| 1e9666208595bc251a8958155b1e41eca90b69db |
|
12-Apr-2018 |
Michael Butler <butlermichael@google.com> |
NNAPI: Add execution preference to prepareModel (runtime)
A model can be compiled in different ways to optimize for different
use-cases. This CL propagates the execution preference across the HAL so
that the NN service can better fit the users needs.
Bug: 77864669
Test: mma
Test: NeuralNetworksTest_static
Test: VtsHalNeuralnetworksV1_1TargetTest
Change-Id: I93079b18224ea10cb0a23bfbdc1ed80e302071a8
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 47b5916d2c4c5d0e0d4f6b43075a23449c16345b |
|
23-Feb-2018 |
David Gross <dgross@google.com> |
Add Capabilities::relaxedFloat32toFloat16Performance for 1.1.
Test: VtsHalNeuralnetworksV1_0TargetTest --hal_service_instance=android.hardware.neuralnetworks@1.0::IDevice/sample-all
Test: VtsHalNeuralnetworksV1_0TargetTest --hal_service_instance=android.hardware.neuralnetworks@1.0::IDevice/hvx
Test: VtsHalNeuralnetworksV1_1TargetTest --hal_service_instance=android.hardware.neuralnetworks@1.1::IDevice/sample-all
Test: NeuralNetworksTest_static
Bug: 63911257
Change-Id: I157641995e239777a0d3e56af1cf090e7309f35f
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 75886e77f9ca074173a49283b5c0a8c182d98977 |
|
23-Jan-2018 |
Michael Butler <butlermichael@google.com> |
Upgrade NeuralNetworks v1.0 to v1.1.
For Android P, the NNAPI is getting new operations. To support this
change, the NN runtime needs to handle both v1.0 and v1.1. This change
brings the NN runtime up to v1.1 and adds a VersionedIDevice class which
manages both v1.0 and v1.1 calls.
Test: mm
Test: cts and vts test binaries
Change-Id: Iaec1dcfe00c1712a0ac89954ae142d4ea126989d
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 033b8a6ce8ebd2a01ecccc6bae96d0fff8d4964e |
|
23-Sep-2017 |
Michael Butler <butlermichael@google.com> |
NNAPI Runtime: Change IEvent to explicit callbacks
IEvent was a synchronization primitive which caused some confusion
in the interface. Originally the event object was paired with an
asynchronous task, and the asynchronous task would signal this event
when the corresponding output was ready to be used.
In the case of IDevice::prepareModel, the function call would return an
IPreparedModel object that was not guaranteed to be prepared until the
runtime had returned from waiting on the corresponding event object.
The event object has been changed to two explicit callbacks--
IPreparedModelCallback and IExecutionCallback. Now,
IDevice::prepareModel no longer returns an unfinished IPreparedModel;
instead, it will pass the IPreparedModel object to the runtime through
IPreparedModelCallback::notify. When the runtime retreives the
IPreparedModel object, the asynchronous task has already finished
preparing the model.
The two callbacks are used for different purposes. Each has its own
version of notify to pass the data back to the runtime:
* IPreparedModelCallback::notify(ErrorStatus, IPreparedModel)
* IExecutionCallback::notify(ErrorStatus)
Bug: 63905942
Test: mm, vts, ml/nn/runtime/tests
Change-Id: I5cd2734a9fc1377546d55a12955445b6ccdd47b3
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 7d6ac906f000f3afe418e92c0a4ae36b2ea1143e |
|
23-Sep-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Remove operationTuple.
Removed operationTuple from Model in the HAL, as the data type was
redundant information.
Removed supportedOperationTuples from Capabilities, as real drivers need
more complex restrictions than this provided. For the OC-MR1, we'll just
rely on getSupportedNodes.
Also removed the unused cachesCompilation.
Bug: 63905942
Test: Compiled and ran tests, incl. vts.
Change-Id: I84433d1e0e5211ac2f77317f3752171d85a5ab04
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 5f916fc9a7ae95f172492bdfe5344c37beff3a6f |
|
12-Sep-2017 |
Michael Butler <butlermichael@google.com> |
Neuralnetworks HAL cleanup -- frameworks/ml/nn
Does three primary cleanups:
1) Removes unused OperandTypes, creates OEM types
2) Creates explicit ErrorStatus; functions return status
3) IDevice::initialize renamed to getCapabilities,
IDevice::getSupportedSubgraph renamed to getSupportedOperations
It makes the corresponding changes to the runtime.
Bug: 63905942
Test: mm, frameworks/ml/nn/runtime/test
Change-Id: I1545373dfa30780aaf9b61b3938be536cf122e74
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| a4e2ee81d015b5e26897d18b18dd50044faba62f |
|
10-Sep-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Refactor how arguments are tracked.
Use an explicit description of the lifetime of an operand
rather than relying on poolIndex to do the job.
Bug: 63905942
Test: Ran unit tests and test.py
Change-Id: I8ce007c439ccc99b1e04cce412ba6c93453c7a02
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 55cef2193279a646292fdeb355399627dd8990c8 |
|
09-Sep-2017 |
Miao Wang <miaowang@google.com> |
Explicitly add "_2D" to applicable ops, and add fused activation
function enum.
- Add the fused activation enum to explicitly set the expectation that
only the included ones are supported for fused ops.
- Add "_2D" to CONV, DEPTHWISE_CONV, AVG_POOL, MAX_POOL, and L2_POOL,
since they are only supposed to do 2D computation.
- Updated the tests and the stack to plumbing the changes through.
Bug: 63905942
Test: mm
Test: NeuralNetworksTests pass for both CPU path and sample hidl driver.
Change-Id: I60aeabab1d193d26d938c0d1d3ffdd519740d222
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| a0002c726a8f41886469e2d83af918472b3a5bbc |
|
07-Sep-2017 |
Miao Wang <miaowang@google.com> |
Increment HAL_NUM_OPERAND_TYPES to 12 after adding TENSOR_INT32.
- Fixes the quantized tests failures.
Bug: 63905942
Test: mm
Test: all unit test pass for both CPU and sample driver.
Change-Id: I9b0dd25cf3abc44f6cef3742f04e8957b30100bb
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 689d892203c06c66c7bb2e374462a8434e40b75f |
|
01-Sep-2017 |
Michael Butler <butlermichael@google.com> |
NNAPI Concurrent Query Management -- Implementation
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.
The design document for NN API asynchronous behavior:
https://docs.google.com/a/google.com/document/d/1mO35KK3Mnr489ZftTDnKXXnXiYxk19jZ1C4DWOqaVB4/edit?usp=sharing
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: I97b1d4cbf189176fb3b21b2cc1af09dddaff18ab
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| a67d8a0e28baef2fab57a2a3255461869abed74d |
|
30-Aug-2017 |
Miao Wang <miaowang@google.com> |
Add OEM_OPERATION enum.
Bug: 63905942
Test: mm
Test: All unit tests passing on sailfish
Change-Id: I391ca039620be6525d95ad581acdb6e11d091af2
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 15b59e721efe1a75b0e8cde8bd44bbb532446e13 |
|
12-Aug-2017 |
Yang Ni <yangni@google.com> |
Implement EMBEDDING_LOOKUP and HASHTABLE_LOOKUP
Bug: 63905942
Adapted previous TF Lite implementation and unit tests for NN API on
CPU.
Test: adb shell /data/nativetest64/embedding_lookup_test/embedding_lookup_test
Test: adb shell /data/nativetest64/hashtable_lookup_test/hashtable_lookup_test
Change-Id: I6e4a9d8c5d0d47e66e3fa771967da3526d19ac4b
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 8b99bb1d98a42b67ba1c00e12c7abb3708cf7c05 |
|
21-Aug-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Add full support for Memory pools.
Weights can now be passed via memory pools.
Test: Compiled and ran the unit test.
Bug: 63905942
Change-Id: I997a10f6102dcae7c6876c96a4f6f913eaf54c74
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 27e9be3904b034e422ee9b6ab70b35ea994d2b39 |
|
03-Aug-2017 |
Miao Wang <miaowang@google.com> |
Initial implementation of the following quantized ops.
- CONV_QUANT8
- DEPTHWISE_CONV_QUANT8
- AVERAGE_POOL_QUANT8
- MAX_POOL_QUANT8
- LOGISTIC_QUANT8
Additionally, added functions to plumb through quantization
parameters.
Bug: 63905942
Test: mm
Test: end-to-end MobileNet quantized test pass
Change-Id: Ib2753c68bf2c51467ae1c158b45541bcfdf10789
/frameworks/ml/nn/common/include/HalInterfaces.h
|
| 707dbd2d55f5dacf78ffb3ad7c8b3f37c2e9d758 |
|
25-Jul-2017 |
Jean-Luc Brouillet <jeanluc@google.com> |
Pass the model as an HIDL component.
Major rework where instead of serializing the model and
passing it in a shared memory, we pass it via HIDL.
The runtime/test code does two run, one goes through the
default CPU path, the other through any available driver.
Note: The code in hardware/../sample was written by Michael Buttler.
Bug: 63905942
Test: Runs both the CPU path and through the sample driver.
Change-Id: Ie3ee975d33056ba299895b13193f4698a690dd04
/frameworks/ml/nn/common/include/HalInterfaces.h
|