jimfly01 [Mon, 19 Nov 2018 11:00:42 +0000 (11:00 +0000)]
IVGCVSW-2115: Updated README.md
* Added line to README.md explaining the function of the get_compute_library.sh script
Change-Id: If6be1c64443dcd6d0a08a12485de3d38614542f0
jimfly01 [Thu, 15 Nov 2018 15:46:11 +0000 (15:46 +0000)]
IVGCVSW-2115: Script to clone/fetch correct clframework
* Script designed to be invoked from anywhere clones clframework
relative to where the script source is located in armnn/scripts
* Allows user to specify GITHUB_USERNAME on command line with -g option
otherwise will set this value to $USER
Change-Id: I34ee0d936152e94ff16b9bd7751c3da7248c9150
narpra01 [Fri, 16 Nov 2018 12:38:41 +0000 (12:38 +0000)]
IVGCVSW-2155 - Always call NEConvolutionLayer in NeonConvolution2dWorkload
* Remove check for preferDirectConvolution
* Remove IsNeonDirectConvolutionPreferred
* Remove unused tests and functions
Change-Id: I3f0868f41403ec5fa740889e7bdcb4415ad77bda
FrancisMurtagh [Fri, 16 Nov 2018 12:26:20 +0000 (12:26 +0000)]
IVGCVSW-2017: CLWorkload to use L2Normalization
* Changed ClL2Normalisation from using CLNormalizationLayer
to use CLL2NormalizeLayer to normalise along the channel axis
in either NCHW or NHWC format.
Change-Id: I399cbee408a277d1ef8c6c85ebcbd86d6c3e407b
Ferran Balaguer [Fri, 16 Nov 2018 11:56:01 +0000 (11:56 +0000)]
IVGCVSW-2158 BuildGuideCrossCompilation.md wrong protobuf version
Change-Id: Ibe826db4412c54aba42c45c155e7e5329861bdb1
Aron Virginas-Tar [Fri, 16 Nov 2018 11:00:48 +0000 (11:00 +0000)]
IVGCVSW-2162: Remove conditional compilation directives from ArmComputeUtils
Change-Id: Ie2f134fd51de919c50fa9d8d82bbb4ed4ee46469
Aron Virginas-Tar [Fri, 16 Nov 2018 10:26:23 +0000 (10:26 +0000)]
IVGCVSW-2157: Remove conditional compilation directives from ClBackendContext
Change-Id: I0b9818c3bc09101aa6563ea8e4a886483d3b09d1
Matteo Martincigh [Wed, 14 Nov 2018 12:39:55 +0000 (12:39 +0000)]
IVGCVSW-2092 Port LSTMCell::Eval to ArmNN
* Ported Google's LSTM implementation to RefLstmFloat32Workload
* Fixed the code throughout because of an error in the docs around the
scratch buffer size
* Updated IsLstmSupported
* Added the unit tests
!android-nn-driver:127
Change-Id: I5577b7e39ca52df1a7f102a9b437df6aa99520b6
Aron Virginas-Tar [Thu, 15 Nov 2018 17:44:36 +0000 (17:44 +0000)]
IVGCVSW-2156: Remove conditional compilation directives from ClContextControl
Change-Id: I812d09a869fe7c45ce71108fb675f3e2d974e2d9
keidav01 [Thu, 15 Nov 2018 14:43:23 +0000 (14:43 +0000)]
IVGCVSW-2134: Updating version number to 18.11
Change-Id: I7cbfd5a6ab0746e0351b64a45617ed089992c1a9
Matteo Martincigh [Thu, 15 Nov 2018 10:54:54 +0000 (10:54 +0000)]
Adedd load/unload warm-up before the memory lek test to put the runtime into a
stable test (memory-wise)
This prevents false-positives when running the tests on Master
Change-Id: I4348378f3689cb3ec98686a93315d2b88bb7bcc7
Conor Kennedy [Wed, 14 Nov 2018 15:28:28 +0000 (15:28 +0000)]
IVGCVSW-2086: Add StridedSlice layer & corresponding no-op factory implementations
Change-Id: I906dcb8b4c9b491bf3a661f208c09d0ea1c0fa35
Aron Virginas-Tar [Wed, 14 Nov 2018 18:35:18 +0000 (18:35 +0000)]
IVGCVSW-2136: Remove memory management methods from workload factories
Change-Id: Idc0f94590566ac362f7e1d1999361d025cc2f67a
David Beck [Wed, 14 Nov 2018 11:33:30 +0000 (11:33 +0000)]
IVGCVSW-2064 + IVGCVSW-2066 + IVGCVSW-2125 + IVGCVSW-2128 backend docs update
Change-Id: Iccc1272e59bd71c59f810a54d6d3742859789212
Matteo Martincigh [Wed, 14 Nov 2018 15:34:14 +0000 (15:34 +0000)]
Fixed minor typos in BuildGuideCrossCompilation.md
Change-Id: I4d02a6cca8e99d70a389ca3b0ce47fb7f15ed21f
jimfly01 [Wed, 14 Nov 2018 17:47:22 +0000 (17:47 +0000)]
IVGCVSW-2152: Convert TfLiteParser to NHWC
* Removed the convertFromTfToArmnnFormat flag from the
CreateConstTensor and CreateConstTensorImpl functions
* Removed the SwizzleInDeswizzleOut function and its invocations
in the ParseConv2D, ParseDepthwiseConv2D and ParsePool functions.
Change-Id: Id935e28ab6eff4731f838af9490d016e108f6df6
jimfly01 [Wed, 14 Nov 2018 14:52:40 +0000 (14:52 +0000)]
IVGCVSW-2125 : Consolidate backend registries into one
* adding missing include <functional> into BackendRegistry.hpp
Change-Id: I7a4ba867a856c90a3099a782ff41b8a5fb9a9efb
Éanna Ó Catháin [Wed, 14 Nov 2018 11:26:23 +0000 (11:26 +0000)]
IVGCVSW-2054: Fixing issue with InferOutputShape implementation in BatchToSpaceNdLayer.
* added Unit test
Change-Id: I80f55e8c7afb39d96006c8dd027fc9683ea8182e
jimfly01 [Mon, 12 Nov 2018 13:32:08 +0000 (13:32 +0000)]
IVGCVSW-2017: Refactored L2Normalisation Tests
* L2Normalisation Tests now take the layout as an argument.
Change-Id: Ib393889eedb3a81c9a500da91593e90a18549c33
Aron Virginas-Tar [Mon, 12 Nov 2018 18:10:43 +0000 (18:10 +0000)]
IVGCVSW-2066: Add IMemoryManager and integrate into the backends framework
Change-Id: I93223c8678165cbc3d39f461c36bb8610dc81c05
Éanna Ó Catháin [Mon, 12 Nov 2018 17:14:43 +0000 (17:14 +0000)]
Tidying up multiple issues
* Fixed error in InferOutputShape implementation
* Added better error checking to the BatchToSpace implementation.
* Added defaults to the batchToSpace descriptors.
* Changed crops to be a vector of pairs to align with the SpaceToBatch implementation
Change-Id: Ib1c16d871f0898a1caeb6629c1fee6380a773e14
David Beck [Mon, 12 Nov 2018 14:59:37 +0000 (14:59 +0000)]
IVGCVSW-2125 : Consolidate backend registries into one
Change-Id: I56da4780f8f5fcef7ff01d232d5d61bf299364bf
Éanna Ó Catháin [Mon, 12 Nov 2018 11:36:34 +0000 (11:36 +0000)]
IVGCVSW-2054: BATCH_TO_SPACE_ND Reference implementation and Unit tests.
Change-Id: I13c6728dbb60643d0e086d171225c5d802987f92
narpra01 [Fri, 9 Nov 2018 09:27:31 +0000 (09:27 +0000)]
IVGCVSW-2082 - add SimpleConvolution2dAsymmetricPaddingNhwc to CL test after ACL fix the convolution issue.
Change-Id: If72f6f4e32567bcfa4af13836e411a3199ede010
David Beck [Fri, 9 Nov 2018 14:46:40 +0000 (14:46 +0000)]
IVGCVSW-2125 : backends now can return optimizations
Change-Id: Ieec34224b433e1d2f3bbe66632cd6016cac5498c
Nattapat Chaimanowong [Fri, 9 Nov 2018 14:10:38 +0000 (14:10 +0000)]
IVGCVSW-2095 Add reference implementation and unit tests for SpaceToBatchNd
Change-Id: I27ffebdece6e68460931a44c15b9b029f9fce638
David Beck [Thu, 8 Nov 2018 09:19:14 +0000 (09:19 +0000)]
IVGCVSW-2056+IVGCVSW-2064 : move ClContextControl to the ClBackend
* add IBackendContext interface
* add ClBackendContext implementation
Change-Id: I13e4d12b73d4c7775069587675276f7cee7d630b
Aron Virginas-Tar [Wed, 31 Oct 2018 11:04:01 +0000 (11:04 +0000)]
IVGCVSW-1709: Clean up memory manager bodge
* Added quick workaround for the memory management errors in the
CL/Neon unit tests
Change-Id: I56250db462cdbdc1acc0a5824807d288fb0c1d11
David Beck [Tue, 6 Nov 2018 13:40:29 +0000 (13:40 +0000)]
IVGCVSW-2037 : Documenting pluggable backends
also in IVGCVSW-2019 + IVGCVSW-2005 + IVGCVSW-1987 IVGCVSW-1642
Change-Id: I33809fd2f30aac248ec45f070f717b8dcf5de3cf
David Beck [Tue, 30 Oct 2018 11:38:41 +0000 (11:38 +0000)]
IVGCVSW-1949 : Refactor ITensorHandle and move backend specifics to their place
Change-Id: I48242425c6a6856e13ebcee1b140cbd2af94a3aa
David Beck [Mon, 5 Nov 2018 13:40:33 +0000 (13:40 +0000)]
Revert "IVGCVSW-2056 + IVGCVSW-2064 : move ClContextControl to the ClBackend"
This reverts commit
d4dfa684941a21314b70593d01b0fc2167eebad4.
Change-Id: Id61ce69215505c3cf5d30ec2a7ec9127fb2554fc
saoste01 [Mon, 5 Nov 2018 12:30:23 +0000 (12:30 +0000)]
IVGCVSW-2113: Update ClDepthWiseConvolutionWOrkload to suit NCHW/NHWC DataLayout (3*3 Optimization)
Change-Id: Icfc137201c52e2c856076b7795572cc4ba75cc95
Aron Virginas-Tar [Thu, 1 Nov 2018 16:15:57 +0000 (16:15 +0000)]
IVGCVSW-1946: Remove armnn/src from the include paths
Change-Id: I663a0a0fccb43ee960ec070121a59df9db0bb04e
Nattapat Chaimanowong [Fri, 2 Nov 2018 10:57:25 +0000 (10:57 +0000)]
IVGCVSW-2093 Add SpaceToBatchNd layer and corresponding no-op factory implementations
Change-Id: Ibd457f3a2d4342c4d6335bd3c471282a14ab6b14
James Conroy [Thu, 1 Nov 2018 11:33:09 +0000 (11:33 +0000)]
IVGCVSW-2103: Add 2-Channel unit tests ResizeBilinear
* Modifies ResizeBilinear unit tests to use 2-Channel
tensor shapes for input and output data, to improve
test coverage when exercising NHWC data layout.
* Refactors unit tests to permute input and output
data when exercising NHWC data layout.
Change-Id: Ib7fb438cac23e78ff0104c895c3b7596bf7c3aa7
jimfly01 [Fri, 26 Oct 2018 15:50:13 +0000 (16:50 +0100)]
IVGCVSW-2074: Updated the DepthwiseConvolution2dDepthMul1 for NHWC
Change-Id: Ibaa0b909680a6f2a155e4ca6fa24e1144d6e6e73
jimfly01 [Wed, 31 Oct 2018 14:43:53 +0000 (14:43 +0000)]
IVGCVSW-2074: Updated CompareDepthwiseConvolution2dTest for NHWC
Change-Id: Ib0568024b4f43ae17e6c75ac35d31cf103ce3482
Matthew Bentham [Tue, 30 Oct 2018 17:33:19 +0000 (17:33 +0000)]
IVGCVSW-1970 TfLite VGG16: Use same number of iterations as other tests
I had forgotten that as a workaround for the testing using a different
number of test cases for different hardware backends, we repeat
the expected test results an arbitrary number of times
Change-Id: I4d1ea8daa1cbfd63ebcd7ac93ac30acb5806fafe
James Conroy [Wed, 31 Oct 2018 11:47:53 +0000 (11:47 +0000)]
IVGCVSW-2102: Fix Pooling2D CpuRef indexing bug
* Fixes bug when calcuating indexes for NHWC in
Pooling2D CpuRef implementation, it now uses
TensorBufferArrayView.
* Adds 2-Channel unit tests for Pooling2d on CpuRef,
Cl and Neon. The single channel tests were not
properly exercising Pooling2d using NHWC data layout.
* Refactors Pooling2D NHWC tests so that the input and
output data are permuted to NHWC when necessary,
instead of hard coding the data in NHWC format.
Change-Id: I5b9d41ed425ff283ea8c8ef6b1266ae0bc80f43b
jimfly01 [Fri, 26 Oct 2018 11:56:21 +0000 (12:56 +0100)]
IVGCVSW-2074: Updated the DepthwiseConvolution2dTest for NHWC
Change-Id: If2befd3bd75da70c57f924cbe18fa0a47bf7a27f
Matthew Bentham [Tue, 30 Oct 2018 12:39:50 +0000 (12:39 +0000)]
IVGCVSW-1970 TfLite VGG16: Correct number of iterations
Change-Id: Iaf03755b0d30598f5d09ca36af8eeeb369191950
Matthew Bentham [Tue, 30 Oct 2018 10:11:23 +0000 (10:11 +0000)]
IVGCVSW-1970 TfLite VGG16: Add Validation.txt for continuous integration
Change-Id: Ib7ea0a0b809bdfe1d705b31a08f42e1b7815b0d2
Matthew Bentham [Tue, 30 Oct 2018 09:42:07 +0000 (09:42 +0000)]
MLCE-67 Fix for Github issue 91
Remove incorrect early out from image resize
Change-Id: I58014680f5cbe43dfc6b4af66638db67c070a146
Matthew Bentham [Mon, 29 Oct 2018 17:39:49 +0000 (17:39 +0000)]
IVGCVSW-2029 Tweak results handling for batch size 2 test
When looking for the top probability, use the 'first' result not the 'second'.
This avoids an issue where for batched tests the classification index was reported wrongly.
Still doesn't correctly handle multiple results with the exact probabibility, or batched testing,
but it's slightly more correct than before.
Change-Id: I57d33552754667613e222d9d2037e12c87a96854
jimfly01 [Fri, 26 Oct 2018 14:55:50 +0000 (15:55 +0100)]
IVGCVSW-2074: Updated DepthwiseConvolution2dAsymmetricTest for NHWC
Change-Id: I1bc8d2f270d4abfaa62e4843b8d96bcf462c6b09
Matteo Martincigh [Mon, 29 Oct 2018 13:45:34 +0000 (13:45 +0000)]
Fixed Android32 build error blocking master
* In ImagePreprocessor, typecasted the number of image elements
(unsigned int) to a const_iterator::difference_type (int)
Change-Id: Idb8611185f6471751678a0a77d8225bceb204a34
Matteo Martincigh [Mon, 29 Oct 2018 11:01:09 +0000 (11:01 +0000)]
Fixed backend id argument parsing in ExecuteNetwork
Change-Id: I99498a633d75a2b4bdcd059627b07beaa9984890
narpra01 [Fri, 26 Oct 2018 16:36:32 +0000 (17:36 +0100)]
IVGCVSW-2083 - Add DataLayout parameter to calculate the expected output shape in ValidateTensorShapesFromInputs
* Convolution2dLayer
* DepthwiseConvolution2dLayer
* Pooling2dLayer
* ResizeBilinearLayer
* Unittests for ValidateTensorShapesFromInputs
Change-Id: I057caf8a90d822175a7dd7271f960b65c6154bb4
David Beck [Fri, 26 Oct 2018 15:03:03 +0000 (16:03 +0100)]
IVGCVSW-1896 : SubGraph selector
Change-Id: Iae9a76b10d84d7ba80136b54355f8d37d2df475d
narpra01 [Fri, 26 Oct 2018 15:24:58 +0000 (16:24 +0100)]
IVGCVSW-2069 - Implement NHWC Convolution2D for CpuRef
* Calculate index for NHWC to compute Convolution
* add more unit test for NHWC
Change-Id: I800d649b9b42be2758c445e3b3e76142888c1377
Aron Virginas-Tar [Fri, 26 Oct 2018 14:37:23 +0000 (15:37 +0100)]
IVGCVSW-2081: Move generic backend tests to the backend related CMake files
Change-Id: Ib642b0c983fed21cf38be8233bd13d86f4bf413d
Nattapat Chaimanowong [Fri, 26 Oct 2018 09:24:14 +0000 (10:24 +0100)]
IVGCVSW-2029 Fix fully connected layer support in TfLite Parser and implement test for TfLite VGG16 quantized
Change-Id: I2061f62f62684b963fa0f090718f1dcffe5c93ce
David Beck [Wed, 24 Oct 2018 16:09:46 +0000 (17:09 +0100)]
IVGCVSW-2056 + IVGCVSW-2064 : move ClContextControl to the ClBackend
Change-Id: Ice19d3f763298bc14585267df389e99df846320d
jimfly01 [Thu, 25 Oct 2018 16:05:05 +0000 (17:05 +0100)]
IVGCVSW-2074: SimpleConvolution2dTestImpl updated for NHWC
* Added a defaulted DataLayoutIndexed argument to the SimpleConvolution2dTestImpl
* Permute test data tensors just before use if necessary
* Add DataLayout to descriptor
* Added a DataLayoutIndexed argument to SimpleConvolution2d3x5TestCommon
* Added NHWC versions of the SimpleConvolution2d3x5Test for Neon and CL
Change-Id: I10e3ece42a50108baeabe4d8b0f0ac3e6d532261
Matteo Martincigh [Thu, 25 Oct 2018 15:19:34 +0000 (16:19 +0100)]
IVGCVSW-2051 Fixed the dynamic allocation of the dimSizes array in MeanLayer
Change-Id: I6e39f1e1bc32c0c2c1953da00016edd68c48d4e0
Aron Virginas-Tar [Thu, 25 Oct 2018 10:47:31 +0000 (11:47 +0100)]
IVGCVSW-2057: Remove ARMCOMPUTE(CL/NEON)_ENABLED and ARMCOMPUTENEON_ENABLED from src/armnnTfliteParser/test/ParserFlatBufferFixture.hpp
Change-Id: I85c720d37097e0c9a53dc473a7eac09f21deb847
Aron Virginas-Tar [Thu, 25 Oct 2018 10:10:49 +0000 (11:10 +0100)]
IVGCVSW-2058: Remove ARMCOMPUTE(CL/NEON)_ENABLED and ARMCOMPUTENEON_ENABLED from src/armnnUtils/ParserPrototxtFixture.hpp
Change-Id: Ief836fb80c4c966f1b76c9429d4a7fb644ceb73d
Aron Virginas-Tar [Wed, 24 Oct 2018 16:26:17 +0000 (17:26 +0100)]
IVGCVSW-2059: Remove CL-specific code from Profiler
Change-Id: I279986b83c7a38b6d395f8a496382ffaebb9ed08
Matteo Martincigh [Fri, 19 Oct 2018 15:40:03 +0000 (16:40 +0100)]
IVGCVSW-2049 + IVGCVSW-2051 Create the CL Mean Float workload and add
the unit tests
* Created the ClFloatWorkload class
* Added ClMeanValidate validation function
* Added helper function to convert the reduction axes from the ArmNN
format to ACL's
* Added workload tests
* Added some unit tests
* These changes need the CL pin to be pointing at least to revision
88d871028eeae57f9e4536d0329110eccb5e2890 (COMPMID-1574 Implement
ReduceMean in OpenCL)
!android-nn-driver:155033
Change-Id: I694fd36be0458c90e158172afde045fcc88c32ae
Aron Virginas-Tar [Wed, 24 Oct 2018 14:33:28 +0000 (15:33 +0100)]
IVGCVSW-2073: Move remaining backend-specific tests from armnn to backends
Change-Id: I45fd5b6dd32c435b78a54dc377a623e60978ce13
Mohamed Nour Abouelseoud [Wed, 24 Oct 2018 10:48:14 +0000 (11:48 +0100)]
IVGCVSW-1007 Support Inception-V3 and V4 in CaffeParser
Removed the kernel dimension/range validation from CaffeParser
Change-Id: I9e63c01fecbaea1448aa203e7212cd4aac854c3d
David Beck [Tue, 23 Oct 2018 15:09:36 +0000 (16:09 +0100)]
IVGCVSW-2071 : remove GetCompute() from the WorkloadFactory interface
Change-Id: I44a9d26d1a5d876d381aee4c6450af62811d0dbb
David Beck [Tue, 23 Oct 2018 12:35:58 +0000 (13:35 +0100)]
IVGCVSW-2067 : dynamically create workload factories based on the backends in the network
Change-Id: Ide594db8c79ff67642721d8bad47624b88621fbd
Aron Virginas-Tar [Tue, 23 Oct 2018 14:14:46 +0000 (15:14 +0100)]
IVGCVSW-2053: Validate BackendId parameters for the InferenceTest (EndToEnd tests) / ExecuteNetwork
!referencetests:154790
Change-Id: Iaf9608ae1e66460079f2546b30d7fbc9b55bb7dd
Sadik Armagan [Mon, 22 Oct 2018 08:04:18 +0000 (09:04 +0100)]
IVGCVSW-2026 + IVGCVSW-2027 Add FullyConnected Support to TfLiteParser
Change-Id: Id48f97ee33e2fd650a1ee3365ef66bdfc514a586
Derek Lamberti [Tue, 2 Oct 2018 14:52:46 +0000 (15:52 +0100)]
IVGCVSW-1823 *Free working mem only when network changes
Change-Id: I62b34713f8ebd96b9d4369f25cc8ba474aad8bb4
Aron Virginas-Tar [Mon, 22 Oct 2018 12:32:01 +0000 (13:32 +0100)]
IVGCVSW-2060: Separate and move backend specific unit tests from the src/armnn/test folder to the backends
* Moved backend-specific memory leak checking tests from RuntimeTests.cpp to
the corresponding backend test folder
Change-Id: I0a7f4ef52c5350c3cebca23b2b4e61a9446ca11f
Sadik Armagan [Mon, 22 Oct 2018 11:32:39 +0000 (12:32 +0100)]
IVGCVSW-2063 Documentation for new operations in TfLite Parser
* Updated TfLite Parser read me file for new operations added
Change-Id: Idd4dcbe03ee8cd8c65377e03b6e52f3766312b70
David Beck [Mon, 22 Oct 2018 12:16:00 +0000 (13:16 +0100)]
IVGCVSW-1991 : refactor m_SupportedComputes in DeviceSpecs
Change-Id: Ied3d54dc356f5e4f87aeb59f66423ac1f893dd01
Aron Virginas-Tar [Fri, 19 Oct 2018 15:46:15 +0000 (16:46 +0100)]
IVGCVSW-1955: Unify backend exceptions (wrap cl::Error)
* Added wrapper function around arm_compute::IFunction::run() that catches
cl::Error and wraps it into an armnn::RuntimeException
* Added MakeWorkload template inside ClWorkloadFactory that catches
cl::Error and wraps it into an armnn::RuntimeException
* Replaced cl::Error with armnn::RuntimeException in catch statements inside
LoadedNetwork
Change-Id: I2340f41ae02b8db1d7ef5157824a50e7410854e3
David Beck [Fri, 19 Oct 2018 14:20:56 +0000 (15:20 +0100)]
IVGCVSW-2019 : replace Compute enum in the backend preferences list
Change-Id: Ie7549fd27378acfa97e68d098e338b8c9d4ea5d2
Aron Virginas-Tar [Fri, 19 Oct 2018 15:58:08 +0000 (16:58 +0100)]
IVGCVSW-2061: Add ARMNN_SCOPED_PROFILING_EVENT_CL inside ClLstmFloatWorkload::Execute()
Change-Id: Ia3f895474b54e73711feb9795e7d47cbd61f13a3
James Conroy [Fri, 19 Oct 2018 09:41:35 +0000 (10:41 +0100)]
IVGCVSW-2024: Support NHWC for Pooling2D CpuRef
* Adds implementation to plumb DataLayout parameter
for Pooling2D on CpuRef.
* Adds unit tests to execute Pooling2D on CpuRef
using NHWC data layout.
* Refactors original tests to use DataLayoutIndexed
and removes duplicate code.
Change-Id: Ife7e0861a886cf58a2042e5be20e5b27af4528c9
Mohamed Nour Abouelseoud [Thu, 18 Oct 2018 11:26:19 +0000 (12:26 +0100)]
IVGCVSW-2013 Add a UInt8 Reference Implementation for the PAD Operator
Change-Id: I41f3606198db1fda8d72aaf5169594ba9156eb38
Nikhil Raj [Thu, 18 Oct 2018 13:27:50 +0000 (14:27 +0100)]
IVGCVSW-2023 CL and Neon implementation of BatchNorm with NHWC
Change-Id: I962e986607e5d045cd97b9eaeaea2f5ae624db35
David Beck [Thu, 18 Oct 2018 14:13:56 +0000 (15:13 +0100)]
IVGCVSW-2019 : replace Compute enum in the Layer object
Change-Id: I76551d511ef718eac36e5b8e5fe426ec3a402855
Aron Virginas-Tar [Thu, 18 Oct 2018 09:39:38 +0000 (10:39 +0100)]
IVGCVSW-2005: Pull in backend specific unit tests into Android.mk similarly as backend sources
Change-Id: Icafc644b64c591f4dc0e1560d02b806bf053a2da
Nattapat Chaimanowong [Wed, 17 Oct 2018 14:19:14 +0000 (15:19 +0100)]
IVGCVSW-2030 and IVGCVSW-2031 Add MaxPooling support and unit test to TfLite Parser
Change-Id: I3aea8ea6d018900682d278f28a50e40cf2f963fe
Matteo Martincigh [Thu, 18 Oct 2018 09:55:19 +0000 (10:55 +0100)]
IVGCVSW-2040 Add unit tests for the newly implemented NHWC support in
ref BatchNormalization
* Added create workload unit tests for the NHWC data layout
Change-Id: I03d66c88dc9b0340302b85012cb0152f0ec6fa72
Matteo Martincigh [Thu, 18 Oct 2018 09:33:23 +0000 (10:33 +0100)]
IVGCVSW-2033 Add unit tests for the newly implemented NHWC support in
ref Normalization
* Added create workload unit tests for the NHWC data layout
Change-Id: Id4813105c027a26e277de64fecea1845fea39fd9
Nikhil Raj [Thu, 18 Oct 2018 09:11:04 +0000 (10:11 +0100)]
IVGCVSW-1865 - Support NHWC for Convolution2D (CpuRef)
* Updated the ConvImpl.hpp to use DataLayoutIndex
* Enabled unit test for CpuRef
* Update CreateWorkload Tests for ref with NHWC
Change-Id: Id309b7ef677489d63dcb5e09bd48ab9624b5ebfb
David Beck [Wed, 17 Oct 2018 13:17:50 +0000 (14:17 +0100)]
IVGCVSW-2037 : separate registry for layer support intrefaces
* because their lifetime is different from backends
Change-Id: I9ba525e7e503b148cde1294933f3954b8bad3d68
Matteo Martincigh [Thu, 18 Oct 2018 07:45:39 +0000 (08:45 +0100)]
IVGCVSW-2032 + IVGCVSW-2033 Add NHWC implementation to the reference
Normalization workload
* Enabled NHWC support in RefNormalizationFloat32Workload for
Across normalization
* Added unit test for the reference implementation
Change-Id: I0e1f319e76491d43b83b121c273fadb5b259d1a0
Matteo Martincigh [Wed, 17 Oct 2018 13:43:29 +0000 (14:43 +0100)]
IVGCVSW-2038 + IVGCVSW-2039 + IVGCVSW-2040 Add NHWC support to the Float32 and UInt8
BatchNormalization workloads
* Enabled NHWC support in RefBatchNormalizationFloat32Workload
* Added NHWC unit tests for both FP32 and U8
* Refactored the existing unit tests
Change-Id: I6aa18f1dcc0666b80a17a7ed229cf53607bae147
Aron Virginas-Tar [Mon, 15 Oct 2018 15:08:20 +0000 (16:08 +0100)]
IVGCVSW-2012: Remove check for DataType::QuantisedAsymm8 in NeonLayerSupport
Change-Id: Ifeabb8c6aa02b5209245d863fc0d7aeceb25cb30
Matteo Martincigh [Wed, 17 Oct 2018 13:45:07 +0000 (14:45 +0100)]
Added missing pragma directive to TensorBufferArrayView.hpp
Change-Id: Ic7882cb415ccb7a899f5fb353fb96a1f19d03346
Nattapat Chaimanowong [Wed, 17 Oct 2018 10:12:19 +0000 (11:12 +0100)]
IVGCVSW-1951 Remove type templating from NeonDepthwiseConvolutionWorkload
Change-Id: I411d02949524eb802672d83ee281300c34b007c8
Nattapat Chaimanowong [Mon, 15 Oct 2018 14:07:34 +0000 (15:07 +0100)]
IVGCVSW-1951 Remove type templating from NeonConvolution2dWorkload
Change-Id: Id3d8137d60b14e93863ad5b6db752a7f6ef9b7fb
David Beck [Tue, 16 Oct 2018 11:58:49 +0000 (12:58 +0100)]
IVGCVSW-2022 : fix for backend cmake files to respect BUILD_UNIT_TESTS variable
Change-Id: I8d34419742152edd50f1ec438e32d3ccc4f38351
Kohei Takahashi [Fri, 31 Aug 2018 08:53:11 +0000 (17:53 +0900)]
Fixed unsigned to signed conversion warning
Change-Id: I6ff147e4418e8d0252bedfd1cf4a3bc3bc38ada7
Matteo Martincigh [Tue, 16 Oct 2018 15:23:33 +0000 (16:23 +0100)]
IVGCVSW-2018 Support NHWC in the current ref implementation
* Enabled the now supported ref layer tests
* Re-enabled the failing test now that the bug has been fixed in
ACL
1903a9976ae24f40cb2203364211ed62fcfbb985
* Added CreateWorkload test for ref L2Normalization NHWC
* Refactoring the ref L2Normalization for clarity
!armnn:153723
Change-Id: Id0067e49072b3e057ffe3ae3b70d928be6091c0f
Nattapat Chaimanowong [Tue, 16 Oct 2018 12:21:27 +0000 (13:21 +0100)]
IVGCVSW-1951 Update NeonWorkloadUtils
Change-Id: I147dbf6811f84ec4588264d636a36efc8ec56f72
Matteo Martincigh [Tue, 16 Oct 2018 15:17:34 +0000 (16:17 +0100)]
IVGCVSW-2018 Support NHWC in the current ref implementation
* Added NHWC support to TensorBufferArrayView class
Change-Id: I41e1d0acd226a471ec834e380389631d9236cb00
David Beck [Tue, 16 Oct 2018 10:32:20 +0000 (11:32 +0100)]
IVGCVSW-1998 : indicate deprecated functions and types in the public API
Change-Id: I60644cead5087c2012b663512fb9448dd31523af
Matteo Martincigh [Tue, 16 Oct 2018 08:45:08 +0000 (09:45 +0100)]
Restored test that was commented out when working on LSTM
Change-Id: I161a832e4e53cf014a3d57784bd2f0bffbd633b5
Aron Virginas-Tar [Mon, 15 Oct 2018 14:00:13 +0000 (15:00 +0100)]
IVGCVSW-2003: Get rid of IsLayerSupportedNeon functions in favor of ILayerSupport interface
Change-Id: I03985ff678acf9393680340638a2e1f425b9966f
Aron Virginas-Tar [Mon, 15 Oct 2018 10:47:37 +0000 (11:47 +0100)]
IVGCVSW-2002: Get rid of IsLayerSupportedCl functions in favor of ILayerSupport interface
Change-Id: Ic9172a5534eb243f3467996dd30c4400bc06224e
Mohamed Nour Abouelseoud [Fri, 12 Oct 2018 11:26:24 +0000 (12:26 +0100)]
IVGCVSW-1885 add RefPadWorkload implementation and associated unit tests
* Added RefPadWorkload implementation
* Added unit tests and applied them to CL and Ref backends
* Fixed a bug in ClPadWorkload
Change-Id: I8cb76bc9d60ae8a39b08d40f05d628e3b72f6410
Aron Virginas-Tar [Mon, 15 Oct 2018 10:11:51 +0000 (11:11 +0100)]
IVGCVSW-2004: Get rid of IsLayerSupportedRef functions in favor of ILayerSupport interface
Change-Id: Ia147a0b408b2ca951c214963432d6e0f9b27b973
David Beck [Fri, 12 Oct 2018 09:38:31 +0000 (10:38 +0100)]
IVGCVSW-1998 : replace Compute enum in LayerSupport free functions
!android-nn-driver:153490
Change-Id: I1c2a5f942e3a1c3626e093c90545ca27c64ba5e8