Clarify that configs with the same group should differ only
by refresh rate.
Test: n/a
Merged-In: I2b09689825fc9564665324ff18eb1907765b9016
Change-Id: I2b09689825fc9564665324ff18eb1907765b9016
The scale and zeroPoint fields are applicable to other types since 1.2.
Also makes some whitespaces changes due to the generated documentation
getting out of sync with the template in frameworks/ml/nn.
Fix: 160406237
Test: generate_api.sh
Test: m
Change-Id: Icf594d40c73ff8c05044c320ac9eb6a9c5a89754
Merged-In: Icf594d40c73ff8c05044c320ac9eb6a9c5a89754
(cherry picked from commit 0d6cefe90b)
This change also removes some non-NN hashes in current.txt that were
unintentionally introduced in https://r.android.com/1178665.
Bug: 156918813
Bug: 158557728
Test: m
Change-Id: I4c2b83e11b27b791b3aa624be68474011b92a269
Merged-In: I4c2b83e11b27b791b3aa624be68474011b92a269
(cherry picked from commit 709842ab98)
The new framework implementation derives capture buffer loss
notification from other information, and treats HAL notify() with
ERROR_BUFFER as no-op.
Test: Build
Bug: 155353799
Change-Id: Ia7ea52ee2750c7404b657467e1cfda4c05e6cc78
- Add documentation about giving hints to camera HAL about concurrent
operation before configuring sessions
- Limitations on ZOOM_RATIO_RANGE during concurrent operation.
Bug: 151891611
Test: builds
Change-Id: I79d0bd8be1ed9aa29a1ba080d01833f16eff580c
Signed-off-by: Jayant Chowdhary <jchowdhary@google.com>
Clarify behaviour of the following ops in the case when inputs would
cause an output to have an empty shape:
* MEAN
* ARGMIN/ARGMAX
* STRIDED_SLICE
Bug: 155508675
Bug: 155660285
Bug: 155508675
Bug: 155238914
Test: VtsHalNeuralnetworksV1_3TargetTest
Change-Id: I92d1c3866a462a99b4eed7782b72f62bb6a076c6
Merged-In: I92d1c3866a462a99b4eed7782b72f62bb6a076c6
(cherry picked from commit d4bba83f6a)
Add a clarification on the behaviour of REDUCE_* ops when all dimensions
of an input tensor are reduced.
Bug: 155508675
Test: VtsHalNeuralnetworksV1_3TargetTest
Change-Id: I3a67225c3af80af40a11362ca5d0f55350cbfa05
Merged-In: I3a67225c3af80af40a11362ca5d0f55350cbfa05
(cherry picked from commit eec7d86b44)
Existing definition was inconsistent, so update it to be consistent
and match what implementations have actually done.
Test: Builds
Bug: 150331548
Change-Id: I5dfbe7c802bf4ac98f0c5556d5ac3af5d79cff7a
Clarify behaviour of the following ops in the case when inputs would
cause an output to have an empty shape:
* MEAN
* ARGMIN/ARGMAX
* STRIDED_SLICE
Bug: 155508675
Bug: 155660285
Bug: 155508675
Bug: 155238914
Test: VtsHalNeuralnetworksV1_3TargetTest
Change-Id: I92d1c3866a462a99b4eed7782b72f62bb6a076c6
Add a clarification on the behaviour of REDUCE_* ops when all dimensions
of an input tensor are reduced.
Bug: 155508675
Test: VtsHalNeuralnetworksV1_3TargetTest
Change-Id: I3a67225c3af80af40a11362ca5d0f55350cbfa05
This was caught by R process, but due to a bug in hidl-gen, not
enforced.
Bug: 155593324
Test: build checks hash
Merged-In: Ied59b4ba94c8b6bb9e3ead869d8341cffa86f1e8
Change-Id: I2bf135bb10ad052bcfe4d8f2035ed8368d5ea9ac
Update comments only for IStreamOutEventCallback::onCodecFormatChanged
Test: compiles, this is a comment change only.
Bug: 150175043
Change-Id: I1895c737997fbf3cb509b4078b545658d11df198
Also removed the hash check on Vehicle HAL files, as they're still being
worked on.
Test: m
Bug: 15249991
Change-Id: I214ebc9b5bbd71e5db1a1332296ceb4e426c50cf
