Cuda graph (openvinotoolkit#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 <[email protected]>
apaniukov · Jan 25, 2024 · b540a1d · b540a1d
1 parent 63a55e1
commit b540a1d
Show file tree

Hide file tree

Showing 3 changed files with 14 additions and 0 deletions.
diff --git a/modules/nvidia_plugin/src/cuda_eager_topology_runner.cpp 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
@@ -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
@@ -12,6 +12,7 @@ namespace ov {
 namespace nvidia_gpu {
 
 class ICudaGraphInfo {
+
 public:
     virtual ~ICudaGraphInfo() = 0;