[FIX] exclude zero_ op impact on benchmark (flagos-ai#1896)

Author: baoqiliu

benchmark/performance_utils.py

@@ -404,10 +404,17 @@ def run(self):
                                 self.gems_op, *args, **kwargs
                             )
                         else:
-                            with flag_gems.use_gems():
-                                metric.latency = self.get_latency(
-                                    self.torch_op, *args, **kwargs
-                                )
+                            if self.op_name == "zero_":
+                                with flag_gems.use_gems():
+                                    metric.latency = self.get_latency(
+                                        self.torch_op, *args, **kwargs
+                                    )
+                            else:
+                                # exclude flaggems' zero_ to avoid the overhead of zero_ in do_bench's clear_cache
+                                with flag_gems.use_gems(exclude=["zero_"]):
+                                    metric.latency = self.get_latency(
+                                        self.torch_op, *args, **kwargs
+                                    )
                     if "speedup" in self.to_bench_metrics:
                         metric.speedup = metric.latency_base / metric.latency
                     if "gbps" in self.to_bench_metrics: