1 /* Copyright 2019 The TensorFlow Authors. All Rights Reserved. 2 3 Licensed under the Apache License, Version 2.0 (the "License"); 4 you may not use this file except in compliance with the License. 5 You may obtain a copy of the License at 6 7 http://www.apache.org/licenses/LICENSE-2.0 8 9 Unless required by applicable law or agreed to in writing, software 10 distributed under the License is distributed on an "AS IS" BASIS, 11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 12 See the License for the specific language governing permissions and 13 limitations under the License. 14 ==============================================================================*/ 15 16 #ifndef TENSORFLOW_LITE_DELEGATES_GPU_CL_ENVIRONMENT_H_ 17 #define TENSORFLOW_LITE_DELEGATES_GPU_CL_ENVIRONMENT_H_ 18 19 #include "tensorflow/lite/delegates/gpu/cl/cl_command_queue.h" 20 #include "tensorflow/lite/delegates/gpu/cl/cl_context.h" 21 #include "tensorflow/lite/delegates/gpu/cl/cl_device.h" 22 #include "tensorflow/lite/delegates/gpu/cl/program_cache.h" 23 #include "tensorflow/lite/delegates/gpu/common/data_type.h" 24 #include "tensorflow/lite/delegates/gpu/common/gpu_info.h" 25 #include "tensorflow/lite/delegates/gpu/common/precision.h" 26 #include "tensorflow/lite/delegates/gpu/common/status.h" 27 #include "tensorflow/lite/delegates/gpu/common/task/tensor_desc.h" 28 #include "tensorflow/lite/delegates/gpu/common/tensor.h" 29 30 namespace tflite { 31 namespace gpu { 32 namespace cl { 33 34 class Environment { 35 public: 36 Environment() = default; 37 explicit Environment(CLDevice&& device, CLContext&& context, 38 CLCommandQueue&& queue, 39 ProfilingCommandQueue&& profiling_queue); 40 // Move only 41 Environment(Environment&& environment); 42 Environment& operator=(Environment&& environment); 43 Environment(const Environment&) = delete; 44 Environment& operator=(const Environment&) = delete; 45 device()46 const CLDevice& device() const { return device_; } GetDevicePtr()47 CLDevice* GetDevicePtr() { return &device_; } GetDevicePtr()48 const CLDevice* GetDevicePtr() const { return &device_; } context()49 CLContext& context() { return context_; } queue()50 CLCommandQueue* queue() { return &queue_; } profiling_queue()51 ProfilingCommandQueue* profiling_queue() { return &profiling_queue_; } program_cache()52 ProgramCache* program_cache() { return &program_cache_; } program_cache()53 const ProgramCache* program_cache() const { return &program_cache_; } 54 55 std::vector<CalculationsPrecision> GetSupportedPrecisions() const; 56 bool IsSupported(CalculationsPrecision precision) const; 57 std::vector<TensorStorageType> GetSupportedStorages() const; 58 // returns storage types that support zero clamping when reading OOB in HW 59 // (Height/Width) dimensions. 60 std::vector<TensorStorageType> GetSupportedStoragesWithHWZeroClampSupport() 61 const; 62 bool IsSupported(TensorStorageType storage_type) const; 63 64 absl::Status Init(); 65 66 void SetHighPerformance() const; 67 void SetDefaultPerformance() const; 68 void SetLowPerformance() const; // for energy saving 69 70 private: 71 CLDevice device_; 72 CLContext context_; 73 CLCommandQueue queue_; 74 ProfilingCommandQueue profiling_queue_; 75 ProgramCache program_cache_; 76 }; 77 78 TensorStorageType GetFastestStorageType(const GpuInfo& gpu_info); 79 TensorStorageType GetStorageTypeWithMinimalMemoryConsumption( 80 const GpuInfo& gpu_info); 81 82 absl::Status CreateEnvironment(Environment* result); 83 84 } // namespace cl 85 } // namespace gpu 86 } // namespace tflite 87 88 #endif // TENSORFLOW_LITE_DELEGATES_GPU_CL_ENVIRONMENT_H_ 89