Fix benchmark compile bug (#66069)

risemeup1 · SigureMo · web-flow · commit 22a84361d48f · 2024-07-16T10:11:05.000+08:00
---------

Co-authored-by: SigureMo &lt;sigure.qaq@gmail.com&gt;
diff --git a/paddle/common/flags.cc b/paddle/common/flags.cc
@@ -1285,6 +1285,14 @@ PHI_DEFINE_EXPORTED_bool(benchmark_nccl,
                          "enable nccl debug mode to synchronize nccl comm");
 #endif
 
+PHI_DEFINE_EXPORTED_bool(
+    benchmark,
+    false,
+    "Doing memory benchmark. It will make deleting scope synchronized, "
+    "and add some memory usage logs."
+    "Default cuda is asynchronous device, set to True will"
+    "force op run in synchronous mode.");
+
 /**
  * Autotune related FLAG
  * Name: FLAGS_use_autotune
diff --git a/paddle/fluid/framework/executor.cc b/paddle/fluid/framework/executor.cc
@@ -30,7 +30,7 @@ limitations under the License. */
 #include "paddle/common/flags.h"
 #include "paddle/fluid/framework/executor_gc_helper.h"
 
-PD_DECLARE_bool(benchmark);
+COMMON_DECLARE_bool(benchmark);
 COMMON_DECLARE_bool(use_mkldnn);
 
 namespace paddle::framework {
diff --git a/paddle/fluid/framework/new_executor/interpreter_base_impl.h b/paddle/fluid/framework/new_executor/interpreter_base_impl.h
@@ -44,7 +44,7 @@ PD_DECLARE_bool(new_executor_use_inplace);
 PD_DECLARE_bool(new_executor_use_local_scope);
 
 COMMON_DECLARE_bool(check_nan_inf);
-PD_DECLARE_bool(benchmark);
+COMMON_DECLARE_bool(benchmark);
 COMMON_DECLARE_uint64(executor_log_deps_every_microseconds);
 COMMON_DECLARE_bool(new_executor_use_cuda_graph);
 COMMON_DECLARE_bool(enable_pir_in_executor);
diff --git a/paddle/fluid/framework/operator.cc b/paddle/fluid/framework/operator.cc
@@ -61,7 +61,7 @@ class DenseTensor;
 #include "paddle/fluid/platform/device/gpu/gpu_dnn.h"
 #endif
 
-PD_DECLARE_bool(benchmark);
+COMMON_DECLARE_bool(benchmark);
 COMMON_DECLARE_bool(check_nan_inf);
 PD_DECLARE_bool(enable_unused_var_check);
 COMMON_DECLARE_bool(run_kp_kernel);
diff --git a/paddle/fluid/framework/scope.cc b/paddle/fluid/framework/scope.cc
@@ -17,8 +17,8 @@ limitations under the License. */
 #include "glog/logging.h"
 #include "paddle/common/flags.h"
 #include "paddle/fluid/framework/threadpool.h"
-PD_DECLARE_bool(benchmark);
 
+COMMON_DECLARE_bool(benchmark);
 COMMON_DECLARE_bool(eager_delete_scope);
 
 #define SCOPE_KIDS_READER_LOCK phi::AutoRDLock auto_lock(&kids_lock_);
diff --git a/paddle/fluid/imperative/prepared_operator.cc b/paddle/fluid/imperative/prepared_operator.cc
@@ -36,7 +36,7 @@
 #include "paddle/fluid/platform/profiler/supplement_tracing.h"
 
 COMMON_DECLARE_bool(check_nan_inf);
-PD_DECLARE_bool(benchmark);
+COMMON_DECLARE_bool(benchmark);
 COMMON_DECLARE_bool(run_kp_kernel);
 
 namespace paddle::imperative {
diff --git a/paddle/fluid/memory/allocation/naive_best_fit_allocator.cc b/paddle/fluid/memory/allocation/naive_best_fit_allocator.cc
@@ -41,7 +41,7 @@ PADDLE_DEFINE_EXPORTED_bool(
 COMMON_DECLARE_double(fraction_of_gpu_memory_to_use);
 COMMON_DECLARE_uint64(initial_gpu_memory_in_mb);
 COMMON_DECLARE_uint64(reallocate_gpu_memory_in_mb);
-PD_DECLARE_bool(benchmark);
+COMMON_DECLARE_bool(benchmark);
 
 namespace paddle {
 namespace memory {
diff --git a/paddle/fluid/platform/place.cc b/paddle/fluid/platform/place.cc
@@ -14,13 +14,6 @@ limitations under the License. */
 
 #include "paddle/fluid/platform/place.h"
 #include "paddle/fluid/platform/flags.h"
-PADDLE_DEFINE_EXPORTED_bool(
-    benchmark,
-    false,
-    "Doing memory benchmark. It will make deleting scope synchronized, "
-    "and add some memory usage logs."
-    "Default cuda is asynchronous device, set to True will"
-    "force op run in synchronous mode.");
 
 namespace paddle {
 namespace platform {
