History log of /frameworks/ml/nn/runtime/test/benchmark/Android.mk
Revision Date Author Comments (<<< Hide modified files) (Show modified files >>>)
8cc3d6a51a4729ef63076e7230518f6c01bf824b 29-Mar-2018 jdesprez <jdesprez@google.com> Add NeuralNetworksApiBenchmark to apct suite metric

Add that module to the suite version APCT that will
eventually replace the current execution model.

Test: build device-tests.zip
Bug: 77280599
Change-Id: I4195ed94c1d78c101c870ae25d0ec84381c44325
/frameworks/ml/nn/runtime/test/benchmark/Android.mk
7a2ef1ba8a81fc3598ee7822b4f67fe8770e380c 24-Jan-2018 Jiyong Park <jiyong@google.com> Use stubs libs for android.test.base/runner/mock

android.test.base/runner/mock libraries are shared libraries that are
built without SDK, but with framework interanl libraries (framework.jar,
etc). Apps or libs that are built with SDK (LOCAL_SDK_VERSION is set)
shouldn't use them, but instead the stubs version of the test framework
libraries, which are built using SDK.

This is because, the build system does not allow a module built with SDK
to depend on another module built without SDK. This has been
exceptionally allowed with warnings, but will soon be errors.

In order to prevent the expected build breakage, direct reference to the
shared libraries are all changed to the stubs libraries.

Note: no one has been using these shared libs statically. So, it is
guaranteed that they won't encounter a problem due to missing
<uses-library name="android.test.runner"/> even after this change.

Bug: 69899800
Test: m -j checkbuild
Change-Id: Ib48252e226a148b311433da760041011b9aa6b04
/frameworks/ml/nn/runtime/test/benchmark/Android.mk
afb5283bc0e3b23dbf94d465debb49d624c9848e 14-Dec-2017 Miao Wang <miaowang@google.com> Rename the package name to be NeuralNetworksApiBenchmark

Test: mm
Change-Id: Id30aeae80e22acd10232e85d1a278ad4f73b71e8
/frameworks/ml/nn/runtime/test/benchmark/Android.mk
082ac6c7e0735b40e2e27f668b294b5f90b2e29b 09-Dec-2017 Miao Wang <miaowang@google.com> Create a benchmark app for NN API

- The benchmark uses TFlite to test NN API performance.
- The benchmark currently only covers mobilenet, both float and
quantized 8bit versions.
- The models to be tested are in assets/.
- The tests are invokable with "adb shell am instrument -w
com.example.android.nn.benchmark/android.support.test.runner.AndroidJUnitRunner"

Test: mm
Change-Id: I8ab408493aeafb22a243d19c9aa7cd09bb364c34
/frameworks/ml/nn/runtime/test/benchmark/Android.mk