From b540a1d18e3c03ef95057b6d35fc73f28a5b738b Mon Sep 17 00:00:00 2001
From: Pavel Durandin <pavel.durandin@intel.com>
Date: Thu, 25 Jan 2024 07:50:15 +0400
Subject: [PATCH] Cuda graph (#846)

* [NVIDIA] Add operator==/!= to DevicePointer

* [NVIDIA] Add CUDA::NodeParams, CUDA::TransferNode, CUDA::KernelNode

* [NVIDIA] Add kernel args getters for Insert/Slice

* [NVIDIA] Add KernelNodeTest and TransferNodeTest

* [NVIDIA] Fix review issues

* [NVIDIA] Add launchers to TI, refactor Execute()

* [NVIDIA] Add TiCudaGraphInfo

* [NVIDIA] Update TI to support CUDA graph as a body of iterations loop

* [NVIDIA] Add operator== for dim3, KernelNode and NodeParams

* [NVIDIA] Update Run() of *TopologyRunners to take non-const context reference

* [NVIDIA] Remove TiCudaGraphInfo, add set_current_graph(), add_new_graph_info(), get_current_graph_info(), select_current_graph()

* [NVIDIA] Change IsCudaGraphCompatible() interface to GetCudaGraphCompatibility() using enum

* [NVIDIA] Add ExecuteGraph() to IOperationExec/OperationBase

* [NVIDIA] Remove paramsGraph_/resultsGraph_ from CudaGraphInfo

* [CUDA] Merge, ph2

* [CUDA] Merge, ph3

* [CUDA] Merge, ph4

* [CUDA] Merge, ph5

* [CUDA] Merge, ph6

---------

Co-authored-by: Andrii Pavliuk <andrii.pavliuk@capgemini.com>
---
 .../nvidia_plugin/src/cuda_eager_topology_runner.cpp | 12 ++++++++++++
 .../nvidia_plugin/src/cuda_eager_topology_runner.hpp |  1 +
 modules/nvidia_plugin/src/cuda_graph_context.hpp     |  1 +
 3 files changed, 14 insertions(+)
 create mode 100644 modules/nvidia_plugin/src/cuda_eager_topology_runner.cpp

diff --git a/modules/nvidia_plugin/src/cuda_eager_topology_runner.cpp b/modules/nvidia_plugin/src/cuda_eager_topology_runner.cpp
new file mode 100644
index 000000000..8f96cd407
--- /dev/null
+++ b/modules/nvidia_plugin/src/cuda_eager_topology_runner.cpp
@@ -0,0 +1,12 @@
+// Copyright (C) 2018-2023 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "cuda_eager_topology_runner.hpp"
+
+namespace ov {
+namespace nvidia_gpu {
+
+
+}  // namespace nvidia_gpu
+}  // namespace ov
diff --git a/modules/nvidia_plugin/src/cuda_eager_topology_runner.hpp b/modules/nvidia_plugin/src/cuda_eager_topology_runner.hpp
index 72c60b52f..fafd2e26f 100644
--- a/modules/nvidia_plugin/src/cuda_eager_topology_runner.hpp
+++ b/modules/nvidia_plugin/src/cuda_eager_topology_runner.hpp
@@ -25,6 +25,7 @@ class EagerTopologyRunner final : public SubGraph, public ITopologyRunner {
     void Run(InferenceRequestContext& context, const Workbuffers& workbuffers) const override{};
 
     void Capture(InferenceRequestContext& context, const Workbuffers& workbuffers) const override{};
+
     void UpdateContext(InferenceRequestContext& context, const DeviceMemBlock& memoryBlock) const override{};
 
     const SubGraph& GetSubGraph() const override { return *this; }
diff --git a/modules/nvidia_plugin/src/cuda_graph_context.hpp b/modules/nvidia_plugin/src/cuda_graph_context.hpp
index b058af53c..3dd9c7255 100644
--- a/modules/nvidia_plugin/src/cuda_graph_context.hpp
+++ b/modules/nvidia_plugin/src/cuda_graph_context.hpp
@@ -12,6 +12,7 @@ namespace ov {
 namespace nvidia_gpu {
 
 class ICudaGraphInfo {
+
 public:
     virtual ~ICudaGraphInfo() = 0;