merge pr165 some commits

Signed-off-by: Maverick123123 <yuming.wu@dynamia.ai>

Author: maverick123123

docs/performance-optimizations.md

@@ -46,9 +46,8 @@ void rate_limiter(int grids, int blocks) {
   if (cached_util_switch == 0)
       return;
 
-  /* Defensive guard for external orchestration (currently unreachable) */
-  while (get_recent_kernel() < 0) {
-    usleep(1000);
+  while (get_recent_kernel()<0) {
+    sleep(1);
   }
   set_recent_kernel(2);