STT-tensorflow/tensorflow/lite/delegates/gpu/metal_delegate.h
Taehee Jeong 115ca425e4 Integer quantized model support for Metal delegate
The new feature is guarded by enable_quantization flag.

PiperOrigin-RevId: 315139125
Change-Id: I944008fcb914e31bd9e54634f85c6b2e0be0b8fe
2020-06-07 01:18:52 -07:00

81 lines
2.7 KiB
C

/* Copyright 2019 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#ifndef TENSORFLOW_LITE_DELEGATES_GPU_METAL_DELEGATE_H_
#define TENSORFLOW_LITE_DELEGATES_GPU_METAL_DELEGATE_H_
#ifdef SWIG
#define TFL_CAPI_EXPORT
#else
#if defined(_WIN32)
#ifdef TFL_COMPILE_LIBRARY
#define TFL_CAPI_EXPORT __declspec(dllexport)
#else
#define TFL_CAPI_EXPORT __declspec(dllimport)
#endif // TFL_COMPILE_LIBRARY
#else
#define TFL_CAPI_EXPORT __attribute__((visibility("default")))
#endif // _WIN32
#endif // SWIG
#ifdef __cplusplus
extern "C" {
#else
// For "C" 'bool' is not built-in type.
#include <stdbool.h>
#endif // __cplusplus
typedef struct TfLiteDelegate TfLiteDelegate;
typedef enum {
// waitUntilCompleted
TFLGpuDelegateWaitTypePassive,
// Minimize latency. It uses active spinning instead of mutex and consumes
// additional CPU resources.
TFLGpuDelegateWaitTypeActive,
// Useful when the output is used with GPU pipeline then or if external
// command encoder is set.
TFLGpuDelegateWaitTypeDoNotWait,
// Tries to avoid GPU sleep mode.
TFLGpuDelegateWaitTypeAggressive,
} TFLGpuDelegateWaitType;
// Creates a new delegate instance that need to be destroyed with
// DeleteFlowDelegate when delegate is no longer used by tflite.
typedef struct {
// Allows to quantify tensors, downcast values, process in float16 etc.
bool allow_precision_loss;
TFLGpuDelegateWaitType wait_type;
// Allows execution of integer quantized models
bool enable_quantization;
} TFLGpuDelegateOptions;
// Creates a new delegate instance that need to be destroyed with
// `TFLDeleteTfLiteGpuDelegate` when delegate is no longer used by TFLite.
// When `options` is set to `nullptr`, the following default values are used:
// .precision_loss_allowed = false,
// .wait_type = kPassive,
TFL_CAPI_EXPORT extern TfLiteDelegate* TFLGpuDelegateCreate(
const TFLGpuDelegateOptions* options);
// Destroys a delegate created with `TFLGpuDelegateCreate` call.
TFL_CAPI_EXPORT extern void TFLGpuDelegateDelete(TfLiteDelegate* delegate);
#ifdef __cplusplus
} // extern "C"
#endif // __cplusplus
#endif // TENSORFLOW_LITE_DELEGATES_GPU_METAL_DELEGATE_H_