3

Author: yanghaoran29

src/a2a3/runtime/tensormap_and_ringbuffer/aicpu/aicpu_executor.cpp

@@ -836,17 +836,16 @@ struct AicpuExecutor {
         if (block_num == 0) return;
 
         uint32_t all_acked = (1u << active_sched_threads_) - 1;
-        uint32_t my_ack = (1u << thread_idx);
 
         // Ack barrier — signal this thread has stopped dispatch.
-        drain_state_.drain_ack_mask.fetch_or(my_ack, std::memory_order_release);
+        drain_state_.drain_ack_mask.fetch_or(1u << thread_idx, std::memory_order_release);
 
         // Spin until all threads have acked.
         // If our bit is cleared while waiting, elected reset due to insufficient resources.
         while (true) {
             uint32_t ack = drain_state_.drain_ack_mask.load(std::memory_order_acquire);
             if ((ack & all_acked) == all_acked) break;
-            if ((ack & my_ack) == 0) return;
+            if ((ack & (1u << thread_idx)) == 0) return;
             SPIN_WAIT_HINT();
         }
 

src/a5/runtime/tensormap_and_ringbuffer/aicpu/aicpu_executor.cpp

@@ -824,17 +824,16 @@ struct AicpuExecutor {
         if (block_num == 0) return;
 
         uint32_t all_acked = (1u << active_sched_threads_) - 1;
-        uint32_t my_ack = (1u << thread_idx);
 
         // Ack barrier — signal this thread has stopped dispatch.
-        drain_state_.drain_ack_mask.fetch_or(my_ack, std::memory_order_release);
+        drain_state_.drain_ack_mask.fetch_or(1u << thread_idx, std::memory_order_release);
 
         // Spin until all threads have acked.
         // If our bit is cleared while waiting, elected reset due to insufficient resources.
         while (true) {
             uint32_t ack = drain_state_.drain_ack_mask.load(std::memory_order_acquire);
             if ((ack & all_acked) == all_acked) break;
-            if ((ack & my_ack) == 0) return;
+            if ((ack & (1u << thread_idx)) == 0) return;
             SPIN_WAIT_HINT();
         }