diff --git a/CMakeLists.txt b/CMakeLists.txt
index 84bed9ad77..f154246472 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -64,6 +64,8 @@ cmake_dependent_option( T8CODE_BUILD_DOCUMENTATION_SPHINX "Build t8code's docume
 set(T8CODE_CUSTOM_PARALLEL_TEST_COMMAND "" CACHE STRING "Define a custom command for parallel tests , e.g.: mpirun -np 8 (overwrites standard mpirun -np 4 if build with mpi)")
 set(T8CODE_CUSTOM_SERIAL_TEST_COMMAND "" CACHE STRING "Define a custom command for serial tests.")
 
+option( T8CODE_ENABLE_PROFILING_BARRIERS "Use MPI-Barriers for profiling" OFF )
+
 # Reading the option T8CODE_TEST_LEVEL: Possible choices are T8_TEST_LEVEL_FULL, T8_TEST_LEVEL_MEDIUM, or T8_TEST_LEVEL_BASIC. Default is T8_TEST_LEVEL_FULL.
 set(T8CODE_TEST_LEVEL "T8_TEST_LEVEL_FULL" CACHE STRING "Test level: T8_TEST_LEVEL_FULL for full tests, T8_TEST_LEVEL_MEDIUM for less thorough tests, T8_TEST_LEVEL_BASIC for minimal tests. (WARNING: Use with care.)")
 set_property(CACHE T8CODE_TEST_LEVEL PROPERTY STRINGS "T8_TEST_LEVEL_FULL" "T8_TEST_LEVEL_MEDIUM" "T8_TEST_LEVEL_BASIC")
diff --git a/benchmarks/CMakeLists.txt b/benchmarks/CMakeLists.txt
index 3293f0127b..2d8b305e60 100644
--- a/benchmarks/CMakeLists.txt
+++ b/benchmarks/CMakeLists.txt
@@ -44,6 +44,9 @@ function( add_t8_benchmark )
     if( T8CODE_EXPORT_COMPILE_COMMANDS )
         set_target_properties( ${ADD_T8_BENCHMARK_NAME} PROPERTIES EXPORT_COMPILE_COMMANDS ON )
     endif( T8CODE_EXPORT_COMPILE_COMMANDS )
+    if( T8CODE_ENABLE_PROFILING_BARRIERS )
+        target_compile_definitions( ${ADD_T8_BENCHMARK_NAME} PUBLIC T8_ENABLE_PROFILE_BARRIER=1 )
+    endif( T8CODE_ENABLE_PROFILING_BARRIERS )
 
     install( TARGETS ${ADD_T8_BENCHMARK_NAME} DESTINATION ${CMAKE_INSTALL_BINDIR} )
 endfunction()
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index d4d4e90ce2..4cb5a5ea2f 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -51,6 +51,10 @@ if( T8CODE_EXPORT_COMPILE_COMMANDS )
   set_target_properties( T8 PROPERTIES EXPORT_COMPILE_COMMANDS ON )
 endif( T8CODE_EXPORT_COMPILE_COMMANDS )
 
+if( T8CODE_ENABLE_PROFILING_BARRIERS )
+  target_compile_definitions( T8 PUBLIC T8_ENABLE_PROFILE_BARRIER=1 )
+endif( T8CODE_ENABLE_PROFILING_BARRIERS )
+
 if( T8CODE_ENABLE_NETCDF )
   target_link_libraries( T8 PUBLIC NetCDF::NetCDF )
   target_compile_definitions(T8 PUBLIC
diff --git a/src/t8_forest/t8_forest.cxx b/src/t8_forest/t8_forest.cxx
index e4c9f5f435..2ef8b7ceb4 100644
--- a/src/t8_forest/t8_forest.cxx
+++ b/src/t8_forest/t8_forest.cxx
@@ -3032,6 +3032,9 @@ t8_forest_commit (t8_forest_t forest)
   T8_ASSERT (!forest->committed);
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of commit */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->commit_runtime = sc_MPI_Wtime ();
   }
 
@@ -3247,6 +3250,9 @@ t8_forest_commit (t8_forest_t forest)
 
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of commit */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->commit_runtime = sc_MPI_Wtime () - forest->profile->commit_runtime;
   }
 
diff --git a/src/t8_forest/t8_forest_balance.cxx b/src/t8_forest/t8_forest_balance.cxx
index 7381628049..f137b98372 100644
--- a/src/t8_forest/t8_forest_balance.cxx
+++ b/src/t8_forest/t8_forest_balance.cxx
@@ -164,6 +164,9 @@ t8_forest_balance (t8_forest_t forest, int repartition)
   adap_stats = ghost_stats = partition_stats = NULL;
 
   if (forest->profile != NULL) {
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     /* Profiling is enable, so we measure the runtime of balance */
     forest->profile->balance_runtime = -sc_MPI_Wtime ();
     /* We store the individual adapt, ghost, and partition runtimes */
@@ -285,6 +288,9 @@ t8_forest_balance (t8_forest_t forest, int repartition)
 
   if (forest->profile != NULL) {
     /* Profiling is enabled, so we measure the runtime of balance. */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->balance_runtime += sc_MPI_Wtime ();
     forest->profile->balance_rounds = count_rounds;
     /* Print the runtime of adapt/ghost/partition */
diff --git a/src/t8_forest/t8_forest_ghost.cxx b/src/t8_forest/t8_forest_ghost.cxx
index f5c578d696..8eccb1acc0 100644
--- a/src/t8_forest/t8_forest_ghost.cxx
+++ b/src/t8_forest/t8_forest_ghost.cxx
@@ -1551,6 +1551,9 @@ t8_forest_ghost_create_ext (t8_forest_t forest, int unbalanced_version)
 
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of ghost_create */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->ghost_runtime = -sc_MPI_Wtime ();
     /* DO NOT DELETE THE FOLLOWING line.
      * even if you do not want this output. It fixes a bug that occurred on JUQUEEN, where the
@@ -1608,6 +1611,9 @@ t8_forest_ghost_create_ext (t8_forest_t forest, int unbalanced_version)
 
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of ghost_create */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->ghost_runtime += sc_MPI_Wtime ();
     /* We also store the number of ghosts and remotes */
     if (ghost != NULL) {
diff --git a/src/t8_forest/t8_forest_partition.cxx b/src/t8_forest/t8_forest_partition.cxx
index 994b1d5f83..a4f61a0e0c 100644
--- a/src/t8_forest/t8_forest_partition.cxx
+++ b/src/t8_forest/t8_forest_partition.cxx
@@ -92,6 +92,9 @@ t8_forest_partition_create_offsets (t8_forest_t forest)
 {
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->forest_offsets_runtime = -sc_MPI_Wtime ();
   }
 
@@ -120,6 +123,9 @@ t8_forest_partition_create_offsets (t8_forest_t forest)
   t8_shmem_array_end_writing (forest->element_offsets);
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->forest_offsets_runtime += sc_MPI_Wtime ();
   }
 }
@@ -266,6 +272,9 @@ t8_forest_partition_create_first_desc (t8_forest_t forest)
 {
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->first_descendant_runtime = -sc_MPI_Wtime ();
   }
   sc_MPI_Comm comm;
@@ -348,6 +357,9 @@ t8_forest_partition_create_first_desc (t8_forest_t forest)
 #endif
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->first_descendant_runtime += sc_MPI_Wtime ();
   }
 }
@@ -358,6 +370,9 @@ t8_forest_partition_create_tree_offsets (t8_forest_t forest)
 
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->cmesh_offsets_runtime = -sc_MPI_Wtime ();
   }
 
@@ -426,6 +441,9 @@ t8_forest_partition_create_tree_offsets (t8_forest_t forest)
   }
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->cmesh_offsets_runtime += sc_MPI_Wtime ();
   }
 }
@@ -1209,6 +1227,9 @@ t8_forest_partition (t8_forest_t forest)
 
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->partition_runtime = sc_MPI_Wtime ();
 
     /* DO NOT DELETE THE FOLLOWING line.
@@ -1239,6 +1260,9 @@ t8_forest_partition (t8_forest_t forest)
 
   if (forest->profile != NULL) {
     /* If profiling is enabled, we measure the runtime of partition */
+#if T8_ENABLE_PROFILE_BARRIER
+    sc_MPI_Barrier (forest->mpicomm);
+#endif
     forest->profile->partition_runtime = sc_MPI_Wtime () - forest->profile->partition_runtime;
 
     /* DO NOT DELETE THE FOLLOWING line.