!7342 [MS][LITE][DEVELOP] add DeviceContext to Context

Merge pull request !7342 from mengyuanli/add_device_context

mindspore/lite/include/context.h

@@ -37,13 +37,35 @@ typedef enum {
   DT_NPU  /**< NPU device type, not supported yet */
 } DeviceType;
 
-/// \brief Context defined for holding environment variables during runtime.
-struct Context {
+/// \brief CpuDeviceInfo defined for CPU's configuration information.
+typedef struct {
   bool enable_float16_ = false; /**< prior enable float16 inference */
+  CpuBindMode cpu_bind_mode_ = MID_CPU;
+} CpuDeviceInfo;
+
+/// \brief GpuDeviceInfo defined for GPU's configuration information.
+typedef struct {
+  bool enable_float16_ = false; /**< prior enable float16 inference */
+} GpuDeviceInfo;
+
+/// \brief DeviceInfo defined for backend's configuration information.
+union DeviceInfo {
+  CpuDeviceInfo cpu_device_info_;
+  GpuDeviceInfo gpu_device_info_;
+};
+
+/// \brief DeviceContext defined for holding backend's configuration information.
+struct DeviceContext {
   DeviceType device_type_ = DT_CPU;
+  DeviceInfo device_info_;
+};
+
+/// \brief Context defined for holding environment variables during runtime.
+struct Context {
+  std::string vendor_name_;
   int thread_num_ = 2; /**< thread number config for thread pool */
   AllocatorPtr allocator = nullptr;
-  CpuBindMode cpu_bind_mode_ = MID_CPU;
+  DeviceContextVector device_list_ = {{DT_CPU, {false, MID_CPU}}};
 };
 }  // namespace mindspore::lite
 #endif  // MINDSPORE_LITE_INCLUDE_CONTEXT_H_

mindspore/lite/include/lite_utils.h

@@ -27,7 +27,11 @@ namespace mindspore::lite {
 /// \note List public class and interface for reference.
 class Allocator;
 
+/// \brief DeviceContext defined a device context.
+struct DeviceContext;
+
 using TensorPtrVector = std::vector<mindspore::schema::Tensor *>;
+using DeviceContextVector = std::vector<DeviceContext>;
 using Uint32Vector = std::vector<uint32_t>;
 using String = std::string;
 using NodeType = schema::NodeType;

mindspore/lite/src/inner_context.cc

@@ -20,8 +20,13 @@
 
 namespace mindspore::lite {
 int InnerContext::Init() {
-  if (this->thread_pool_ == nullptr) {
-    this->thread_pool_ = CreateLiteThreadPool(this->thread_num_, this->cpu_bind_mode_);
+  if (this->device_list_.empty()) {
+    MS_LOG(ERROR) << "Device list is empty.";
+    return RET_NOT_SUPPORT;
+  }
+  if (this->thread_pool_ == nullptr && this->device_list_[0].device_type_ == DT_CPU) {
+    this->thread_pool_ =
+      CreateLiteThreadPool(this->thread_num_, this->device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_);
     if (this->thread_pool_ == nullptr) {
       MS_LOG(ERROR) << "Create ThreadPool failed";
       return RET_NULL_PTR;

mindspore/lite/src/lite_session.cc

@@ -315,14 +315,27 @@ int LiteSession::Init(Context *context) {
     return RET_ERROR;
   }
 
-  MS_ASSERT(nullptr != context);
-  if (context->device_type_ == DT_NPU) {
+  if (context == nullptr) {
+    MS_LOG(ERROR) << "context is nullptr";
+    is_running_.store(false);
+    return RET_NULL_PTR;
+  }
+
+  if (context->device_list_.empty()) {
+    MS_LOG(ERROR) << "Device list is empty.";
+    is_running_.store(false);
+    return RET_NOT_SUPPORT;
+  }
+
+  auto &device_type = context->device_list_[0].device_type_;
+
+  if (device_type == DT_NPU) {
     MS_LOG(ERROR) << "NPU is not supported.";
     is_running_.store(false);
     return RET_NOT_SUPPORT;
   }
 #ifndef SUPPORT_GPU
-  if (context->device_type_ == DT_GPU) {
+  if (device_type == DT_GPU) {
     MS_LOG(ERROR) << "GPU is not supported.";
     is_running_.store(false);
     return RET_NOT_SUPPORT;
@@ -337,9 +350,10 @@ int LiteSession::Init(Context *context) {
   }
   this->context_->allocator = context->allocator;
   this->context_->thread_num_ = context->thread_num_;
-  this->context_->cpu_bind_mode_ = context->cpu_bind_mode_;
-  this->context_->device_type_ = context->device_type_;
-  this->context_->enable_float16_ = context->enable_float16_;
+  this->context_->device_list_.clear();
+  for (auto &device_ctx : context->device_list_) {
+    this->context_->device_list_.push_back(device_ctx);
+  }
   auto ret = this->context_->Init();
   if (ret != RET_OK) {
     MS_LOG(ERROR) << "Init Context failed";
@@ -353,11 +367,12 @@ int LiteSession::Init(Context *context) {
     return ret;
   }
 #if SUPPORT_GPU
-  if (context_->device_type_ == DT_GPU) {
+  if (device_type == DT_GPU) {
+    auto gpu_device_info = this->context_->device_list_[0].device_info_.gpu_device_info_;
     auto opencl_runtime = ocl_runtime_wrap_.GetInstance();
-    opencl_runtime->SetFp16Enable(context_->enable_float16_);
+    opencl_runtime->SetFp16Enable(gpu_device_info.enable_float16_);
     if (opencl_runtime->Init() != RET_OK) {
-      context_->device_type_ = DT_CPU;
+      device_type = DT_CPU;
       MS_LOG(WARNING) << "Init OpenCL runtime failed, change to CPU mode.";
     } else {
       MS_LOG(INFO) << "Init OpenCL runtime success.";
@@ -375,9 +390,18 @@ int LiteSession::Init(Context *context) {
 }
 
 void LiteSession::BindThread(bool if_bind) {
-  if (this->context_->cpu_bind_mode_ != NO_BIND) {
+  if (this->context_->device_list_.empty()) {
+    MS_LOG(ERROR) << "Device list is empty.";
+    return;
+  }
+  auto &device_ctx = this->context_->device_list_[0];
+  if (device_ctx.device_type_ != DT_CPU) {
+    MS_LOG(ERROR) << "Device is not CPU.";
+    return;
+  }
+  if (device_ctx.device_info_.cpu_device_info_.cpu_bind_mode_ != NO_BIND) {
     MS_ASSERT(this->context_->thread_pool_ != NULL);
-    BindThreads(this->context_->thread_pool_, if_bind, this->context_->cpu_bind_mode_);
+    BindThreads(this->context_->thread_pool_, if_bind, device_ctx.device_info_.cpu_device_info_.cpu_bind_mode_);
   }
 }
 

mindspore/lite/src/scheduler.cc

@@ -192,10 +192,12 @@ void Scheduler::ConstructSubgraphs(std::vector<kernel::LiteKernel *> *kernels) {
 
   std::vector<kernel::LiteKernel *> subgraph_kernels;
   size_t sub_cnt{0};
+  auto &device_ctx = context_->device_list_[0];
   for (auto temp_kernels : sub_kernels_list) {
     std::vector<Tensor *> output_tensor = kernel::LiteKernelUtil::SubgraphOutputTensors(temp_kernels);
     for (auto tensor : output_tensor) {
-      if (context_->enable_float16_ && tensor->data_type() == kNumberTypeFloat16) {
+      if (device_ctx.device_type_ == DT_CPU && device_ctx.device_info_.cpu_device_info_.enable_float16_ &&
+          tensor->data_type() == kNumberTypeFloat16) {
         tensor->set_data_type(kNumberTypeFloat32);
       }
     }
@@ -246,8 +248,9 @@ kernel::LiteKernel *Scheduler::ScheduleNode(const std::vector<Tensor *> &in_tens
   MS_ASSERT(primitive != nullptr);
   TypeId data_type = GetFirstFp32Fp16OrInt8Type(in_tensors);
   kernel::KernelKey desc{kernel::KERNEL_ARCH::kCPU, data_type, static_cast<schema::PrimitiveType>(primitive->Type())};
+  auto &device_ctx = context_->device_list_[0];
 #if SUPPORT_GPU
-  if (context_->device_type_ == DT_GPU) {
+  if (device_ctx.device_type_ == DT_GPU) {
     desc.arch = kernel::KERNEL_ARCH::kGPU;
     auto *kernel = KernelRegistry::GetInstance()->GetKernel(in_tensors, out_tensors, primitive, context_, desc);
     if (kernel != nullptr) {
@@ -262,7 +265,8 @@ kernel::LiteKernel *Scheduler::ScheduleNode(const std::vector<Tensor *> &in_tens
 #endif
   desc.arch = kernel::KERNEL_ARCH::kCPU;
   kernel::LiteKernel *kernel = nullptr;
-  if ((context_->enable_float16_ && data_type == kNumberTypeFloat32) || data_type == kNumberTypeFloat16) {
+  if ((device_ctx.device_info_.cpu_device_info_.enable_float16_ && data_type == kNumberTypeFloat32) ||
+      data_type == kNumberTypeFloat16) {
     // check if support fp16
     kernel::KernelKey key{desc.arch, kNumberTypeFloat16, desc.type};
     kernel = KernelRegistry::GetInstance()->GetKernel(in_tensors, out_tensors, primitive, context_, key);

mindspore/lite/test/ut/src/infer_test.cc

@@ -106,8 +106,9 @@ TEST_F(InferTest, TestConvNode) {
   meta_graph.reset();
   content = nullptr;
   auto context = new lite::InnerContext;
-  context->cpu_bind_mode_ = lite::NO_BIND;
-  context->device_type_ = lite::DT_CPU;
+  auto &device_list = context->device_list_;
+  lite::DeviceContext device_ctx = {lite::DT_CPU, {false, lite::NO_BIND}};
+  device_list.push_back(device_ctx);
   context->thread_num_ = 4;
   ASSERT_EQ(lite::RET_OK, context->Init());
   auto session = session::LiteSession::CreateSession(context);
@@ -205,8 +206,9 @@ TEST_F(InferTest, TestAddNode) {
   meta_graph.reset();
   content = nullptr;
   auto context = new lite::InnerContext;
-  context->cpu_bind_mode_ = lite::NO_BIND;
-  context->device_type_ = lite::DT_CPU;
+  auto &device_list = context->device_list_;
+  lite::DeviceContext device_ctx = {lite::DT_CPU, {false, lite::NO_BIND}};
+  device_list.push_back(device_ctx);
   context->thread_num_ = 4;
   ASSERT_EQ(lite::RET_OK, context->Init());
   auto session = session::LiteSession::CreateSession(context);
@@ -295,8 +297,9 @@ TEST_F(InferTest, TestParallelExecutor) {
   meta_graph.reset();
   content = nullptr;
   auto context = new lite::InnerContext;
-  context->cpu_bind_mode_ = lite::NO_BIND;
-  context->device_type_ = lite::DT_CPU;
+  auto &device_list = context->device_list_;
+  lite::DeviceContext device_ctx = {lite::DT_CPU, {false, lite::NO_BIND}};
+  device_list.push_back(device_ctx);
   context->thread_num_ = 4;
   ASSERT_EQ(lite::RET_OK, context->Init());
   auto session = new SessionWithParallelExecutor();
@@ -336,8 +339,7 @@ TEST_F(InferTest, TestModel) {
   ASSERT_NE(nullptr, model);
   delete[] buf[0];
   auto context = new lite::InnerContext;
-  context->cpu_bind_mode_ = lite::NO_BIND;
-  context->device_type_ = lite::DT_CPU;
+  context->device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = lite::NO_BIND;
   context->thread_num_ = 4;
   ASSERT_EQ(lite::RET_OK, context->Init());
   auto session = session::LiteSession::CreateSession(context);

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32_grad/bn_grad_fp32_test.cc

@@ -68,7 +68,6 @@ TEST_F(TestBNGradFp32, BNGradFp32) {
   std::vector<lite::Tensor *> outputs = {&dx_tensor, &dscale_tensor, &dbias_tensor};
 
   lite::InnerContext ctx;
-  ctx.device_type_ = lite::DT_CPU;
   ctx.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, ctx.Init());
 
@@ -171,7 +170,6 @@ TEST_F(TestBNGradFp32, BNTtrainFp32) {
   kernel::KernelKey desc = {kernel::kCPU, TypeId::kNumberTypeFloat32, schema::PrimitiveType_FusedBatchNorm};
 
   mindspore::lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32_grad/convolution_grad_fp32_tests.cc

@@ -108,7 +108,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -182,7 +181,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32InputGrad) {
   uint64_t time_avg = 0;
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -255,7 +253,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32GroupFilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -328,7 +325,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32GroupInputGrad) {
   uint64_t time_avg = 0;
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -401,7 +397,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32GroupDilationFilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -474,7 +469,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32GroupDilationInputGrad) {
   uint64_t time_avg = 0;
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -542,7 +536,6 @@ TEST_F(TestConvolutionGradFp32, ConvGroupDilation) {
   uint64_t time_avg = 0;
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -644,7 +637,6 @@ TEST_F(TestConvolutionGradFp32, ConvFp32Dilation2Group2Stride2FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -745,7 +737,6 @@ TEST_F(TestConvolutionGradFp32, ConvGroup2Dilation2Stride2) {
   uint64_t time_avg = 0;
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32_grad/deconvolution_grad_fp32_tests.cc

@@ -90,7 +90,6 @@ TEST_F(TestDeConvolutionGradFp32, DeConvFp32FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -190,7 +189,6 @@ TEST_F(TestDeConvolutionGradFp32, DeConvFp32Dilation2FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -290,7 +288,6 @@ TEST_F(TestDeConvolutionGradFp32, DeConvFp32Dilation2Group3FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -390,7 +387,6 @@ TEST_F(TestDeConvolutionGradFp32, DeConvFp32Dilation2Group3Stride1FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -490,7 +486,6 @@ TEST_F(TestDeConvolutionGradFp32, DeConvFp32Dilation2Group2Stride2FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -590,7 +585,6 @@ TEST_F(TestDeConvolutionGradFp32, DeConvFp32Dilation2Group12Stride2FilterGrad) {
   std::vector<lite::Tensor *> outputs = {&dw_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32_grad/network_test.cc

@@ -357,8 +357,7 @@ TEST_F(NetworkTest, tuning_layer) {
   meta_graph.reset();
   content = nullptr;
   lite::Context context;
-  context.device_type_ = lite::DT_CPU;
-  context.cpu_bind_mode_ = lite::NO_BIND;
+  context.device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = lite::NO_BIND;
   context.thread_num_ = 1;
   auto session = session::TrainSession::CreateSession(&context);
   ASSERT_NE(nullptr, session);
@@ -518,8 +517,7 @@ TEST_F(NetworkTest, efficient_net) {
   auto model = lite::TrainModel::Import(buf, net_size);
   delete[] buf;
   auto context = new lite::Context;
-  context->device_type_ = lite::DT_CPU;
-  context->cpu_bind_mode_ = lite::NO_BIND;
+  context->device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = lite::NO_BIND;
   context->thread_num_ = 1;
 
   auto session = session::TrainSession::CreateSession(context);
@@ -544,8 +542,7 @@ TEST_F(NetworkTest, lenetnet) {
   auto model = lite::TrainModel::Import(buf, net_size);
   delete[] buf;
   auto context = new lite::Context;
-  context->device_type_ = lite::DT_CPU;
-  context->cpu_bind_mode_ = lite::NO_BIND;
+  context->device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = lite::NO_BIND;
   context->thread_num_ = 1;
 
   // check registration
@@ -589,8 +586,7 @@ TEST_F(NetworkTest, retina_net) {
   auto model = lite::Model::Import(buf, net_size);
   delete[] buf;
   auto context = new lite::Context;
-  context->device_type_ = lite::DT_CPU;
-  context->cpu_bind_mode_ = lite::NO_BIND;
+  context->device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = lite::NO_BIND;
   context->thread_num_ = 1;
 
   // auto session = session::TrainSession::CreateSession(context);
@@ -640,8 +636,7 @@ TEST_F(NetworkTest, mobileface_net) {
   auto model = lite::Model::Import(buf, net_size);
   delete[] buf;
   auto context = new lite::Context;
-  context->device_type_ = lite::DT_CPU;
-  context->cpu_bind_mode_ = lite::NO_BIND;
+  context->device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = lite::NO_BIND;
   context->thread_num_ = 1;
 
   // auto session = session::TrainSession::CreateSession(context);

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32_grad/pooling_grad_fp32_tests.cc

@@ -141,7 +141,6 @@ TEST_F(TestPoolingGradFp32, AvgPoolingKernelGradFp32) {
   std::vector<lite::Tensor *> outputs = {&dx_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -206,7 +205,6 @@ TEST_F(TestPoolingGradFp32, AvgPoolingBatchGradFp32) {
   std::vector<lite::Tensor *> outputs = {&dx_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -269,7 +267,6 @@ TEST_F(TestPoolingGradFp32, AvgPoolGradStride2Fp32) {
   std::vector<lite::Tensor *> outputs = {&out_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -334,7 +331,6 @@ TEST_F(TestPoolingGradFp32, AvgPoolGradStride3Fp32) {
   std::vector<lite::Tensor *> outputs = {&out_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -455,7 +451,6 @@ TEST_F(TestPoolingGradFp32, MaxPoolGradBatchFp32) {
   std::vector<lite::Tensor *> maxpool_outputs = {&out_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -530,7 +525,6 @@ TEST_F(TestPoolingGradFp32, MaxPoolGradStride2Fp32) {
   std::vector<lite::Tensor *> maxpool_outputs = {&out_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 
@@ -605,7 +599,6 @@ TEST_F(TestPoolingGradFp32, MaxPoolGradStride3Fp32) {
   std::vector<lite::Tensor *> maxpool_outputs = {&out_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32_grad/softmax_crossentropy_fp32_tests.cc

@@ -61,7 +61,6 @@ TEST_F(TestSoftmaxCrossEntropyFp32, SoftmaxCrossEntropyFp32) {
   std::vector<lite::Tensor *> outputs = {&loss_tensor, &grad_tensor};
 
   lite::InnerContext context;
-  context.device_type_ = lite::DT_CPU;
   context.thread_num_ = 1;
   ASSERT_EQ(lite::RET_OK, context.Init());
 

mindspore/lite/tools/benchmark/benchmark.cc

@@ -378,21 +378,28 @@ int Benchmark::RunBenchmark() {
     std::cerr << "New context failed while running " << model_name.c_str() << std::endl;
     return RET_ERROR;
   }
+  auto &device_ctx = context->device_list_[0];
   if (flags_->device_ == "CPU") {
-    context->device_type_ = lite::DT_CPU;
+    device_ctx.device_type_ = lite::DT_CPU;
   } else if (flags_->device_ == "GPU") {
-    context->device_type_ = lite::DT_GPU;
+    device_ctx.device_type_ = lite::DT_GPU;
   }
 
-  if (flags_->cpu_bind_mode_ == 2) {
-    context->cpu_bind_mode_ = MID_CPU;
-  } else if (flags_->cpu_bind_mode_ == 1) {
-    context->cpu_bind_mode_ = HIGHER_CPU;
-  } else {
-    context->cpu_bind_mode_ = NO_BIND;
+  if (device_ctx.device_type_ == DT_CPU) {
+    if (flags_->cpu_bind_mode_ == MID_CPU) {
+      device_ctx.device_info_.cpu_device_info_.cpu_bind_mode_ = MID_CPU;
+    } else if (flags_->cpu_bind_mode_ == HIGHER_CPU) {
+      device_ctx.device_info_.cpu_device_info_.cpu_bind_mode_ = HIGHER_CPU;
+    } else {
+      device_ctx.device_info_.cpu_device_info_.cpu_bind_mode_ = NO_BIND;
+    }
+    device_ctx.device_info_.cpu_device_info_.enable_float16_ = flags_->enable_fp16_;
+  }
+  if (device_ctx.device_type_ == DT_GPU) {
+    device_ctx.device_info_.gpu_device_info_.enable_float16_ = flags_->enable_fp16_;
   }
   context->thread_num_ = flags_->num_threads_;
-  context->enable_float16_ = flags_->enable_fp16_;
+
   session_ = session::LiteSession::CreateSession(context.get());
   if (session_ == nullptr) {
     MS_LOG(ERROR) << "CreateSession failed while running ", model_name.c_str();

mindspore/lite/tools/converter/quantizer/post_training_quantizer.cc

@@ -1369,9 +1369,7 @@ STATUS PostTrainingQuantizer::DoQuantize(FuncGraphPtr func_graph) {
   auto model = lite::Model::Import(content, size);
 
   Context ctx;
-  ctx.device_type_ = DT_CPU;
   ctx.thread_num_ = calibrator_->GetThreadNum();
-  ctx.cpu_bind_mode_ = MID_CPU;
 
   fp32_session_ = dynamic_cast<mindspore::lite::LiteSession *>(session::LiteSession::CreateSession(&ctx));
   if (fp32_session_ == nullptr) {
@@ -1452,9 +1450,8 @@ STATUS PostTrainingQuantizer::DoQuantize(FuncGraphPtr func_graph) {
     auto int8_model = lite::Model::Import(int8_content, size);
 
     Context int8_ctx;
-    int8_ctx.device_type_ = DT_CPU;
     int8_ctx.thread_num_ = calibrator_->GetThreadNum();
-    int8_ctx.cpu_bind_mode_ = HIGHER_CPU;
+    int8_ctx.device_list_[0].device_info_.cpu_device_info_.cpu_bind_mode_ = HIGHER_CPU;
 
     int8_session_ = dynamic_cast<mindspore::lite::LiteSession *>(session::LiteSession::CreateSession(&int8_ctx));
     if (int8_session_ == nullptr) {