external_function bug fix

src/ngraph/runtime/gpu/gpu_external_function.cpp

-/*******************************************************************************
-* Copyright 2017-2018 Intel Corporation
-*
-* Licensed under the Apache License, Version 2.0 (the "License");
-* you may not use this file except in compliance with the License.
-* You may obtain a copy of the License at
-*
-*     http://www.apache.org/licenses/LICENSE-2.0
-*
-* Unless required by applicable law or agreed to in writing, software
-* distributed under the License is distributed on an "AS IS" BASIS,
-* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-* See the License for the specific language governing permissions and
-* limitations under the License.
-*******************************************************************************/
+// ----------------------------------------------------------------------------
+// Copyright 2017 Nervana Systems Inc.
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// ----------------------------------------------------------------------------
 
 #include <cstdlib>
 #include <fstream>
@@ -22,6 +20,12 @@
 #include <typeindex>
 #include <typeinfo>
 #include <unordered_map>
+
+    #include <cuda_runtime.h>
+    #include <cudnn_v7.h>
+    #include "cublas_v2.h"
+    #include "cuda.h"
+
 #include "ngraph/codegen/code_writer.hpp"
 #include "ngraph/codegen/compiler.hpp"
 #include "ngraph/codegen/execution_engine.hpp"
@@ -224,8 +228,8 @@ void runtime::gpu::GPU_ExternalFunction::compile()
 
     writer +=
         R"(// Generated by the NGraph GPU backend
-    #include "cuda_runtime.h"
-    #include "cudnn_v7.h"
+    #include <cuda_runtime.h>
+    #include <cudnn_v7.h>
     #include "cublas_v2.h"
     #include "cuda.h"
 
@@ -249,6 +253,11 @@ void runtime::gpu::GPU_ExternalFunction::compile()
 
     string pch_header_source = writer.get_code();
 
+    writer += R"(
+    using namespace ngraph;
+    using namespace std;
+)";
+
     if (m_emit_timing)
     {
         writer << "// Declare debug timers\n";
@@ -324,7 +333,6 @@ void runtime::gpu::GPU_ExternalFunction::compile()
 
     writer << "void *__dso_handle = 0;\n\n";
     writer << "// Declare all constants\n";
-    unordered_map<Node*, string> match_functions;
     for (shared_ptr<Function> current_function : pass_manager.get_state().get_functions())
     {
         for (shared_ptr<Node> node : current_function->get_ordered_ops())
@@ -355,8 +363,12 @@ void runtime::gpu::GPU_ExternalFunction::compile()
 
     writer << "\n";
 
+    unordered_map<Node*, string> match_functions;
     for (shared_ptr<Function> current_function : pass_manager.get_state().get_functions())
     {
+        bool temporaries_used = false;
+        size_t worst_case_tmp_size = 0;
+
         set<string> output_names;
         for (shared_ptr<Node> op : current_function->get_results())
         {