diff --git a/paddle/fluid/operators/math/matrix_bit_code.h b/paddle/fluid/operators/math/matrix_bit_code.h
index 5454d58f371afb5f5d6a1c3208318f80d4e0aa36..07854c83584f90db02b416b85a4aa61f5cdc0685 100644
--- a/paddle/fluid/operators/math/matrix_bit_code.h
+++ b/paddle/fluid/operators/math/matrix_bit_code.h
@@ -17,6 +17,11 @@ limitations under the License. */
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/platform/device_context.h"
 
+#if defined(_WIN32)
+#include <intrin.h>
+#include <windows.h>
+#endif  // _WIN32
+
 namespace paddle {
 namespace operators {
 namespace math {
@@ -55,12 +60,38 @@ namespace math {
  *    FindLastSet(x) = 1 + \floor*{\log_{2}x}
  * \f]
  */
+#if !defined(_WIN32)
 inline constexpr size_t FindLastSet(size_t x) {
   return std::is_same<size_t, unsigned int>::value
              ? (x ? 8 * sizeof(x) - __builtin_clz(x) : 0)
              : (std::is_same<size_t, unsigned long>::value  // NOLINT
                     ? (x ? 8 * sizeof(x) - __builtin_clzl(x) : 0)
                     : (x ? 8 * sizeof(x) - __builtin_clzll(x) : 0));
+
+#else
+// windows don't have built-in clz, ctz function
+template <typename T>
+inline int ctz(const T& value) {
+  DWORD trailing_zero = 0;
+  if (_BitScanForward(&trailing_zero, value)) {
+    return static_cast<int>(trailing_zero);
+  } else {
+    return static_cast<int>(0);
+  }
+}
+
+template <typename T>
+inline int clz(const T& value) {
+  DWORD leadning_zero = 0;
+  if (_BitScanReverse(&leadning_zero, value)) {
+    return static_cast<int>(sizeof(T) * 8 - leadning_zero);
+  } else {
+    return static_cast<int>(0);
+  }
+}
+
+inline size_t FindLastSet(size_t x) { return sizeof(size_t) * 8 - clz(x); }
+#endif  // !_WIN32
 }
 
 struct SimpleCode {
diff --git a/paddle/fluid/operators/math/maxouting.h b/paddle/fluid/operators/math/maxouting.h
index 4166fb54946b7082f5f7dc0e232f636a1d2f8a13..e4d378dc23210e95605c6e09eda8a190cc5c6b4f 100644
--- a/paddle/fluid/operators/math/maxouting.h
+++ b/paddle/fluid/operators/math/maxouting.h
@@ -16,13 +16,12 @@ limitations under the License. */
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/hostdevice.h"
+#include "paddle/fluid/platform/macros.h"
 
 namespace paddle {
 namespace operators {
 namespace math {
 
-#define FLT_MAX __FLT_MAX__
-
 template <typename DeviceContext, typename T>
 class MaxOutFunctor {
  public:
diff --git a/paddle/fluid/operators/math/pooling.h b/paddle/fluid/operators/math/pooling.h
index 2538d739cce95d1b2fc5b3f905af5e6d94cf7af5..120f5919803806e0d3b7dc8eaf530ae89819b84d 100644
--- a/paddle/fluid/operators/math/pooling.h
+++ b/paddle/fluid/operators/math/pooling.h
@@ -18,15 +18,12 @@ limitations under the License. */
 #include "paddle/fluid/framework/tensor.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/hostdevice.h"
+#include "paddle/fluid/platform/macros.h"
 
 namespace paddle {
 namespace operators {
 namespace math {
 
-#define FLT_MAX \
-  __FLT_MAX__  // TODO(zcd) :It might need to be placed in another file, but I'm
-               // still wondering where to put it.
-
 /*
  * \brief Extracting simple operations from pooling.
  *        Both MaxPool and AvgPool need "initial", "compute" and "finalize"
diff --git a/paddle/fluid/platform/macros.h b/paddle/fluid/platform/macros.h
index 4cc04b090519637ab0b8d3740b8a12f216218cae..32b7efc04c1f2ecc22f93c08387aec69ded4930a 100644
--- a/paddle/fluid/platform/macros.h
+++ b/paddle/fluid/platform/macros.h
@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #pragma once
+#include <cfloat>
 
 // Disable the copy and assignment operator for a class.
 #ifndef DISABLE_COPY_AND_ASSIGN
@@ -23,3 +24,7 @@ limitations under the License. */
   classname& operator=(const classname&) = delete; \
   classname& operator=(classname&&) = delete
 #endif
+
+#if defined(__FLT_MAX__)
+#define FLT_MAX __FLT_MAX__
+#endif  // __FLT_MAX__