diff --git a/tinygrad/runtime/support/hcq.py b/tinygrad/runtime/support/hcq.py index 177d6c88f2..8dbfa1df1c 100644 --- a/tinygrad/runtime/support/hcq.py +++ b/tinygrad/runtime/support/hcq.py @@ -406,6 +406,8 @@ class HCQCompiled(Compiled, Generic[SignalType]): return self.signal_t(base_buf=HCQCompiled.signal_pool[pg].pop(), owner=self, **kwargs) def _at_profile_finalize(self): + self.synchronize() # Expect device to be synchronizes + def _sync(d:HCQCompiled, q_t:Callable[[], HWQueue]): q_t().timestamp(d.timeline_signal).signal(d.timeline_signal, d.next_timeline()).submit(d) st = time.perf_counter_ns()