Merge pull request #5711 from hedaoyuan/inference

The pool_limit_size need be zero in mobile inference.

Merge pull request #5711 from hedaoyuan/inference
The pool_limit_size need be zero in mobile inference.
bf712b5c · hedaoyuan · GitHub · 6f22e6c4 · b34f21bd · bf712b5c
显示空白变更内容
内联并排

Showing with 4 addition and 0 deletion

paddle/math/Storage.cpp paddle/math/Storage.cpp +4 -0

未找到文件。
--- a/paddle/math/Storage.cpp
+++ b/paddle/math/Storage.cpp
@@ -17,9 +17,13 @@ limitations under the License. */
 #include "paddle/utils/StringUtil.h"
 #include "paddle/utils/Util.h"

+#ifndef PADDLE_MOBILE_INFERENCE
 DEFINE_int32(pool_limit_size,
             536870912,
             "maximum memory size managed by a memory pool, default is 512M");
+#else
+DEFINE_int32(pool_limit_size, 0, "default is 0");
+#endif

 namespace paddle {