diff --git a/gromacs/force_lj.c b/gromacs/force_lj.c
index cbdfe8a..9bf3732 100644
--- a/gromacs/force_lj.c
+++ b/gromacs/force_lj.c
@@ -423,8 +423,6 @@ double computeForceLJ_4xn_half(Parameter *param, Atom *atom, Neighbor *neighbor,
     MD_SIMD_FLOAT eps_vec = simd_broadcast(epsilon);
     MD_SIMD_FLOAT c48_vec = simd_broadcast(48.0);
     MD_SIMD_FLOAT c05_vec = simd_broadcast(0.5);
-    double S = getTimeStamp();
-    LIKWID_MARKER_START("force");
 
     for(int ci = 0; ci < atom->Nclusters_local; ci++) {
         int ci_vec_base = CI_VECTOR_BASE_INDEX(ci);
@@ -436,6 +434,9 @@ double computeForceLJ_4xn_half(Parameter *param, Atom *atom, Neighbor *neighbor,
         }
     }
 
+    double S = getTimeStamp();
+    LIKWID_MARKER_START("force");
+
     #pragma omp parallel for
     for(int ci = 0; ci < atom->Nclusters_local; ci++) {
         int ci_cj0 = CJ0_FROM_CI(ci);
@@ -607,8 +608,6 @@ double computeForceLJ_4xn_full(Parameter *param, Atom *atom, Neighbor *neighbor,
     MD_SIMD_FLOAT eps_vec = simd_broadcast(epsilon);
     MD_SIMD_FLOAT c48_vec = simd_broadcast(48.0);
     MD_SIMD_FLOAT c05_vec = simd_broadcast(0.5);
-    double S = getTimeStamp();
-    LIKWID_MARKER_START("force");
 
     for(int ci = 0; ci < atom->Nclusters_local; ci++) {
         int ci_vec_base = CI_VECTOR_BASE_INDEX(ci);
@@ -620,6 +619,9 @@ double computeForceLJ_4xn_full(Parameter *param, Atom *atom, Neighbor *neighbor,
         }
     }
 
+    double S = getTimeStamp();
+    LIKWID_MARKER_START("force");
+
     #pragma omp parallel for
     for(int ci = 0; ci < atom->Nclusters_local; ci++) {
         int ci_cj0 = CJ0_FROM_CI(ci);