You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

168 lines
9.9 KiB

/*
* Copyright (C) 2021 The Android Open Source Project
*
* 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 ANDROID_PACKAGES_MODULES_NEURALNETWORKS_COMMON_NNAPI_IEXECUTION_H
#define ANDROID_PACKAGES_MODULES_NEURALNETWORKS_COMMON_NNAPI_IEXECUTION_H
#include <utility>
#include <vector>
#include "nnapi/Types.h"
namespace android::nn {
/**
* IExecution describes a reusable execution with request and several other execution configurations
* fixed. It is used to launch computations.
*
* IExecution manages a sequence of computations of the same prepared model with the same request
* and configurations. An IExecution object is used to control a set of computation, and to preserve
* resources between computations, enabling computations to have lower overhead. IExecution objects
* enable some optimizations:
* (1) An IExecution object can preserve resources between computations. For example, a driver can
* map a memory object when the IExecution object is created and cache the mapping for reuse in
* subsequent computations. Any cached resource can be released when the IExecution object is
* destroyed.
* (2) An IExecution object may be used for at most one computation at a time. This enables any
* transient computation resources such as intermediate tensors to be allocated once when the
* IExecution object is created and freed when the IExecution object is destroyed.
* (3) An IExecution object is created for a fixed request. This enables the implementation to apply
* request-specific optimizations. For example, an implementation can avoid request validation
* and conversions when the IExecution object is reused. An implementation may also choose to
* specialize the dynamic tensor shapes in a prepared model according to the request.
*
* This interface is thread-safe, and any class that implements this interface must be thread-safe.
*/
class IExecution {
public:
/**
* Performs a synchronous computation on a reusable execution.
*
* At most one computation may occur on a execution object at any given time, either by means of
* IExecution::compute or IExecution::computeFenced.
*
* The computation is performed synchronously with respect to the caller.
* IExecution::compute must verify the inputs to the function are correct. If there is an
* error, IExecution::compute must immediately return {@link ErrorStatus::INVALID_ARGUMENT}
* as a ExecutionError. If the inputs to the function are valid and there is no error,
* IExecution::compute must perform the computation, and must not return until the computation
* is complete.
*
* The caller must not change the content of any data object referenced by the request provided
* in IPreparedModel::createReusableExecution (described by the {@link DataLocation} of a {@link
* RequestArgument}) until IExecution::compute returns. IExecution::compute must not change the
* content of any of the data objects corresponding to request inputs.
*
* If the prepared model that the execution is created from was prepared from a model wherein
* all tensor operands have fully specified dimensions, and the inputs to the function are
* valid, and at computation time every operation's input operands have legal values, then the
* computation should complete successfully. There must be no failure unless the device itself
* is in a bad state.
*
* IExecution::compute may be called with an optional deadline. If the computation is not
* able to be completed before the provided deadline, the computation may be aborted, and either
* {@link ErrorStatus::MISSED_DEADLINE_TRANSIENT} or {@link
* ErrorStatus::MISSED_DEADLINE_PERSISTENT} may be returned as a ExecutionError.
*
* @param deadline Optional time point. If provided, compute is expected to complete by this
* time point. If it is not able to be completed by the deadline, the computation may be
* aborted.
* @return A pair consisting of:
* - A list of shape information of model output operands. The index into "outputShapes"
* corresponds to the index of the output operand in the Request outputs vector.
* outputShapes must be empty unless the computation is successful or the ExecutionResult
* is {@link ErrorStatus::OUTPUT_INSUFFICIENT_SIZE}. outputShapes may be empty if the
* computation is successful and all model output operands are fully-specified at
* computation time. outputShapes must have the same number of elements as the number of
* model output operands if the ExecutionResult is
* {@link ErrorStatus::OUTPUT_INSUFFICIENT_SIZE}, or if the computation is successful and
* the model has at least one output operand that is not fully-specified.
* - Duration of computation. Unless measure is YES and the computation is successful, all
* times must be reported as UINT64_MAX. A driver may choose to report any time as
* UINT64_MAX, indicating that measurement is not available.
*/
virtual ExecutionResult<std::pair<std::vector<OutputShape>, Timing>> compute(
const OptionalTimePoint& deadline) const = 0;
/**
* Launch a fenced asynchronous computation on a reusabled execution.
*
* At most one computation may occur on a execution object at any given time, either by means of
* IExecution::compute or IExecution::computeFenced.
*
* The computation is performed asynchronously with respect to the caller.
* IExecution::computeFenced must verify its inputs are correct, and the usages of memory
* pools allocated by IDevice::allocate are valid. If there is an error,
* IExecution::computeFenced must immediately return {@link ErrorStatus::INVALID_ARGUMENT}
* as a GeneralError. If the inputs to the function are valid and there is no error,
* IExecution::computeFenced must dispatch an asynchronous task to perform the computation in
* the background, and immediately return with a sync fence that will be signaled once the
* computation is completed and a callback that can be used by the client to query the duration
* and runtime error status. If the task has finished before the call returns, an empty handle
* may be returned for syncFence. The computation must wait for all the sync fences (if any) in
* waitFor to be signaled before starting the actual computation.
*
* When the asynchronous task has finished its computation, it must immediately signal the
* syncFence returned from the IExecution::computeFenced call. After the syncFence is
* signaled, the task must not modify the content of any data object referenced by request
* (described by the {@link DataLocation} of a {@link Request::Argument}).
*
* IExecution::computeFenced may be called with an optional deadline and an optional
* duration. If the computation is not able to be completed before the provided deadline or
* within the timeout duration (measured from when all sync fences in waitFor are signaled),
* whichever comes earlier, the computation may be aborted, and either {@link
* ErrorStatus::MISSED_DEADLINE_TRANSIENT} or {@link ErrorStatus::MISSED_DEADLINE_PERSISTENT}
* may be returned as an GeneralError. The error due to an abort must be sent the same way as
* other errors, described above.
*
* If any of the sync fences in waitFor changes to error status after the
* IExecution::computeFenced call succeeds, or the computation is aborted because it cannot
* finish before the deadline has been reached or the duration has elapsed, the driver must
* immediately set the returned syncFence to error status.
*
* @param waitFor A vector of sync fence file descriptors. The computation must wait for all
* sync fence to be signaled before starting the task.
* @param deadline The time by which computation is expected to complete. If the computation
* cannot be finished by the deadline, the computation may be aborted.
* @param timeoutDurationAfterFence The timeout duration within which the computation is
* expected to complete after all sync fences in waitFor are signaled.
* @return A pair consisting of:
* - A syncFence that will be triggered when the task is completed. The syncFence will be
* set to error if critical error occurs when doing actual evaluation.
* - A callback can be used to query information like duration and detailed runtime error
* status when the task is completed.
*/
virtual GeneralResult<std::pair<SyncFence, ExecuteFencedInfoCallback>> computeFenced(
const std::vector<SyncFence>& waitFor, const OptionalTimePoint& deadline,
const OptionalDuration& timeoutDurationAfterFence) const = 0;
// Public virtual destructor to allow objects to be stored (and destroyed) as smart pointers.
// E.g., std::unique_ptr<IExecution>.
virtual ~IExecution() = default;
protected:
// Protect the non-destructor special member functions to prevent object slicing.
IExecution() = default;
IExecution(const IExecution&) = default;
IExecution(IExecution&&) noexcept = default;
IExecution& operator=(const IExecution&) = default;
IExecution& operator=(IExecution&&) noexcept = default;
};
} // namespace android::nn
#endif // ANDROID_PACKAGES_MODULES_NEURALNETWORKS_COMMON_NNAPI_IEXECUTION_H