Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 0e2ac1bb authored by Michael Butler's avatar Michael Butler Committed by David Gross
Browse files

NNAPI Concurrent Query Management -- HAL and VTS update 

The NNAPI requires requests on a model to be asynchronously
processed. This CL implements a basic Event that can later
be used to block the runtime thread until the asynchronous
request has completed.

Bug: 63905942
Test: VtsHalNeuralnetworksV1_0TargetTest (32-bit, 64-bit) with sample driver enabled by cherry-pick
      frameworks/ml/nn/runtime/test with and without sample driver enabled
Change-Id: Ie27a574aaaac312e7cbb731750f9c06278357a1c
parent 5d143d2a

neuralnetworks/1.0/Android.bp

+7 −0
Original line number Diff line number Diff line
@@ -5,6 +5,7 @@ filegroup { 
    srcs: [

        "types.hal",

        "IDevice.hal",

        "IEvent.hal",

        "IPreparedModel.hal",

    ],

}
@@ -19,6 +20,7 @@ genrule { 
    out: [

        "android/hardware/neuralnetworks/1.0/types.cpp",

        "android/hardware/neuralnetworks/1.0/DeviceAll.cpp",

        "android/hardware/neuralnetworks/1.0/EventAll.cpp",

        "android/hardware/neuralnetworks/1.0/PreparedModelAll.cpp",

    ],

}
@@ -38,6 +40,11 @@ genrule { 
        "android/hardware/neuralnetworks/1.0/BnHwDevice.h",

        "android/hardware/neuralnetworks/1.0/BpHwDevice.h",

        "android/hardware/neuralnetworks/1.0/BsDevice.h",

        "android/hardware/neuralnetworks/1.0/IEvent.h",

        "android/hardware/neuralnetworks/1.0/IHwEvent.h",

        "android/hardware/neuralnetworks/1.0/BnHwEvent.h",

        "android/hardware/neuralnetworks/1.0/BpHwEvent.h",

        "android/hardware/neuralnetworks/1.0/BsEvent.h",

        "android/hardware/neuralnetworks/1.0/IPreparedModel.h",

        "android/hardware/neuralnetworks/1.0/IHwPreparedModel.h",

        "android/hardware/neuralnetworks/1.0/BnHwPreparedModel.h",

neuralnetworks/1.0/IEvent.hal

0 → 100644
+49 −0
Original line number Diff line number Diff line 
/*

 * Copyright (C) 2017 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.

 */



/* This HAL is a work in progress */



package android.hardware.neuralnetworks@1.0;



/**

 * The IEvent interface is a callback object passed by the

 * Neuralnetworks runtime to the vendor service. It is used as a

 * synchronization primitive between one or more runtime threads and a

 * single asynchronous vendor thread.  An event object is passed as an

 * argument to a HIDL call that is expected to take a non-trivial

 * amount of time. When the asynchronous execution thread has

 * completed its computation, it must call "notify" on the event to

 * indicate to the Neuralnetworks runtime whether the computation was

 * successful or not, and that the corresponding output is ready to be

 * consumed if the execution was successful.

 *

 * TODO: Mention that "notify" is also called by a runtime thread

 * during CPU fallback execution? Depends on whether the HIDL comments

 * are strictly for vendors or not.

 */

interface IEvent {



    /**

     * IEvent::notify is called by the server thread (i.e. the thread doing the

     * work) to mark the event as completed so that any threads requiring the

     * corresponding resources can continue executing.

     *

     * @param status Status of the execution associated with the Event.

     *               Should be SUCCESS or ERROR.

     */

    oneway notify(Status status);



};

neuralnetworks/1.0/IPreparedModel.hal

+3 −2
Original line number Diff line number Diff line
@@ -18,8 +18,9 @@ 


package android.hardware.neuralnetworks@1.0;



import IEvent;



interface IPreparedModel {

    // TODO: The execution is synchronous.  Change that to have a callback on completion.

    // Multiple threads can call this execute function concurrently.

    execute(Request request) generates(bool success);

    execute(Request request, IEvent event) generates(bool success);

};

neuralnetworks/1.0/types.hal

+5 −0
Original line number Diff line number Diff line
@@ -181,3 +181,8 @@ struct Request { 
    vec<InputOutputInfo> outputs;

    vec<memory> pools;

};



enum Status : uint32_t {

    SUCCESS,

    ERROR,

};

neuralnetworks/1.0/vts/functional/Android.bp

+4 −1
Original line number Diff line number Diff line
@@ -16,7 +16,10 @@ 


cc_test {

    name: "VtsHalNeuralnetworksV1_0TargetTest",

    srcs: ["VtsHalNeuralnetworksV1_0TargetTest.cpp"],

    srcs: [

        "Event.cpp",

        "VtsHalNeuralnetworksV1_0TargetTest.cpp",

    ],

    defaults: ["VtsHalTargetTestDefaults"],

    static_libs: [

        "android.hardware.neuralnetworks@1.0",