fix old style cast error (#632)

src/ngraph/runtime/gpu/gpu_cuda_kernel_emitters.cpp

@@ -53,12 +53,12 @@ void cuda_)" + name + "(" + data_type +
 
                 //convert runtime ptr to driver api ptr
                 CUdeviceptr d_ptr_in, d_ptr_out;
-                d_ptr_in = (CUdeviceptr)in;
-                d_ptr_out = (CUdeviceptr)out;
+                d_ptr_in = CUdeviceptr(in);
+                d_ptr_out = CUdeviceptr(out);
 
                 void* args_list[] = {&d_ptr_in, &d_ptr_out, &repeat_size, &repeat_times, &count};
                 CUDA_SAFE_CALL(cuLaunchKernel(*CudaFunctionPool::instance().get(name).get(),
-                                              count,
+                                              static_cast<unsigned int>(count),
                                               1,
                                               1, // grid dim
                                               1,

src/ngraph/runtime/gpu/gpu_emitter.cpp

@@ -183,7 +183,6 @@ cudnnSetOpTensorDescriptor(opTensorDesc,
                     return;
                 }
 
-                const ngraph::op::Dot* dot = static_cast<const ngraph::op::Dot*>(node);
                 const Shape& arg0_shape = args[0].get_shape();
                 const Shape& arg1_shape = args[1].get_shape();
                 if (arg0_shape.empty() || arg1_shape.empty())
@@ -504,15 +503,11 @@ cudnnSetOpTensorDescriptor(opTensorDesc,
                 writer.indent++;
                 auto arg_shape = args[0].get_shape();
                 auto arg_rank = arg_shape.size();
-
                 auto result_shape = out[0].get_shape();
-                auto& result_element_type = out[0].get_element_type();
-
                 auto input_order = reshape->get_input_order();
-
                 bool same_layout = is_sorted(input_order.begin(), input_order.end());
-
                 size_t result_shape_product = 1;
+
                 for (auto i : result_shape)
                 {
                     result_shape_product *= i;

src/ngraph/runtime/gpu/gpu_external_function.cpp

@@ -427,9 +427,6 @@ using namespace std;
                 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())
                     {
@@ -454,18 +451,6 @@ using namespace std;
                             continue;
                         }
                         string match_function_name;
-                        for (size_t j = i + 1; j < op_list.size(); j++)
-                        {
-                            if (0) //op_list[i]->is_functionally_identical(*op_list[j]))
-                            {
-                                if (match_function_name.empty())
-                                {
-                                    match_function_name = "func_" + op_list[i]->get_name();
-                                    match_functions.insert({op_list[i].get(), match_function_name});
-                                }
-                                match_functions.insert({op_list[j].get(), match_function_name});
-                            }
-                        }
                         if (!match_function_name.empty())
                         {
                             writer << "static void " << match_function_name << "(";
@@ -545,28 +530,24 @@ using namespace std;
                     writer << "{\n";
                     writer.indent++;
 
-                    for (shared_ptr<Function> current_function :
-                         pass_manager.get_state().get_functions())
+                    for (shared_ptr<Node> node : current_function->get_ordered_ops())
                     {
-                        for (shared_ptr<Node> node : current_function->get_ordered_ops())
+                        const op::Constant* c = dynamic_cast<op::Constant*>(node.get());
+                        if (c)
                         {
-                            const op::Constant* c = dynamic_cast<op::Constant*>(node.get());
-                            if (c)
-                            {
-                                shared_ptr<descriptor::TensorView> tv =
-                                    node->get_outputs()[0].get_tensor_view();
-                                writer << "if(" << tv->get_tensor().get_name() << " == NULL)\n";
-                                writer << "{\n";
-                                writer.indent++;
-                                writer << "runtime::gpu::cuda_memcpyHtD("
-                                       << tv->get_tensor().get_name() << ", "
-                                       << tv->get_tensor().get_name() << "_cpu, "
-                                       << tv->get_tensor().size() << ");\n";
-                                writer.indent--;
-                                writer << "}\n";
-                            }
+                            shared_ptr<descriptor::TensorView> tv =
+                                node->get_outputs()[0].get_tensor_view();
+                            writer << "if(" << tv->get_tensor().get_name() << " == NULL)\n";
+                            writer << "{\n";
+                            writer.indent++;
+                            writer << "runtime::gpu::cuda_memcpyHtD(" << tv->get_tensor().get_name()
+                                   << ", " << tv->get_tensor().get_name() << "_cpu, "
+                                   << tv->get_tensor().size() << ");\n";
+                            writer.indent--;
+                            writer << "}\n";
                         }
                     }
+
                     bool temporaries_used = false;
                     size_t worst_case_tmp_size = 0;
                     for (shared_ptr<Node> node : current_function->get_ordered_ops())

src/ngraph/runtime/gpu/gpu_tensor_view.cpp

@@ -41,7 +41,7 @@ runtime::gpu::GPU_TensorView::GPU_TensorView(const ngraph::element::Type& elemen
     m_buffer_size = shape_size(shape) * element_type.size();
     if (m_buffer_size > 0)
     {
-        cudaMalloc((void**)&m_allocated_buffer_pool, m_buffer_size);
+        cudaMalloc(static_cast<void**>(&m_allocated_buffer_pool), m_buffer_size);
     }
 }
 

src/ngraph/runtime/gpu/gpu_util.cpp

@@ -50,7 +50,7 @@ void runtime::gpu::check_cuda_errors(CUresult err)
 void* runtime::gpu::create_gpu_buffer(size_t buffer_size)
 {
     void* allocated_buffer_pool;
-    cudaMalloc((void**)&allocated_buffer_pool, buffer_size);
+    cudaMalloc(static_cast<void**>(&allocated_buffer_pool), buffer_size);
     return allocated_buffer_pool;
 }