diff --git a/paddle/fluid/lite/host/target_wrapper.cc b/paddle/fluid/lite/host/target_wrapper.cc
index 2e8393ef46689ef3b77fbb7945885809d7b70cea..af4269390cbe198b5f46e551ec611040b0fa332e 100644
--- a/paddle/fluid/lite/host/target_wrapper.cc
+++ b/paddle/fluid/lite/host/target_wrapper.cc
@@ -14,15 +14,29 @@
 
 #include "paddle/fluid/lite/core/target_wrapper.h"
 #include <cstring>
+#include <memory>
 
 namespace paddle {
 namespace lite {
 
+const int MALLOC_ALIGN = 64;
+
 void* TargetWrapper<TARGET(kHost)>::Malloc(size_t size) {
-  return new char[size];
+  size_t offset = sizeof(void*) + MALLOC_ALIGN - 1;
+  char* p = static_cast<char*>(malloc(offset + size));
+  if (!p) {
+    return nullptr;
+  }
+  void* r = reinterpret_cast<void*>(reinterpret_cast<size_t>(p + offset) &
+                                    (~(MALLOC_ALIGN - 1)));
+  static_cast<void**>(r)[-1] = p;
+  memset(r, 0, size);
+  return r;
 }
 void TargetWrapper<TARGET(kHost)>::Free(void* ptr) {
-  delete[] static_cast<char*>(ptr);
+  if (ptr) {
+    free(static_cast<void**>(ptr)[-1]);
+  }
 }
 void TargetWrapper<TARGET(kHost)>::MemcpySync(void* dst, const void* src,
                                               size_t size, IoDirection dir) {