added warmup iterations to benchmarkable code

pystencils/kerncraft_coupling/generate_benchmark.py

@@ -55,8 +55,6 @@ int main(int argc, char **argv)
 
   {%- endfor %}
 
-  int repeat = atoi(argv[1]);
-
   {%- if likwid %}
   {%- if openmp %}
   #pragma omp parallel
@@ -67,6 +65,13 @@ int main(int argc, char **argv)
   likwid_markerStartRegion("loop");
   {%- endif %}
 
+  for(int warmup = 1; warmup >= 0; --warmup) {
+    int repeat = 2;
+    if(warmup == 0) {
+      repeat = atoi(argv[1]);
+      likwid_markerStartRegion("loop");
+    }
+
     for (; repeat > 0; --repeat)
     {
       {{kernelName}}({{call_argument_list}});
@@ -79,6 +84,7 @@ int main(int argc, char **argv)
       if(var_false) dummy(&{{constantName}});
       {%- endfor %}
     }
+  }
 
   {%- if likwid %}
   likwid_markerStopRegion("loop");