diff --git a/paddle/fluid/operators/mkldnn/matmul_mkldnn_op.cc b/paddle/fluid/operators/mkldnn/matmul_mkldnn_op.cc
index 8efd2b226cad6f27c12036d863dba5a60ebf586f..2c84218c48e0bcc2d22d032bf5b3e949424aec3a 100644
--- a/paddle/fluid/operators/mkldnn/matmul_mkldnn_op.cc
+++ b/paddle/fluid/operators/mkldnn/matmul_mkldnn_op.cc
@@ -71,9 +71,6 @@ static Tensor FoldFirstAndLastDims(const MKLDNNDeviceContext& dev_ctx,
   auto reorder_p = reorder_handler.AcquireReorder(reorder_src_memory_p,
                                                   reorder_dst_memory_p);
 
-  paddle::platform::RecordEvent record_reorder(
-      "int_reorder", paddle::platform::EventRole::kUniqueOp);
-
   auto& astream = MKLDNNDeviceContext::tls().get_stream();
   reorder_p->execute(astream, *reorder_src_memory_p, *reorder_dst_memory_p);
   astream.wait();
diff --git a/paddle/fluid/operators/mkldnn/quantize_mkldnn_op.cc b/paddle/fluid/operators/mkldnn/quantize_mkldnn_op.cc
index c44f22dd02face48fe344ea2ee91ead4e9836837..deb8c735c8b0260d80c016439b8f1ae6765b56c5 100644
--- a/paddle/fluid/operators/mkldnn/quantize_mkldnn_op.cc
+++ b/paddle/fluid/operators/mkldnn/quantize_mkldnn_op.cc
@@ -106,12 +106,8 @@ class QuantOpKernel : public framework::OpKernel<T> {
     reorder_p = std::shared_ptr<reorder>(new reorder(*reorder_pd));
 
     auto& astream = platform::MKLDNNDeviceContext::tls().get_stream();
-    {
-      platform::RecordEvent record_reorder("int_reorder",
-                                           platform::EventRole::kUniqueOp);
-      reorder_p->execute(astream, *src_memory, *dst_memory);
-      astream.wait();
-    }
+    reorder_p->execute(astream, *src_memory, *dst_memory);
+    astream.wait();
 
     output->set_layout(DataLayout::kMKLDNN);
     output->set_format(GetMKLDNNFormat(*dst_memory));
diff --git a/paddle/fluid/operators/mkldnn/requantize_mkldnn_op.cc b/paddle/fluid/operators/mkldnn/requantize_mkldnn_op.cc
index 1b1bd69aec2f4d88a65d66fd9a59d9ea9c78ee66..963f10441f9bdf2ac3369a770c43ca92ac21a7bf 100644
--- a/paddle/fluid/operators/mkldnn/requantize_mkldnn_op.cc
+++ b/paddle/fluid/operators/mkldnn/requantize_mkldnn_op.cc
@@ -138,12 +138,9 @@ class ReQuantOpKernel : public framework::OpKernel<T> {
     }
 
     auto& astream = platform::MKLDNNDeviceContext::tls().get_stream();
-    {
-      platform::RecordEvent record_reorder("int_reorder",
-                                           platform::EventRole::kUniqueOp);
-      reorder_p->execute(astream, *src_memory, *dst_memory);
-      astream.wait();
-    }
+
+    reorder_p->execute(astream, *src_memory, *dst_memory);
+    astream.wait();
 
     output->set_layout(framework::DataLayout::kMKLDNN);
     output->set_format(platform::GetMKLDNNFormat(*dst_memory));
diff --git a/paddle/fluid/operators/mkldnn/sum_mkldnn_op.cc b/paddle/fluid/operators/mkldnn/sum_mkldnn_op.cc
index 5df2a546812adec055573a0d9c2c5c373fbed928..9c63afff13c22c7ad4ec283f2b25c2bc4535e6d1 100644
--- a/paddle/fluid/operators/mkldnn/sum_mkldnn_op.cc
+++ b/paddle/fluid/operators/mkldnn/sum_mkldnn_op.cc
@@ -174,12 +174,9 @@ class SumMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
           output, in_out.format(), ctx.GetPlace());
 
       auto reorder_p = reorder_handler.AcquireReorder(target_mem, dst_mem);
-      {
-        platform::RecordEvent record_reorder("int_reorder",
-                                             platform::EventRole::kUniqueOp);
-        reorder_p->execute(astream, *dst_mem, *target_mem);
-        astream.wait();
-      }
+
+      reorder_p->execute(astream, *dst_mem, *target_mem);
+      astream.wait();
     }
     output->set_layout(framework::DataLayout::kMKLDNN);
     output->set_format(platform::GetMKLDNNFormat(*dst_mem));
diff --git a/paddle/fluid/operators/reduce_ops/mkldnn/reduce_mkldnn_op.h b/paddle/fluid/operators/reduce_ops/mkldnn/reduce_mkldnn_op.h
index 5dcb81c75407f7c4b4a2f787d04e3085f366b348..a27b6ae90f29a14af5e0a119fb3f5d0182dafa7c 100644
--- a/paddle/fluid/operators/reduce_ops/mkldnn/reduce_mkldnn_op.h
+++ b/paddle/fluid/operators/reduce_ops/mkldnn/reduce_mkldnn_op.h
@@ -85,9 +85,6 @@ class ReduceMKLDNNKernel : public framework::OpKernel<T> {
       auto reorder_p = reorder_handler.AcquireReorder(reorder_src_memory_p,
                                                       reorder_dst_memory_p);
 
-      platform::RecordEvent record_reorder("int_reorder",
-                                           platform::EventRole::kUniqueOp);
-
       reorder_p->execute(astream, *reorder_src_memory_p, *reorder_dst_memory_p);
       astream.wait();