL` (0–3) and `OP_PROFILE_JSON_OUTPUT` environment variables. Fortran bindings provided with identical names.
 - `op_mpi_probe_halo_index`, `op_force_part`: new MPI utility routines.
 - `op_reset_data_ptr` `real(4)` variants added (Fortran).
 - `op_get_global_set_offset` and Fortran bindings for `op_mpi_get_data`.

CODEBASE_OVERVIEW.md

@@ -110,7 +110,7 @@ OP2-Common/
 | `op_mpi_core.h` | MPI halo data structures: `halo_list`, import/export lists, MPI comms |
 | `op_lib_mpi.h` | MPI runtime state: exec/non-exec halo lists, partition tables |
 | `op_hdf5.h` | Parallel HDF5 I/O API |
-| `op_timing2.h` | Tree-based timing instrumentation (JSON output, 4 detail levels) |
+| `op_profile.h` | Tree-based timing instrumentation (JSON output, 4 detail levels) |
 | `op_util.h` | Utility functions |
 | `SafeLong.h` | Debug wrapper type `SafeLong` for `idx_g_t` — detects integer overflow/underflow at runtime (enabled via `-DUSE_SAFELONG`) |
 | `fortran/` | Fortran C-interop headers |

apps/c/aero/aero_hdf5/aero.cpp

@@ -46,6 +46,7 @@ double gm1, gm1i, wtg1[2], xi1[2], Ng1[4], Ng1_xi[4], wtg2[4], Ng2[16],
 //
 
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -202,8 +203,7 @@ int main(int argc, char **argv) {
   ncell = op_get_size(cells);
   nbnodes = op_get_size(bnodes);
 
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Aero");
 
   // main time-marching loop
 
@@ -324,8 +324,7 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timing_output();
-  op_timers(&cpu_t2, &wall_t2);
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_end();
+  op_profile_output();
   op_exit();
 }

apps/c/aero/aero_plain/aero.cpp

@@ -46,6 +46,7 @@ double gm1, gm1i, wtg1[2], xi1[2], Ng1[4], Ng1_xi[4], wtg2[4], Ng2[16],
 //
 
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -263,8 +264,7 @@ int main(int argc, char **argv) {
 
   op_diagnostic_output();
 
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Aero");
 
   // main fixpoint iteration loop
 
@@ -387,8 +387,7 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timers(&cpu_t2, &wall_t2);
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_end();
+  op_profile_output();
   op_exit();
 }

apps/c/aero/aero_plain/aero_mpi.cpp

@@ -53,6 +53,7 @@ double gm1, gm1i, wtg1[2], xi1[2], Ng1[4], Ng1_xi[4], wtg2[4], Ng2[16],
 
 #include "op_lib_mpi.h"
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -147,7 +148,6 @@ int main(int argc, char **argv) {
   MPI_Comm_size(MPI_COMM_WORLD, &comm_size);
 
   // timer
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
 
   int *bnode, *cell, *g_bnode, *g_cell;
   double *xm, *g_xm;
@@ -369,7 +369,7 @@ int main(int argc, char **argv) {
 
   niter = 20;
   // initialise timers for total execution wall time
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Aero");
   for (int iter = 1; iter <= niter; iter++) {
 
     op_par_loop(res_calc, "res_calc", cells,
@@ -486,8 +486,7 @@ int main(int argc, char **argv) {
       }
     }
   }
-  op_timers(&cpu_t2, &wall_t2);
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_end();
+  op_profile_output();
   op_exit();
 }

apps/c/airfoil/airfoil_hdf5/dp/airfoil.cpp

@@ -55,6 +55,7 @@ double gam, gm1, cfl, eps, mach, alpha, qinf[4];
 //
 
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -83,7 +84,6 @@ int main(int argc, char **argv) {
   double rms;
 
   // timer
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
 
   // set constants and initialise flow field and residual
   op_printf("initialising flow field \n");
@@ -158,7 +158,7 @@ int main(int argc, char **argv) {
   int g_ncell = op_get_size(cells);
 
   // initialise timers for total execution wall time
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Airfoil");
 
   // main time-marching loop
 
@@ -240,7 +240,7 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timers(&cpu_t2, &wall_t2);
+  op_profile_end();
 
   // write given op_dat's indicated segment of data to a memory block in the
   // order it was originally
@@ -262,7 +262,6 @@ int main(int argc, char **argv) {
   // compress using
   // ~/hdf5/bin/h5repack -f GZIP=9 new_grid.h5 new_grid_pack.h5
 
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_output();
   op_exit();
 }

apps/c/airfoil/airfoil_hdf5/sp/airfoil.cpp

@@ -55,6 +55,7 @@ float gam, gm1, cfl, eps, mach, alpha, qinf[4];
 //
 
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -76,7 +77,6 @@ int main(int argc, char **argv) {
   float rms;
 
   // timer
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
 
   // set constants and initialise flow field and residual
   op_printf("initialising flow field \n");
@@ -129,7 +129,7 @@ int main(int argc, char **argv) {
   int g_ncell = op_get_size(cells);
 
   // initialise timers for total execution wall time
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Airfoil");
 
   // main time-marching loop
 
@@ -210,9 +210,8 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timers(&cpu_t2, &wall_t2);
+  op_profile_end();
 
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_output();
   op_exit();
 }

apps/c/airfoil/airfoil_plain/dp/airfoil.cpp

@@ -55,6 +55,7 @@ double gam, gm1, cfl, eps, mach, alpha, qinf[4];
 //
 
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -79,7 +80,6 @@ int main(int argc, char **argv) {
   double rms, maxerr;
 
   // timer
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
 
   // read in grid
 
@@ -212,7 +212,7 @@ int main(int argc, char **argv) {
   op_diagnostic_output();
 
   // initialise timers for total execution wall time
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Airfoil");
 
   // main time-marching loop
 
@@ -296,7 +296,7 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timers(&cpu_t2, &wall_t2);
+  op_profile_end();
 
   // output the result dat array to files
   op_print_dat_to_txtfile(p_q, "out_grid_seq.dat"); // ASCI
@@ -309,8 +309,7 @@ int main(int argc, char **argv) {
   op_fetch_data_idx(p_q, q_part, 0, op_get_size(cells) - 1);
   free(q_part);
 
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_output();
 
   op_exit();
 }

apps/c/airfoil/airfoil_plain/dp/airfoil_mpi.cpp

@@ -63,6 +63,7 @@ double gam, gm1, cfl, eps, mach, alpha, qinf[4];
 
 #include "op_lib_mpi.h"
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -154,7 +155,6 @@ int main(int argc, char **argv) {
   MPI_Comm_size(MPI_COMM_WORLD, &comm_size);
 
   // timer
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
 
   int *becell, *ecell, *bound, *bedge, *edge, *cell;
   double *x, *q, *qold, *adt, *res;
@@ -164,7 +164,7 @@ int main(int argc, char **argv) {
 
   /**------------------------BEGIN I/O and PARTITIONING -------------------**/
 
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Airfoil");
 
   /* read in grid from disk on root processor */
   FILE *fp;
@@ -310,9 +310,6 @@ int main(int argc, char **argv) {
     free(g_res);
   }
 
-  op_timers(&cpu_t2, &wall_t2);
-  op_printf("Max total file read time = %f\n", wall_t2 - wall_t1);
-
   /**------------------------END I/O and PARTITIONING -----------------------**/
 
   // declare sets, pointers, datasets and global constants
@@ -361,7 +358,6 @@ int main(int argc, char **argv) {
   op_partition("PARMETIS", "KWAY", cells, pecell, p_x);
 
   // initialise timers for total execution wall time
-  op_timers(&cpu_t1, &wall_t1);
 
   niter = 1000;
   for (int iter = 1; iter <= niter; iter++) {
@@ -440,7 +436,7 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timers(&cpu_t2, &wall_t2);
+  op_profile_end();
 
   // output the result dat array to files
   op_print_dat_to_txtfile(p_q, "out_grid_mpi.dat"); // ASCI
@@ -453,8 +449,7 @@ int main(int argc, char **argv) {
   op_fetch_data_idx(p_q, q_part, 0, op_get_size(cells) - 1);
   free(q_part);
 
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_output();
 
   op_exit();
 }

apps/c/airfoil/airfoil_plain/sp/airfoil.cpp

@@ -55,6 +55,7 @@ float gam, gm1, cfl, eps, mach, alpha, qinf[4];
 //
 
 #include "op_seq.h"
+#include <op_profile.h>
 
 //
 // kernel routines for parallel loops
@@ -79,7 +80,6 @@ int main(int argc, char **argv) {
   float rms;
 
   // timer
-  double cpu_t1, cpu_t2, wall_t1, wall_t2;
 
   // read in grid
 
@@ -212,7 +212,7 @@ int main(int argc, char **argv) {
   op_diagnostic_output();
 
   // initialise timers for total execution wall time
-  op_timers(&cpu_t1, &wall_t1);
+  op_profile_start("Airfoil");
 
   // main time-marching loop
 
@@ -291,9 +291,8 @@ int main(int argc, char **argv) {
     }
   }
 
-  op_timers(&cpu_t2, &wall_t2);
-  op_timing_output();
-  op_printf("Max total runtime = %f\n", wall_t2 - wall_t1);
+  op_profile_end();
+  op_profile_output();
 
   op_exit();
 }