diff --git a/cblas.h b/cblas.h
index 8a5055cf8..ade2fca3a 100644
--- a/cblas.h
+++ b/cblas.h
@@ -12,6 +12,7 @@ extern "C" {
 /*Set the number of threads on runtime.*/
 void openblas_set_num_threads(int num_threads);
 void goto_set_num_threads(int num_threads);
+int openblas_set_num_threads_local(int num_threads);
 
 /*Get the number of threads on runtime.*/
 int openblas_get_num_threads(void);
diff --git a/common_thread.h b/common_thread.h
index 9e7dae74a..d37fcb189 100644
--- a/common_thread.h
+++ b/common_thread.h
@@ -137,19 +137,20 @@ typedef struct blas_queue {
 
 extern int blas_server_avail;
 extern int blas_omp_number_max;
+extern int blas_omp_threads_local;
 
 static __inline int num_cpu_avail(int level) {
 
 #ifdef USE_OPENMP
 int openmp_nthreads;
 	openmp_nthreads=omp_get_max_threads();
+	if (omp_in_parallel()) openmp_nthreads = blas_omp_threads_local;
 #endif
 
 #ifndef USE_OPENMP 
   if (blas_cpu_number == 1
-#endif
-#ifdef USE_OPENMP
-     if (openmp_nthreads == 1 || omp_in_parallel()
+#else
+     if (openmp_nthreads == 1 
 #endif
       ) return 1;