Use wait-for-event to not spin in the blas_lock

common_arm64.h

@@ -53,16 +53,16 @@ static void __inline blas_lock(volatile BLASULONG *address){
   BLASULONG ret;
 
   do {
-    while (*address) {YIELDING;};
-
     __asm__ __volatile__(
 			 "mov	x4, #1							\n\t"
+			 "sevl								\n\t"
                          "1:                                                            \n\t"
+                         "wfe                                                           \n\t"
+			 "2:								\n\t"
                          "ldaxr x2, [%1]                                                \n\t"
                          "cbnz  x2, 1b                                                  \n\t"
-			 "2:								\n\t"
                          "stxr  w3, x4, [%1]                                            \n\t"
-                         "cbnz  w3, 1b                                                  \n\t"
+                         "cbnz  w3, 2b                                                  \n\t"
                          "mov   %0, #0                                                  \n\t"
                          : "=r"(ret), "=r"(address)
                          : "1"(address)