Merge pull request !20648 from zyli2020/mindrt_debugtags/v1.4.0
| @@ -329,10 +329,10 @@ std::vector<KernelWithIndex> AnfRuntimeAlgorithm::GetAllOutputWithIndex(const An | |||
| auto value_tuple = value->cast<ValueTuplePtr>(); | |||
| auto value_tuple_size = CountValueNum(value_tuple); | |||
| for (size_t i = 0; i < value_tuple_size; ++i) { | |||
| ret.push_back({node, i}); | |||
| ret.emplace_back(node, i); | |||
| } | |||
| } else { | |||
| ret.push_back({node, 0}); | |||
| ret.emplace_back(node, 0); | |||
| } | |||
| return ret; | |||
| } | |||
| @@ -126,7 +126,7 @@ void KernelActor::RunOpControlWithInputTensor(AID *const input_control, OpContex | |||
| MS_EXCEPTION_IF_NULL(context); | |||
| MS_EXCEPTION_IF_NULL(input_tensors); | |||
| auto &sequential_num = context->sequential_num_; | |||
| input_op_controls_[sequential_num].emplace_back(input_control); | |||
| (void)input_op_controls_[sequential_num].emplace_back(input_control); | |||
| PushInputDeviceTensor(input_tensors); | |||
| // When all the inputs are collected, then allocate memory and callback launch. | |||
| @@ -56,7 +56,7 @@ void CreateParameterDeviceAddress(const DeviceContext *device_context, const Ker | |||
| MS_EXCEPTION_IF_NULL(graph); | |||
| std::vector<AnfNodePtr> graph_inputs = graph->inputs(); | |||
| const std::vector<bool> &graph_valid_input = graph->valid_inputs(); | |||
| graph_inputs.insert(graph_inputs.end(), graph->child_graph_result().begin(), graph->child_graph_result().end()); | |||
| (void)graph_inputs.insert(graph_inputs.end(), graph->child_graph_result().begin(), graph->child_graph_result().end()); | |||
| // Anf nodes which need create device address. | |||
| std::vector<AnfNodePtr> nodes_list; | |||
| @@ -408,7 +408,7 @@ GraphId GraphCompiler::CompileGraph(const session::OpRunInfo &op_run_info, const | |||
| auto &outputs_with_index = run_op_graph_output_nodes_[graph->graph_id()]; | |||
| for (auto &node : output_nodes) { | |||
| MS_EXCEPTION_IF_NULL(node); | |||
| outputs_with_index.emplace_back(AnfAlgo::VisitKernelWithReturnType(node, 0, false)); | |||
| (void)outputs_with_index.emplace_back(AnfAlgo::VisitKernelWithReturnType(node, 0, false)); | |||
| } | |||
| UpdateRefCountForGraphOutput(outputs_with_index); | |||
| @@ -289,7 +289,7 @@ void EraseValueNodeTensor(const std::vector<int64_t> *tensors_mask, const std::v | |||
| } | |||
| for (size_t index = 0; index < tensors_mask->size(); ++index) { | |||
| if (tensors_mask->at(index) != kValueNodeTensorMask) { | |||
| input_tensors_without_value_node->emplace_back(input_tensors->at(index)); | |||
| (void)input_tensors_without_value_node->emplace_back(input_tensors->at(index)); | |||
| } | |||
| } | |||
| } | |||
| @@ -2173,7 +2173,7 @@ void GraphScheduler::LinkArrowByControlNode(const GraphCompilerInfo &graph_compi | |||
| LinkDataArrowByControlNode(graph_compiler_info, input_with_index, from_func_graph, gather_actor, i); | |||
| } | |||
| } | |||
| LinkBranchArrowForSwitchActor(graph_compiler_info, actor_set); | |||
| LinkBranchArrowForSwitchActor(graph_compiler_info); | |||
| LinkBranchArrowForGatherActor(graph_compiler_info, actor_set); | |||
| @@ -2482,7 +2482,7 @@ void GraphScheduler::LinkControlArrowForSwitchActor(std::vector<SwitchActorPtr> | |||
| if (actor != nullptr) { | |||
| const auto &gather_actor = dynamic_cast<GatherActor *>(actor); | |||
| MS_EXCEPTION_IF_NULL(gather_actor); | |||
| switch_actor->output_branch_control_arrows_[i].emplace_back(gather_actor->GetAID()); | |||
| (void)switch_actor->output_branch_control_arrows_[i].emplace_back(gather_actor->GetAID()); | |||
| gather_actor->input_controls_num_++; | |||
| } | |||
| } | |||
| @@ -2516,13 +2516,12 @@ void GraphScheduler::LinkControlArrowForSwitchActor(std::vector<SwitchActorPtr> | |||
| switch_actor->branch_id_to_index_[kMainBranchID] = branch_index; | |||
| } | |||
| switch_actor->output_branch_control_arrows_[branch_index].emplace_back(to_actor->GetAID()); | |||
| (void)switch_actor->output_branch_control_arrows_[branch_index].emplace_back(to_actor->GetAID()); | |||
| } | |||
| } | |||
| } | |||
| void GraphScheduler::LinkBranchArrowForSwitchActor(const GraphCompilerInfo &graph_compiler_info, | |||
| const ActorSet *actor_set) { | |||
| void GraphScheduler::LinkBranchArrowForSwitchActor(const GraphCompilerInfo &graph_compiler_info) { | |||
| for (const auto &control_node : graph_compiler_info.control_nodes_) { | |||
| if (AnfAlgo::CheckPrimitiveType(control_node, prim::kPrimSwitch) || | |||
| AnfAlgo::CheckPrimitiveType(control_node, prim::kPrimSwitchLayer)) { | |||
| @@ -256,7 +256,7 @@ class GraphScheduler { | |||
| const KernelMapPosition &origin_outputs_order); | |||
| // In control flow, there are scenarios where there are multi-branch outputs, and the gather actor needs to | |||
| // send the branch id to the loop count actor. | |||
| void LinkBranchArrowForSwitchActor(const GraphCompilerInfo &graph_compiler_info, const ActorSet *actor_set); | |||
| void LinkBranchArrowForSwitchActor(const GraphCompilerInfo &graph_compiler_info); | |||
| void LinkBranchArrowForGatherActor(const GraphCompilerInfo &graph_compiler_info, const ActorSet *actor_set); | |||
| void LinkOutputResultArrowForSwitchActor(const GraphCompilerInfo &graph_compiler_info, const ActorSet *actor_set); | |||
| void PrepareDataForControlNode(HostQueueDataSourceActor *host_data_source_actor, | |||
| @@ -37,9 +37,9 @@ namespace device { | |||
| namespace cpu { | |||
| using mindspore::kernel::KernelBuildInfo; | |||
| bool CPUDeviceContext::Initialize() { | |||
| void CPUDeviceContext::Initialize() { | |||
| if (initialized_) { | |||
| return true; | |||
| return; | |||
| } | |||
| mem_manager_ = std::make_shared<CPUMemoryManager>(); | |||
| @@ -55,7 +55,6 @@ bool CPUDeviceContext::Initialize() { | |||
| #endif | |||
| initialized_ = true; | |||
| return true; | |||
| } | |||
| bool CPUDeviceContext::AllocateMemory(DeviceAddress *const &address, size_t size) const { | |||
| @@ -130,14 +129,14 @@ void SetControlOpInfo(const CNodePtr &kernel_node) { | |||
| std::vector<TypeId> inputs_type; | |||
| size_t input_num = AnfAlgo::GetInputTensorNum(kernel_node); | |||
| for (size_t input_index = 0; input_index < input_num; ++input_index) { | |||
| inputs_format.emplace_back(kOpFormat_DEFAULT); | |||
| (void)inputs_format.emplace_back(kOpFormat_DEFAULT); | |||
| inputs_type.push_back(AnfAlgo::GetPrevNodeOutputInferDataType(kernel_node, input_index)); | |||
| } | |||
| std::vector<std::string> outputs_format; | |||
| std::vector<TypeId> outputs_type; | |||
| size_t output_num = AnfAlgo::GetOutputTensorNum(kernel_node); | |||
| for (size_t output_index = 0; output_index < output_num; ++output_index) { | |||
| outputs_format.emplace_back(kOpFormat_DEFAULT); | |||
| (void)outputs_format.emplace_back(kOpFormat_DEFAULT); | |||
| outputs_type.push_back(AnfAlgo::GetOutputInferDataType(kernel_node, output_index)); | |||
| } | |||
| @@ -240,7 +239,7 @@ bool CPUDeviceContext::LaunchKernelWithProfiling(const CNodePtr &kernel, const s | |||
| auto kernel_mod = AnfAlgo::GetKernelMod(kernel); | |||
| MS_EXCEPTION_IF_NULL(kernel_mod); | |||
| uint32_t pid = getpid(); | |||
| uint32_t pid = IntToUint(getpid()); | |||
| profiler_inst->OpDataProducerBegin(kernel->fullname_with_scope(), pid); | |||
| bool ret = DoLaunchKernel(kernel_mod, inputs, workspace, outputs); | |||
| profiler_inst->OpDataProducerEnd(); | |||
| @@ -33,7 +33,7 @@ class CPUDeviceContext : public DeviceContext { | |||
| : DeviceContext(device_context_key), mem_manager_(nullptr), initialized_(false) {} | |||
| ~CPUDeviceContext() override = default; | |||
| bool Initialize() override; | |||
| void Initialize() override; | |||
| bool AllocateMemory(DeviceAddress *const &address, size_t size) const override; | |||
| void FreeMemory(DeviceAddress *const &address) const override; | |||
| @@ -46,8 +46,8 @@ class DeviceContext { | |||
| explicit DeviceContext(const DeviceContextKey &device_context_key) : device_context_key_(device_context_key) {} | |||
| virtual ~DeviceContext() = default; | |||
| // Initialize the device context and return success or not. | |||
| virtual bool Initialize() = 0; | |||
| // Initialize the device context. | |||
| virtual void Initialize() = 0; | |||
| // Destroy device context and release device resource. | |||
| virtual void Destroy() {} | |||
| @@ -64,7 +64,7 @@ void DeviceContextManager::UpdateDeviceContextKey(const DeviceContextKey &old_ke | |||
| } | |||
| handle.key() = new_key_str; | |||
| device_contexts_.insert(std::move(handle)); | |||
| (void)device_contexts_.insert(std::move(handle)); | |||
| } | |||
| } // namespace device | |||
| } // namespace mindspore | |||
| @@ -47,13 +47,13 @@ using KernelGraph = mindspore::session::KernelGraph; | |||
| static thread_local bool cur_thread_device_inited{false}; | |||
| bool GPUDeviceContext::Initialize() { | |||
| void GPUDeviceContext::Initialize() { | |||
| if (initialized_ == true) { | |||
| if (!BindDeviceToCurrentThread()) { | |||
| return false; | |||
| MS_LOG(EXCEPTION) << "BindDeviceToCurrentThread failed."; | |||
| } | |||
| GPUMemoryAllocator::GetInstance().CheckMaxDeviceMemory(); | |||
| return true; | |||
| return; | |||
| } | |||
| // Set device id | |||
| @@ -74,10 +74,8 @@ bool GPUDeviceContext::Initialize() { | |||
| } | |||
| // Set device id and initialize device resource. | |||
| bool ret = InitDevice(); | |||
| if (!ret) { | |||
| MS_LOG(ERROR) << "GPU InitDevice failed."; | |||
| return ret; | |||
| if (!InitDevice()) { | |||
| MS_LOG(EXCEPTION) << "GPU InitDevice failed."; | |||
| } | |||
| // Initialize memory pool. | |||
| @@ -101,7 +99,6 @@ bool GPUDeviceContext::Initialize() { | |||
| json_parser.CopyMSCfgJsonToDir(rank_id); | |||
| initialized_ = true; | |||
| return ret; | |||
| } | |||
| bool GPUDeviceContext::InitDevice() { | |||
| @@ -34,7 +34,7 @@ class GPUDeviceContext : public DeviceContext { | |||
| ~GPUDeviceContext() override = default; | |||
| // Set device id and initialize device resource, such as stream, cudnn and cublas handle. | |||
| bool Initialize() override; | |||
| void Initialize() override; | |||
| // Release device memory, stream, cudnn and cublas handle, etc. | |||
| void Destroy() override; | |||
| @@ -382,7 +382,7 @@ const ActorInfo &MindRTBackend::CompileGraphs(const FuncGraphPtr &func_graph) { | |||
| } | |||
| MS_EXCEPTION_IF_NULL(graph_compiler_info); | |||
| const ActorInfo &actor_info = graph_compiler_info->name_; | |||
| actor_to_graph_compiler_info_.emplace(graph_compiler_info->name_, std::move(graph_compiler_info)); | |||
| (void)actor_to_graph_compiler_info_.emplace(graph_compiler_info->name_, std::move(graph_compiler_info)); | |||
| return actor_info; | |||
| } | |||
| @@ -531,7 +531,7 @@ void PlantTensorTupleToVector(const py::tuple &tuple_inputs, std::vector<tensor: | |||
| } | |||
| auto tensor = py::cast<tensor::TensorPtr>(input_object); | |||
| MS_EXCEPTION_IF_NULL(tensor); | |||
| tensors->emplace_back(tensor); | |||
| (void)tensors->emplace_back(tensor); | |||
| } | |||
| } | |||
| @@ -547,7 +547,7 @@ void ConvertValueTupleToTensor(const py::object &input_object, std::vector<tenso | |||
| MS_EXCEPTION_IF_NULL(value_tuple); | |||
| tensor::TensorPtr tensor_ptr = opt::CreateTupleTensor(value_tuple); | |||
| MS_EXCEPTION_IF_NULL(tensor_ptr); | |||
| tensors->emplace_back(tensor_ptr); | |||
| (void)tensors->emplace_back(tensor_ptr); | |||
| } | |||
| void ConvertMultiPyObjectToTensor(const py::object &input_object, std::vector<tensor::TensorPtr> *tensors) { | |||
| @@ -714,7 +714,7 @@ void MindRTBackend::RunGraphBySingleOp(const std::vector<KernelGraphPtr> &graphs | |||
| auto iter = cnode_ref_counts_.find(graph->graph_id()); | |||
| if (iter == cnode_ref_counts_.end()) { | |||
| graph_compiler_->CalculateRefCount(graph, &cnode_ref_count); | |||
| cnode_ref_counts_.emplace(graph->graph_id(), cnode_ref_count); | |||
| (void)cnode_ref_counts_.emplace(graph->graph_id(), cnode_ref_count); | |||
| } else { | |||
| cnode_ref_count = iter->second; | |||
| } | |||
| @@ -944,7 +944,7 @@ std::unique_ptr<GraphCompilerInfo> MindRTBackend::ConstructGraphCompilerInfo(con | |||
| std::vector<AnfNodePtr> call_nodes; | |||
| size_t call_output_num = runtime::FetchOutputSizebyCallNode(root_output, &call_nodes); | |||
| for (size_t i = 0; i < call_output_num; ++i) { | |||
| outputs.push_back({root_output, i}); | |||
| (void)outputs.emplace_back(root_output, i); | |||
| } | |||
| } | |||
| outputs_num = outputs.size(); | |||
| @@ -952,7 +952,7 @@ std::unique_ptr<GraphCompilerInfo> MindRTBackend::ConstructGraphCompilerInfo(con | |||
| if (outputs_order.count(output) == 0) { | |||
| outputs_order[output] = {position++}; | |||
| } else { | |||
| outputs_order[output].emplace_back(position++); | |||
| (void)outputs_order[output].emplace_back(position++); | |||
| } | |||
| } | |||
| @@ -983,7 +983,7 @@ std::unique_ptr<GraphCompilerInfo> MindRTBackend::ConstructGraphCompilerInfo( | |||
| if (outputs_order.count(output) == 0) { | |||
| outputs_order[output] = {position++}; | |||
| } else { | |||
| outputs_order[output].emplace_back(position++); | |||
| (void)outputs_order[output].emplace_back(position++); | |||
| } | |||
| } | |||
| } | |||
| @@ -1021,7 +1021,7 @@ void MindRTBackend::RunGraph(const ActorInfo &actor_info, OpRunInfo *op_run_info | |||
| } | |||
| for (size_t index = 0; index < tensors_mask->size(); ++index) { | |||
| if (tensors_mask->at(index) != kValueNodeTensorMask) { | |||
| tensors_without_value_node.emplace_back(input_tensors->at(index)); | |||
| (void)tensors_without_value_node.emplace_back(input_tensors->at(index)); | |||
| } | |||
| } | |||