PME GPU/CUDA data framework.

[alexxy/gromacs.git] / src / gromacs / mdlib / sim_util.cpp

diff --git a/src/gromacs/mdlib/sim_util.cpp b/src/gromacs/mdlib/sim_util.cpp
index b0267051a3a1e3a2643afe85c274064ef1cb06a3..5278ad6873f8740aa30021833dd98348429d6b7b 100644 (file)
--- a/src/gromacs/mdlib/sim_util.cpp
+++ b/src/gromacs/mdlib/sim_util.cpp
@@ -2436,6 +2436,7 @@ void finish_run(FILE *fplog, const gmx::MDLogger &mdlog, t_commrec *cr,
                 t_nrnb nrnb[], gmx_wallcycle_t wcycle,
                 gmx_walltime_accounting_t walltime_accounting,
                 nonbonded_verlet_t *nbv,
+                gmx_pme_t *pme,
                 gmx_bool bWriteStat)
 {
     t_nrnb *nrnb_tot = nullptr;
@@ -2522,11 +2523,17 @@ void finish_run(FILE *fplog, const gmx::MDLogger &mdlog, t_commrec *cr,
 
     if (printReport)
     {
-        struct gmx_wallclock_gpu_t* gputimes = use_GPU(nbv) ? nbnxn_gpu_get_timings(nbv->gpu_nbv) : nullptr;
-
+        auto                    nbnxn_gpu_timings = use_GPU(nbv) ? nbnxn_gpu_get_timings(nbv->gpu_nbv) : nullptr;
+        gmx_wallclock_gpu_pme_t pme_gpu_timings   = {};
+        if (pme_gpu_task_enabled(pme))
+        {
+            pme_gpu_get_timings(pme, &pme_gpu_timings);
+        }
         wallcycle_print(fplog, mdlog, cr->nnodes, cr->npmenodes, nthreads_pp, nthreads_pme,
                         elapsed_time_over_all_ranks,
-                        wcycle, cycle_sum, gputimes);
+                        wcycle, cycle_sum,
+                        nbnxn_gpu_timings,
+                        &pme_gpu_timings);
 
         if (EI_DYNAMICS(inputrec->eI))
         {