Blame - docs/00_introduction.dox - ml/ComputeLibrary

blob: 1d309cb80fd46d15b9a241f2caf55c378462aa5a [file] [log] [blame]

Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	1	namespace arm_compute
				2	{
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	3	/** @mainpage Introduction
				4
				5	@tableofcontents
				6
				7	The Computer Vision and Machine Learning library is a set of functions optimised for both ARM CPUs and GPUs using SIMD technologies.
				8
				9	Several builds of the library are available using various configurations:
				10	- OS: Linux, Android or bare metal.
				11	- Architecture: armv7a (32bit) or arm64-v8a (64bit)
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	12	- Technology: NEON / OpenCL / GLES_COMPUTE / NEON and OpenCL and GLES_COMPUTE
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	13	- Debug / Asserts / Release: Use a build with asserts enabled to debug your application and enable extra validation. Once you are sure your application works as expected you can switch to a release build of the library for maximum performance.
				14
				15	@section S0_1_contact Contact / Support
				16
				17	Please email developer@arm.com
				18
				19	In order to facilitate the work of the support team please provide the build information of the library you are using. To get the version of the library you are using simply run:
				20
				21	$ strings android-armv7a-cl-asserts/libarm_compute.so \| grep arm_compute_version
				22	arm_compute_version=v16.12 Build options: {'embed_kernels': '1', 'opencl': '1', 'arch': 'armv7a', 'neon': '0', 'asserts': '1', 'debug': '0', 'os': 'android', 'Werror': '1'} Git hash=f51a545d4ea12a9059fe4e598a092f1fd06dc858
				23
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	24	@section S0_2_prebuilt_binaries Pre-built binaries
				25
				26	For each release we provide some pre-built binaries of the library [here](https://github.com/ARM-software/ComputeLibrary/releases)
				27
				28	These binaries have been built using the following toolchains:
				29	- Linux armv7a: gcc-linaro-arm-linux-gnueabihf-4.9-2014.07_linux
				30	- Linux arm64-v8a: gcc-linaro-4.9-2016.02-x86_64_aarch64-linux-gnu
Anthony Barbier	a8a28f6	2018-02-26 19:16:32 +0000	[diff] [blame]	31	- Android armv7a: clang++ / gnustl NDK r16b
				32	- Android am64-v8a: clang++ / gnustl NDK r16b
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	33
				34	@warning Make sure to use a compatible toolchain to build your application or you will get some std::bad_alloc errors at runtime.
				35
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	36	@section S1_file_organisation File organisation
				37
				38	This archive contains:
				39	- The arm_compute header and source files
				40	- The latest Khronos OpenCL 1.2 C headers from the <a href="https://www.khronos.org/registry/cl/">Khronos OpenCL registry</a>
				41	- The latest Khronos cl2.hpp from the <a href="https://www.khronos.org/registry/cl/">Khronos OpenCL registry</a> (API version 2.1 when this document was written)
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	42	- The latest Khronos OpenGL ES 3.1 C headers from the <a href="https://www.khronos.org/registry/gles/">Khronos OpenGL ES registry</a>
				43	- The latest Khronos EGL 1.5 C headers from the <a href="https://www.khronos.org/registry/gles/">Khronos EGL registry</a>
				44	- The sources for a stub version of libOpenCL.so, libGLESv1_CM.so, libGLESv2.so and libEGL.so to help you build your application.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	45	- An examples folder containing a few examples to compile and link against the library.
				46	- A @ref utils folder containing headers with some boiler plate code used by the examples.
				47	- This documentation.
				48
				49	You should have the following file organisation:
				50
				51	.
				52	├── arm_compute --> All the arm_compute headers
				53	│ ├── core
				54	│ │ ├── CL
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	55	│ │ │ ├── CLKernelLibrary.h --> Manages all the OpenCL kernels compilation and caching, provides accessors for the OpenCL Context.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	56	│ │ │ ├── CLKernels.h --> Includes all the OpenCL kernels at once
				57	│ │ │ ├── CL specialisation of all the generic objects interfaces (ICLTensor, ICLImage, etc.)
				58	│ │ │ ├── kernels --> Folder containing all the OpenCL kernels
				59	│ │ │ │ └── CL*Kernel.h
				60	│ │ │ └── OpenCL.h --> Wrapper to configure the Khronos OpenCL C++ header
				61	│ │ ├── CPP
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	62	│ │ │ ├── CPPKernels.h --> Includes all the CPP kernels at once
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	63	│ │ │ └── kernels --> Folder containing all the CPP kernels
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	64	│ │ │ └── CPP*Kernel.h
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	65	│ │ ├── GLES_COMPUTE
				66	│ │ │ ├── GCKernelLibrary.h --> Manages all the GLES kernels compilation and caching, provides accessors for the GLES Context.
				67	│ │ │ ├── GCKernels.h --> Includes all the GLES kernels at once
				68	│ │ │ ├── GLES specialisation of all the generic objects interfaces (IGCTensor, IGCImage, etc.)
				69	│ │ │ ├── kernels --> Folder containing all the GLES kernels
				70	│ │ │ │ └── GC*Kernel.h
				71	│ │ │ └── OpenGLES.h --> Wrapper to configure the Khronos EGL and OpenGL ES C header
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	72	│ │ ├── NEON
				73	│ │ │ ├── kernels --> Folder containing all the NEON kernels
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	74	│ │ │ │ ├── arm64 --> Folder containing the interfaces for the assembly arm64 NEON kernels
				75	│ │ │ │ ├── arm32 --> Folder containing the interfaces for the assembly arm32 NEON kernels
				76	│ │ │ │ ├── assembly --> Folder containing the NEON assembly routines.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	77	│ │ │ │ └── NE*Kernel.h
				78	│ │ │ └── NEKernels.h --> Includes all the NEON kernels at once
				79	│ │ ├── All common basic types (Types.h, Window, Coordinates, Iterator, etc.)
				80	│ │ ├── All generic objects interfaces (ITensor, IImage, etc.)
				81	│ │ └── Objects metadata classes (ImageInfo, TensorInfo, MultiImageInfo)
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	82	│ ├── graph
				83	│ │ ├── CL --> OpenCL specific operations
				84	│ │ │ └── CLMap.h / CLUnmap.h
				85	│ │ ├── nodes
				86	│ │ │ └── The various nodes supported by the graph API
				87	│ │ ├── Nodes.h --> Includes all the Graph nodes at once.
				88	│ │ └── Graph objects ( INode, ITensorAccessor, Graph, etc.)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	89	│ └── runtime
				90	│ ├── CL
				91	│ │ ├── CL objects & allocators (CLArray, CLImage, CLTensor, etc.)
				92	│ │ ├── functions --> Folder containing all the OpenCL functions
				93	│ │ │ └── CL*.h
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	94	│ │ ├── CLScheduler.h --> Interface to enqueue OpenCL kernels and get/set the OpenCL CommandQueue and ICLTuner.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	95	│ │ └── CLFunctions.h --> Includes all the OpenCL functions at once
				96	│ ├── CPP
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	97	│ │ ├── CPPKernels.h --> Includes all the CPP functions at once.
				98	│ │ └── CPPScheduler.h --> Basic pool of threads to execute CPP/NEON code on several cores in parallel
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	99	│ ├── GLES_COMPUTE
				100	│ │ ├── GLES objects & allocators (GCArray, GCImage, GCTensor, etc.)
				101	│ │ ├── functions --> Folder containing all the GLES functions
				102	│ │ │ └── GC*.h
				103	│ │ ├── GCScheduler.h --> Interface to enqueue GLES kernels and get/set the GLES CommandQueue.
				104	│ │ └── GCFunctions.h --> Includes all the GLES functions at once
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	105	│ ├── NEON
				106	│ │ ├── functions --> Folder containing all the NEON functions
				107	│ │ │ └── NE*.h
				108	│ │ └── NEFunctions.h --> Includes all the NEON functions at once
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	109	│ ├── OMP
				110	│ │ └── OMPScheduler.h --> OpenMP scheduler (Alternative to the CPPScheduler)
				111	│ ├── Memory manager files (LifetimeManager, PoolManager, etc.)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	112	│ └── Basic implementations of the generic object interfaces (Array, Image, Tensor, etc.)
Anthony Barbier	a8a28f6	2018-02-26 19:16:32 +0000	[diff] [blame]	113	├── data -> Contains test images and reference data dumps used by validation tests
				114	├── docs -> Contains Doxyfile and Doxygen sources used to generate the HTML pages in the documentation folder.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	115	├── documentation
				116	│ ├── index.xhtml
				117	│ └── ...
				118	├── documentation.xhtml -> documentation/index.xhtml
				119	├── examples
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	120	│ ├── cl_*.cpp --> OpenCL examples
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	121	│ ├── gc_*.cpp --> GLES compute shaders examples
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	122	│ ├── graph_*.cpp --> Graph examples
				123	│ ├── neoncl_*.cpp --> NEON / OpenCL interoperability examples
				124	│ └── neon_*.cpp --> NEON examples
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	125	├── include
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	126	│ ├── CL
				127	│ │ └── Khronos OpenCL C headers and C++ wrapper
				128	│ ├── half --> FP16 library available from http://half.sourceforge.net
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	129	│ ├── libnpy --> Library to load / write npy buffers, available from https://github.com/llohse/libnpy
				130	│ └── linux --> Headers only needed for Linux builds
				131	│ └── Khronos EGL and OpenGLES headers
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	132	├── opencl-1.2-stubs
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	133	│ └── opencl_stubs.c --> OpenCL stubs implementation
				134	├── opengles-3.1-stubs
				135	│ ├── EGL.c --> EGL stubs implementation
				136	│ └── GLESv2.c --> GLESv2 stubs implementation
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	137	├── scripts
				138	│ ├── caffe_data_extractor.py --> Basic script to export weights from Caffe to npy files
				139	│ └── tensorflow_data_extractor.py --> Basic script to export weights from Tensor Flow to npy files
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	140	├── src
				141	│ ├── core
				142	│ │ └── ... (Same structure as headers)
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	143	│ │ ├── CL
				144	│ │ │ └── cl_kernels --> All the OpenCL kernels
				145	│ │ └── GLES_COMPUTE
				146	│ │ └── cs_shaders --> All the OpenGL ES Compute Shaders
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	147	│ ├── graph
				148	│ │ └── ... (Same structure as headers)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	149	│ └── runtime
				150	│ └── ... (Same structure as headers)
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	151	├── support
				152	│ └── Various headers to work around toolchains / platform issues.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	153	├── tests
				154	│ ├── All test related files shared between validation and benchmark
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	155	│ ├── CL --> OpenCL accessors
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	156	│ ├── GLES_COMPUTE --> GLES accessors
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	157	│ ├── NEON --> NEON accessors
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	158	│ ├── benchmark --> Sources for benchmarking
				159	│ │ ├── Benchmark specific files
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	160	│ │ ├── CL --> OpenCL benchmarking tests
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	161	│ │ ├── GLES_COMPUTE --> GLES benchmarking tests
Anthony Barbier	a8a28f6	2018-02-26 19:16:32 +0000	[diff] [blame]	162	│ │ ├── fixtures
				163	│ │ │ └── Fixtures to initialise and run the runtime Functions.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	164	│ │ └── NEON --> NEON benchmarking tests
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	165	│ ├── datasets
				166	│ │ └── Datasets for all the validation / benchmark tests, layer configurations for various networks, etc.
				167	│ ├── framework
				168	│ │ └── Boiler plate code for both validation and benchmark test suites (Command line parsers, instruments, output loggers, etc.)
				169	│ ├── networks
				170	│ │ └── Examples of how to instantiate networks.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	171	│ ├── validation --> Sources for validation
				172	│ │ ├── Validation specific files
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	173	│ │ ├── CL --> OpenCL validation tests
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	174	│ │ ├── GLES_COMPUTE --> GLES validation tests
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	175	│ │ ├── CPP --> C++ reference implementations
				176	│ │ ├── fixtures
				177	│ │ │ └── Fixtures to initialise and run the runtime Functions.
				178	│ │ └── NEON --> NEON validation tests
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	179	│ └── dataset --> Datasets defining common sets of input parameters
				180	└── utils --> Boiler plate code used by examples
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	181	└── Various utilities to print types, load / store assets, etc.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	182
				183	@section S2_versions_changelog Release versions and changelog
				184
				185	@subsection S2_1_versions Release versions
				186
				187	All releases are numbered vYY.MM Where YY are the last two digits of the year, and MM the month number.
				188	If there is more than one release in a month then an extra sequential number is appended at the end:
				189
				190	v17.03 (First release of March 2017)
				191	v17.03.1 (Second release of March 2017)
				192	v17.04 (First release of April 2017)
				193
				194	@note We're aiming at releasing one major public release with new features per quarter. All releases in between will only contain bug fixes.
				195
				196	@subsection S2_2_changelog Changelog
				197
Pablo Tello	eb82fd2	2018-02-23 13:43:50 +0000	[diff] [blame]	198	v18.05 Public maintenance release
				199	- Major redesign in the interface for the neon kernels implemented in assembly.
				200	- Removed arm_compute::NEGEMMLowpAArch64A53Kernel / arm_compute::NEGEMMLowpAArch64Kernel / arm_compute::NEGEMMLowpAArch64V8P4Kernel / arm_compute::NEGEMMInterleavedBlockedKernel / arm_compute::NEGEMMLowpAssemblyMatrixMultiplyCore / arm_compute::NEHGEMMAArch64FP16Kernel
				201	- Added NEGEMMAssemblyWrapper and AssemblyKernelGlue which are used to execute assembly kernels in neon functions.
				202	- Minor changes to the CPUInfo type to make it compatible with the new assembly gemm interface.
				203
Anthony Barbier	577fbdf	2018-03-01 15:17:54 +0000	[diff] [blame]	204	v18.03 Public maintenance release
				205	- Various bug fixes.
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	206	- Fixed bug in @ref NEActivationLayer
				207	- Fix in @ref CLTuner when using batches.
Anthony Barbier	577fbdf	2018-03-01 15:17:54 +0000	[diff] [blame]	208	- Updated recommended NDK version to r16b (And fixed warnings).
				209	- Fixed bug in validation code.
				210	- Added Inception v4 graph example.
Georgios Pinitas	9fb1159	2018-04-26 20:34:58 +0100	[diff] [blame]	211	- Renamed NEWinogradLayer.cpp to @ref NEWinogradConvolutionLayer
Anthony Barbier	577fbdf	2018-03-01 15:17:54 +0000	[diff] [blame]	212
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	213	v18.02 Public major release
				214	- Various NEON / OpenCL / GLES optimisations.
				215	- Various bug fixes.
				216	- Changed default number of threads on big LITTLE systems.
				217	- Refactored examples and added:
				218	- graph_mobilenet_qassym8
				219	- graph_resnet
				220	- graph_squeezenet_v1_1
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	221	- Renamed @ref CLConvolutionLayer into @ref CLGEMMConvolutionLayer and created a new @ref CLConvolutionLayer to select the fastest convolution method.
				222	- Renamed @ref NEConvolutionLayer into @ref NEGEMMConvolutionLayer and created a new @ref NEConvolutionLayer to select the fastest convolution method.
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	223	- Added in place support to:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	224	- @ref CLActivationLayer
				225	- @ref CLBatchNormalizationLayer
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	226	- Added QASYMM8 support to:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	227	- @ref CLActivationLayer
				228	- @ref CLDepthwiseConvolutionLayer
				229	- @ref NEDepthwiseConvolutionLayer
				230	- @ref NESoftmaxLayer
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	231	- Added FP16 support to:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	232	- @ref CLDepthwiseConvolutionLayer3x3
				233	- @ref CLDepthwiseConvolutionLayer
				234	- Added broadcasting support to @ref NEArithmeticAddition / @ref CLArithmeticAddition / @ref CLPixelWiseMultiplication
				235	- Added fused batched normalization and activation to @ref CLBatchNormalizationLayer and @ref NEBatchNormalizationLayer
				236	- Added support for non-square pooling to @ref NEPoolingLayer and @ref CLPoolingLayer
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	237	- New OpenCL kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	238	- @ref CLDirectConvolutionLayerOutputStageKernel
Pablo Tello	f6c572c	2018-02-14 12:47:30 +0000	[diff] [blame]	239	- New NEON kernels / functions
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	240	- Added name() method to all kernels.
				241	- Added support for Winograd 5x5.
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	242	- @ref NEPermuteKernel / @ref NEPermute
Georgios Pinitas	9fb1159	2018-04-26 20:34:58 +0100	[diff] [blame]	243	- @ref NEWinogradLayerTransformInputKernel / NEWinogradLayer
				244	- @ref NEWinogradLayerTransformOutputKernel / NEWinogradLayer
				245	- @ref NEWinogradLayerTransformWeightsKernel / NEWinogradLayer
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	246	- Renamed NEWinogradLayerKernel into @ref NEWinogradLayerBatchedGEMMKernel
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	247	- New GLES kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	248	- @ref GCTensorShiftKernel / @ref GCTensorShift
Pablo Tello	f6c572c	2018-02-14 12:47:30 +0000	[diff] [blame]	249
Anthony Barbier	64c95a0	2018-01-22 18:48:55 +0000	[diff] [blame]	250	v18.01 Public maintenance release
				251	- Various bug fixes
				252	- Added some of the missing validate() methods
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	253	- Added @ref CLDeconvolutionLayerUpsampleKernel / @ref CLDeconvolutionLayer @ref CLDeconvolutionLayerUpsample
				254	- Added @ref CLPermuteKernel / @ref CLPermute
Anthony Barbier	64c95a0	2018-01-22 18:48:55 +0000	[diff] [blame]	255	- Added method to clean the programs cache in the CL Kernel library.
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	256	- Added @ref GCArithmeticAdditionKernel / @ref GCArithmeticAddition
				257	- Added @ref GCDepthwiseConvolutionLayer3x3Kernel / @ref GCDepthwiseConvolutionLayer3x3
				258	- Added @ref GCNormalizePlanarYUVLayerKernel / @ref GCNormalizePlanarYUVLayer
				259	- Added @ref GCScaleKernel / @ref GCScale
				260	- Added @ref GCWeightsReshapeKernel / @ref GCConvolutionLayer
Anthony Barbier	64c95a0	2018-01-22 18:48:55 +0000	[diff] [blame]	261	- Added FP16 support to the following GLES compute kernels:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	262	- @ref GCCol2ImKernel
				263	- @ref GCGEMMInterleave4x4Kernel
				264	- @ref GCGEMMTranspose1xWKernel
				265	- @ref GCIm2ColKernel
				266	- Refactored NEON Winograd (NEWinogradLayerKernel)
				267	- Added @ref NEDirectConvolutionLayerOutputStageKernel
Anthony Barbier	64c95a0	2018-01-22 18:48:55 +0000	[diff] [blame]	268	- Added QASYMM8 support to the following NEON kernels:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	269	- @ref NEDepthwiseConvolutionLayer3x3Kernel
				270	- @ref NEFillBorderKernel
				271	- @ref NEPoolingLayerKernel
Anthony Barbier	64c95a0	2018-01-22 18:48:55 +0000	[diff] [blame]	272	- Added new examples:
				273	- graph_cl_mobilenet_qasymm8.cpp
				274	- graph_inception_v3.cpp
				275	- gc_dc.cpp
				276	- More tests added to both validation and benchmarking suites.
				277
Gian Marco	ff85093	2017-12-11 12:37:17 +0000	[diff] [blame]	278	v17.12 Public major release
				279	- Most machine learning functions on OpenCL support the new data type QASYMM8
				280	- Introduced logging interface
				281	- Introduced opencl timer
				282	- Reworked GEMMLowp interface
				283	- Added new NEON assembly kernels for GEMMLowp, SGEMM and HGEMM
				284	- Added validation method for most Machine Learning kernels / functions
				285	- Added new graph examples such as googlenet, mobilenet, squeezenet, vgg16 and vgg19
				286	- Added sgemm example for OpenCL
				287	- Added absolute difference example for GLES compute
				288	- Added new tests and benchmarks in validation and benchmark frameworks
				289	- Added new kernels / functions for GLES compute
				290
				291	- New OpenGL ES kernels / functions
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	292	- @ref GCAbsoluteDifferenceKernel / @ref GCAbsoluteDifference
				293	- @ref GCActivationLayerKernel / @ref GCActivationLayer
				294	- @ref GCBatchNormalizationLayerKernel / @ref GCBatchNormalizationLayer
				295	- @ref GCCol2ImKernel
				296	- @ref GCDepthConcatenateLayerKernel / @ref GCDepthConcatenateLayer
				297	- @ref GCDirectConvolutionLayerKernel / @ref GCDirectConvolutionLayer
				298	- @ref GCDropoutLayerKernel / @ref GCDropoutLayer
				299	- @ref GCFillBorderKernel / @ref GCFillBorder
				300	- @ref GCGEMMInterleave4x4Kernel / @ref GCGEMMInterleave4x4
				301	- @ref GCGEMMMatrixAccumulateBiasesKernel / @ref GCGEMMMatrixAdditionKernel / @ref GCGEMMMatrixMultiplyKernel / @ref GCGEMM
				302	- @ref GCGEMMTranspose1xWKernel / @ref GCGEMMTranspose1xW
				303	- @ref GCIm2ColKernel
				304	- @ref GCNormalizationLayerKernel / @ref GCNormalizationLayer
				305	- @ref GCPixelWiseMultiplicationKernel / @ref GCPixelWiseMultiplication
				306	- @ref GCPoolingLayerKernel / @ref GCPoolingLayer
				307	- @ref GCLogits1DMaxKernel / @ref GCLogits1DShiftExpSumKernel / @ref GCLogits1DNormKernel / @ref GCSoftmaxLayer
				308	- @ref GCTransposeKernel / @ref GCTranspose
Gian Marco	ff85093	2017-12-11 12:37:17 +0000	[diff] [blame]	309
				310	- New NEON kernels / functions
Pablo Tello	eb82fd2	2018-02-23 13:43:50 +0000	[diff] [blame]	311	- arm_compute::NEGEMMLowpAArch64A53Kernel / arm_compute::NEGEMMLowpAArch64Kernel / arm_compute::NEGEMMLowpAArch64V8P4Kernel / arm_compute::NEGEMMInterleavedBlockedKernel / arm_compute::NEGEMMLowpAssemblyMatrixMultiplyCore
				312	- arm_compute::NEHGEMMAArch64FP16Kernel
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	313	- @ref NEDepthwiseConvolutionLayer3x3Kernel / @ref NEDepthwiseIm2ColKernel / @ref NEGEMMMatrixVectorMultiplyKernel / @ref NEDepthwiseVectorToTensorKernel / @ref NEDepthwiseConvolutionLayer
				314	- @ref NEGEMMLowpOffsetContributionKernel / @ref NEGEMMLowpMatrixAReductionKernel / @ref NEGEMMLowpMatrixBReductionKernel / @ref NEGEMMLowpMatrixMultiplyCore
				315	- @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel / @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint
				316	- @ref NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel / @ref NEGEMMLowpQuantizeDownInt32ToUint8Scale
Georgios Pinitas	9fb1159	2018-04-26 20:34:58 +0100	[diff] [blame]	317	- NEWinogradLayer / NEWinogradLayerKernel
Gian Marco	ff85093	2017-12-11 12:37:17 +0000	[diff] [blame]	318
				319	- New OpenCL kernels / functions
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	320	- @ref CLGEMMLowpOffsetContributionKernel / @ref CLGEMMLowpMatrixAReductionKernel / @ref CLGEMMLowpMatrixBReductionKernel / @ref CLGEMMLowpMatrixMultiplyCore
				321	- @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel / @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint
				322	- @ref CLGEMMLowpQuantizeDownInt32ToUint8ScaleKernel / @ref CLGEMMLowpQuantizeDownInt32ToUint8Scale
Gian Marco	ff85093	2017-12-11 12:37:17 +0000	[diff] [blame]	323
				324	- New graph nodes for NEON and OpenCL
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	325	- graph::BranchLayer
				326	- graph::DepthConvertLayer
				327	- graph::DepthwiseConvolutionLayer
				328	- graph::DequantizationLayer
				329	- graph::FlattenLayer
				330	- graph::QuantizationLayer
				331	- graph::ReshapeLayer
Gian Marco	ff85093	2017-12-11 12:37:17 +0000	[diff] [blame]	332
Anthony Barbier	3c5b4ff	2017-10-12 13:20:52 +0100	[diff] [blame]	333	v17.10 Public maintenance release
				334	- Bug fixes:
				335	- Check the maximum local workgroup size supported by OpenCL devices
				336	- Minor documentation updates (Fixed instructions to build the examples)
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	337	- Introduced a graph::GraphContext
Anthony Barbier	3c5b4ff	2017-10-12 13:20:52 +0100	[diff] [blame]	338	- Added a few new Graph nodes, support for branches and grouping.
				339	- Automatically enable cl_printf in debug builds
				340	- Fixed bare metal builds for armv7a
				341	- Added AlexNet and cartoon effect examples
				342	- Fixed library builds: libraries are no longer built as supersets of each other.(It means application using the Runtime part of the library now need to link against both libarm_compute_core and libarm_compute)
				343
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	344	v17.09 Public major release
				345	- Experimental Graph support: initial implementation of a simple stream API to easily chain machine learning layers.
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	346	- Memory Manager (@ref BlobLifetimeManager, @ref BlobMemoryPool, @ref ILifetimeManager, @ref IMemoryGroup, @ref IMemoryManager, @ref IMemoryPool, @ref IPoolManager, @ref MemoryManagerOnDemand, @ref PoolManager)
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	347	- New validation and benchmark frameworks (Boost and Google frameworks replaced by homemade framework).
				348	- Most machine learning functions support both fixed point 8 and 16 bit (QS8, QS16) for both NEON and OpenCL.
				349	- New NEON kernels / functions:
Pablo Tello	eb82fd2	2018-02-23 13:43:50 +0000	[diff] [blame]	350	- arm_compute::NEGEMMAssemblyBaseKernel arm_compute::NEGEMMAArch64Kernel
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	351	- @ref NEDequantizationLayerKernel / @ref NEDequantizationLayer
				352	- @ref NEFloorKernel / @ref NEFloor
				353	- @ref NEL2NormalizeLayerKernel / @ref NEL2NormalizeLayer
				354	- @ref NEQuantizationLayerKernel @ref NEMinMaxLayerKernel / @ref NEQuantizationLayer
				355	- @ref NEROIPoolingLayerKernel / @ref NEROIPoolingLayer
				356	- @ref NEReductionOperationKernel / @ref NEReductionOperation
				357	- @ref NEReshapeLayerKernel / @ref NEReshapeLayer
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	358
				359	- New OpenCL kernels / functions:
Giorgio Arena	dfca60b	2018-01-31 10:30:59 +0000	[diff] [blame]	360	- @ref CLDepthwiseConvolutionLayer3x3NCHWKernel @ref CLDepthwiseConvolutionLayer3x3NHWCKernel @ref CLDepthwiseIm2ColKernel @ref CLDepthwiseVectorToTensorKernel @ref CLDepthwiseWeightsReshapeKernel / @ref CLDepthwiseConvolutionLayer3x3 @ref CLDepthwiseConvolutionLayer @ref CLDepthwiseSeparableConvolutionLayer
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	361	- @ref CLDequantizationLayerKernel / @ref CLDequantizationLayer
				362	- @ref CLDirectConvolutionLayerKernel / @ref CLDirectConvolutionLayer
				363	- @ref CLFlattenLayer
				364	- @ref CLFloorKernel / @ref CLFloor
				365	- @ref CLGEMMTranspose1xW
				366	- @ref CLGEMMMatrixVectorMultiplyKernel
				367	- @ref CLL2NormalizeLayerKernel / @ref CLL2NormalizeLayer
				368	- @ref CLQuantizationLayerKernel @ref CLMinMaxLayerKernel / @ref CLQuantizationLayer
				369	- @ref CLROIPoolingLayerKernel / @ref CLROIPoolingLayer
				370	- @ref CLReductionOperationKernel / @ref CLReductionOperation
				371	- @ref CLReshapeLayerKernel / @ref CLReshapeLayer
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	372
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	373	v17.06 Public major release
				374	- Various bug fixes
				375	- Added support for fixed point 8 bit (QS8) to the various NEON machine learning kernels.
				376	- Added unit tests and benchmarks (AlexNet, LeNet)
				377	- Added support for sub tensors.
				378	- Added infrastructure to provide GPU specific optimisation for some OpenCL kernels.
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	379	- Added @ref OMPScheduler (OpenMP) scheduler for NEON
				380	- Added @ref SingleThreadScheduler scheduler for NEON (For bare metal)
				381	- User can specify his own scheduler by implementing the @ref IScheduler interface.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	382	- New OpenCL kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	383	- @ref CLBatchNormalizationLayerKernel / @ref CLBatchNormalizationLayer
				384	- @ref CLDepthConcatenateLayerKernel / @ref CLDepthConcatenateLayer
				385	- @ref CLHOGOrientationBinningKernel @ref CLHOGBlockNormalizationKernel, @ref CLHOGDetectorKernel / @ref CLHOGDescriptor @ref CLHOGDetector @ref CLHOGGradient @ref CLHOGMultiDetection
				386	- @ref CLLocallyConnectedMatrixMultiplyKernel / @ref CLLocallyConnectedLayer
				387	- @ref CLWeightsReshapeKernel / @ref CLConvolutionLayerReshapeWeights
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	388	- New C++ kernels:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	389	- @ref CPPDetectionWindowNonMaximaSuppressionKernel
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	390	- New NEON kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	391	- @ref NEBatchNormalizationLayerKernel / @ref NEBatchNormalizationLayer
				392	- @ref NEDepthConcatenateLayerKernel / @ref NEDepthConcatenateLayer
				393	- @ref NEDirectConvolutionLayerKernel / @ref NEDirectConvolutionLayer
				394	- @ref NELocallyConnectedMatrixMultiplyKernel / @ref NELocallyConnectedLayer
				395	- @ref NEWeightsReshapeKernel / @ref NEConvolutionLayerReshapeWeights
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	396
				397	v17.05 Public bug fixes release
				398	- Various bug fixes
				399	- Remaining of the functions ported to use accurate padding.
				400	- Library does not link against OpenCL anymore (It uses dlopen / dlsym at runtime instead to determine whether or not OpenCL is available).
				401	- Added "free" method to allocator.
				402	- Minimum version of g++ required for armv7 Linux changed from 4.8 to 4.9
				403
				404	v17.04 Public bug fixes release
				405
				406	The following functions have been ported to use the new accurate padding:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	407	- @ref CLColorConvertKernel
				408	- @ref CLEdgeNonMaxSuppressionKernel
				409	- @ref CLEdgeTraceKernel
				410	- @ref CLGaussianPyramidHorKernel
				411	- @ref CLGaussianPyramidVertKernel
				412	- @ref CLGradientKernel
				413	- @ref NEChannelCombineKernel
				414	- @ref NEFillArrayKernel
				415	- @ref NEGaussianPyramidHorKernel
				416	- @ref NEGaussianPyramidVertKernel
				417	- @ref NEHarrisScoreFP16Kernel
				418	- @ref NEHarrisScoreKernel
				419	- @ref NEHOGDetectorKernel
				420	- @ref NELogits1DMaxKernel
				421	- NELogits1DShiftExpSumKernel
				422	- NELogits1DNormKernel
				423	- @ref NENonMaximaSuppression3x3FP16Kernel
				424	- @ref NENonMaximaSuppression3x3Kernel
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	425
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	426	v17.03.1 First Major public release of the sources
				427	- Renamed the library to arm_compute
				428	- New CPP target introduced for C++ kernels shared between NEON and CL functions.
				429	- New padding calculation interface introduced and ported most kernels / functions to use it.
				430	- New OpenCL kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	431	- @ref CLGEMMLowpMatrixMultiplyKernel / CLGEMMLowp
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	432	- New NEON kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	433	- @ref NENormalizationLayerKernel / @ref NENormalizationLayer
				434	- @ref NETransposeKernel / @ref NETranspose
				435	- @ref NELogits1DMaxKernel, NELogits1DShiftExpSumKernel, NELogits1DNormKernel / @ref NESoftmaxLayer
				436	- @ref NEIm2ColKernel, @ref NECol2ImKernel, NEConvolutionLayerWeightsReshapeKernel / @ref NEConvolutionLayer
				437	- @ref NEGEMMMatrixAccumulateBiasesKernel / @ref NEFullyConnectedLayer
				438	- @ref NEGEMMLowpMatrixMultiplyKernel / NEGEMMLowp
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	439
				440	v17.03 Sources preview
				441	- New OpenCL kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	442	- @ref CLGradientKernel, @ref CLEdgeNonMaxSuppressionKernel, @ref CLEdgeTraceKernel / @ref CLCannyEdge
				443	- GEMM refactoring + FP16 support: @ref CLGEMMInterleave4x4Kernel, @ref CLGEMMTranspose1xWKernel, @ref CLGEMMMatrixMultiplyKernel, @ref CLGEMMMatrixAdditionKernel / @ref CLGEMM
				444	- @ref CLGEMMMatrixAccumulateBiasesKernel / @ref CLFullyConnectedLayer
				445	- @ref CLTransposeKernel / @ref CLTranspose
				446	- @ref CLLKTrackerInitKernel, @ref CLLKTrackerStage0Kernel, @ref CLLKTrackerStage1Kernel, @ref CLLKTrackerFinalizeKernel / @ref CLOpticalFlow
				447	- @ref CLNormalizationLayerKernel / @ref CLNormalizationLayer
				448	- @ref CLLaplacianPyramid, @ref CLLaplacianReconstruct
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	449	- New NEON kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	450	- @ref NEActivationLayerKernel / @ref NEActivationLayer
				451	- GEMM refactoring + FP16 support (Requires armv8.2 CPU): @ref NEGEMMInterleave4x4Kernel, @ref NEGEMMTranspose1xWKernel, @ref NEGEMMMatrixMultiplyKernel, @ref NEGEMMMatrixAdditionKernel / @ref NEGEMM
				452	- @ref NEPoolingLayerKernel / @ref NEPoolingLayer
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	453
				454	v17.02.1 Sources preview
				455	- New OpenCL kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	456	- @ref CLLogits1DMaxKernel, @ref CLLogits1DShiftExpSumKernel, @ref CLLogits1DNormKernel / @ref CLSoftmaxLayer
				457	- @ref CLPoolingLayerKernel / @ref CLPoolingLayer
				458	- @ref CLIm2ColKernel, @ref CLCol2ImKernel, CLConvolutionLayerWeightsReshapeKernel / @ref CLConvolutionLayer
				459	- @ref CLRemapKernel / @ref CLRemap
				460	- @ref CLGaussianPyramidHorKernel, @ref CLGaussianPyramidVertKernel / @ref CLGaussianPyramid, @ref CLGaussianPyramidHalf, @ref CLGaussianPyramidOrb
				461	- @ref CLMinMaxKernel, @ref CLMinMaxLocationKernel / @ref CLMinMaxLocation
				462	- @ref CLNonLinearFilterKernel / @ref CLNonLinearFilter
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	463	- New NEON FP16 kernels (Requires armv8.2 CPU)
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	464	- @ref NEAccumulateWeightedFP16Kernel
				465	- @ref NEBox3x3FP16Kernel
				466	- @ref NENonMaximaSuppression3x3FP16Kernel
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	467
				468	v17.02 Sources preview
				469	- New OpenCL kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	470	- @ref CLActivationLayerKernel / @ref CLActivationLayer
				471	- @ref CLChannelCombineKernel / @ref CLChannelCombine
				472	- @ref CLDerivativeKernel / @ref CLChannelExtract
				473	- @ref CLFastCornersKernel / @ref CLFastCorners
				474	- @ref CLMeanStdDevKernel / @ref CLMeanStdDev
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	475	- New NEON kernels / functions:
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	476	- HOG / SVM: @ref NEHOGOrientationBinningKernel, @ref NEHOGBlockNormalizationKernel, @ref NEHOGDetectorKernel, NEHOGNonMaximaSuppressionKernel / @ref NEHOGDescriptor, @ref NEHOGDetector, @ref NEHOGGradient, @ref NEHOGMultiDetection
				477	- @ref NENonLinearFilterKernel / @ref NENonLinearFilter
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	478	- Introduced a CLScheduler to manage the default context and command queue used by the runtime library and create synchronisation events.
				479	- Switched all the kernels / functions to use tensors instead of images.
				480	- Updated documentation to include instructions to build the library from sources.
				481
				482	v16.12 Binary preview release
				483	- Original release
				484
				485	@section S3_how_to_build How to build the library and the examples
				486
				487	@subsection S3_1_build_options Build options
				488
				489	scons 2.3 or above is required to build the library.
				490	To see the build options available simply run ```scons -h```:
				491
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	492	debug: Debug (yes\|no)
				493	default: False
				494	actual: False
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	495
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	496	asserts: Enable asserts (this flag is forced to 1 for debug=1) (yes\|no)
				497	default: False
				498	actual: False
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	499
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	500	arch: Target Architecture (armv7a\|arm64-v8a\|arm64-v8.2-a\|x86_32\|x86_64)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	501	default: armv7a
				502	actual: armv7a
				503
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	504	os: Target OS (linux\|android\|bare_metal)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	505	default: linux
				506	actual: linux
				507
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	508	build: Build type (native\|cross_compile\|embed_only)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	509	default: cross_compile
				510	actual: cross_compile
				511
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	512	examples: Build example programs (yes\|no)
				513	default: True
				514	actual: True
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	515
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	516	Werror: Enable/disable the -Werror compilation flag (yes\|no)
				517	default: True
				518	actual: True
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	519
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	520	opencl: Enable OpenCL support (yes\|no)
				521	default: True
				522	actual: True
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	523
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	524	neon: Enable Neon support (yes\|no)
				525	default: False
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	526	actual: False
				527
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	528	gles_compute: Enable OpenGL ES Compute Shader support (yes\|no)
				529	default: False
				530	actual: False
				531
				532	embed_kernels: Embed OpenCL kernels and OpenGL ES compute shader in library binary (yes\|no)
Anthony Barbier	cc0a80b	2017-12-15 11:37:29 +0000	[diff] [blame]	533	default: True
				534	actual: True
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	535
				536	set_soname: Set the library's soname and shlibversion (requires SCons 2.4 or above) (yes\|no)
				537	default: False
				538	actual: False
				539
				540	openmp: Enable OpenMP backend (yes\|no)
				541	default: False
				542	actual: False
				543
				544	cppthreads: Enable C++11 threads backend (yes\|no)
				545	default: True
				546	actual: True
				547
				548	build_dir: Specify sub-folder for the build ( /path/to/build_dir )
				549	default: .
				550	actual: .
				551
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	552	extra_cxx_flags: Extra CXX flags to be appended to the build command
				553	default:
				554	actual:
				555
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	556	pmu: Enable PMU counters (yes\|no)
				557	default: False
				558	actual: False
				559
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	560	mali: Enable Mali hardware counters (yes\|no)
				561	default: False
				562	actual: False
				563
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	564	validation_tests: Build validation test programs (yes\|no)
				565	default: False
				566	actual: False
				567
				568	benchmark_tests: Build benchmark test programs (yes\|no)
				569	default: False
				570	actual: False
				571
				572	@b debug / @b asserts:
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	573	- With debug=1 asserts are enabled, and the library is built with symbols and no optimisations enabled.
				574	- With debug=0 and asserts=1: Optimisations are enabled and symbols are removed, however all the asserts are still present (This is about 20% slower than the release build)
				575	- With debug=0 and asserts=0: All optimisations are enable and no validation is performed, if the application misuses the library it is likely to result in a crash. (Only use this mode once you are sure your application is working as expected).
				576
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	577	@b arch: The x86_32 and x86_64 targets can only be used with neon=0 and opencl=1.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	578
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	579	@b os: Choose the operating system you are targeting: Linux, Android or bare metal.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	580	@note bare metal can only be used for NEON (not OpenCL), only static libraries get built and NEON's multi-threading support is disabled.
				581
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	582	@b build: you can either build directly on your device (native) or cross compile from your desktop machine (cross-compile). In both cases make sure the compiler is available in your path.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	583
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	584	@note If you want to natively compile for 32bit on a 64bit ARM device running a 64bit OS then you will have to use cross-compile too.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	585
Anthony Barbier	2d0ce77	2018-02-21 15:35:36 +0000	[diff] [blame]	586	There is also an 'embed_only' option which will generate all the .embed files for the OpenCL kernels and / or OpenGLES compute shaders. This might be useful if using a different build system to compile the library.
				587
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	588	@b Werror: If you are compiling using the same toolchains as the ones used in this guide then there shouldn't be any warning and therefore you should be able to keep Werror=1. If with a different compiler version the library fails to build because of warnings interpreted as errors then, if you are sure the warnings are not important, you might want to try to build with Werror=0 (But please do report the issue either on Github or by an email to developer@arm.com so that the issue can be addressed).
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	589
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	590	@b opencl / @b neon / @b gles_compute: Choose which SIMD technology you want to target. (NEON for ARM Cortex-A CPUs or OpenCL / GLES_COMPUTE for ARM Mali GPUs)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	591
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	592	@b embed_kernels: For OpenCL / GLES_COMPUTE only: set embed_kernels=1 if you want the OpenCL / GLES_COMPUTE kernels to be built in the library's binaries instead of being read from separate ".cl" / ".cs" files. If embed_kernels is set to 0 then the application can set the path to the folder containing the OpenCL / GLES_COMPUTE kernel files by calling CLKernelLibrary::init() / GCKernelLibrary::init(). By default the path is set to "./cl_kernels" / "./cs_shaders".
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	593
				594	@b set_soname: Do you want to build the versioned version of the library ?
				595
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	596	If enabled the library will contain a SONAME and SHLIBVERSION and some symlinks will automatically be created between the objects.
				597	Example:
				598	libarm_compute_core.so -> libarm_compute_core.so.1.0.0
				599	libarm_compute_core.so.1 -> libarm_compute_core.so.1.0.0
				600	libarm_compute_core.so.1.0.0
				601
				602	@note This options is disabled by default as it requires SCons version 2.4 or above.
				603
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	604	@b extra_cxx_flags: Custom CXX flags which will be appended to the end of the build command.
				605
				606	@b build_dir: Build the library in a subfolder of the "build" folder. (Allows to build several configurations in parallel).
				607
				608	@b examples: Build or not the examples
				609
				610	@b validation_tests: Enable the build of the validation suite.
				611
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	612	@b benchmark_tests: Enable the build of the benchmark tests
				613
				614	@b pmu: Enable the PMU cycle counter to measure execution time in benchmark tests. (Your device needs to support it)
				615
Anthony Barbier	6a5627a	2017-09-26 14:42:02 +0100	[diff] [blame]	616	@b mali: Enable the collection of Mali hardware counters to measure execution time in benchmark tests. (Your device needs to have a Mali driver that supports it)
				617
Anthony Barbier	79c6178	2017-06-23 11:48:24 +0100	[diff] [blame]	618	@b openmp Build in the OpenMP scheduler for NEON.
				619
				620	@note Only works when building with g++ not clang++
				621
				622	@b cppthreads Build in the C++11 scheduler for NEON.
				623
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	624	@sa Scheduler::set
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	625
Moritz Pflanzer	07674de	2017-07-21 09:39:36 +0100	[diff] [blame]	626	@subsection S3_2_linux Building for Linux
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	627
				628	@subsubsection S3_2_1_library How to build the library ?
				629
				630	For Linux, the library was successfully built and tested using the following Linaro GCC toolchain:
				631
				632	- gcc-linaro-arm-linux-gnueabihf-4.9-2014.07_linux
				633	- gcc-linaro-4.9-2016.02-x86_64_aarch64-linux-gnu
				634	- gcc-linaro-6.3.1-2017.02-i686_aarch64-linux-gnu
				635
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	636	To cross-compile the library in debug mode, with NEON only support, for Linux 32bit:
				637
				638	scons Werror=1 -j8 debug=1 neon=1 opencl=0 os=linux arch=armv7a
				639
				640	To cross-compile the library in asserts mode, with OpenCL only support, for Linux 64bit:
				641
				642	scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=1 embed_kernels=1 os=linux arch=arm64-v8a
				643
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	644	To cross-compile the library in asserts mode, with GLES_COMPUTE only support, for Linux 64bit:
				645
				646	scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=0 gles_compute=1 embed_kernels=1 os=linux arch=arm64-v8a
				647
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	648	You can also compile the library natively on an ARM device by using <b>build=native</b>:
				649
				650	scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=arm64-v8a build=native
				651	scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv7a build=native
				652
				653	@note g++ for ARM is mono-arch, therefore if you want to compile for Linux 32bit on a Linux 64bit platform you will have to use a cross compiler.
				654
				655	For example on a 64bit Debian based system you would have to install <b>g++-arm-linux-gnueabihf</b>
				656
				657	apt-get install g++-arm-linux-gnueabihf
				658
				659	Then run
				660
				661	scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv7a build=cross_compile
				662
				663	or simply remove the build parameter as build=cross_compile is the default value:
				664
				665	scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=linux arch=armv7a
				666
				667	@attention To cross compile with opencl=1 you need to make sure to have a version of libOpenCL matching your target architecture.
				668
				669	@subsubsection S3_2_2_examples How to manually build the examples ?
				670
				671	The examples get automatically built by scons as part of the build process of the library described above. This section just describes how you can build and link your own application against our library.
				672
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	673	@note The following command lines assume the arm_compute binaries are present in the current directory or in the system library path. If this is not the case you can specify the location of the pre-built library with the compiler option -L. When building the OpenCL example the commands below assume that the CL headers are located in the include folder where the command is executed.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	674
				675	To cross compile a NEON example for Linux 32bit:
				676
Anthony Barbier	b2881fc	2017-09-29 17:12:12 +0100	[diff] [blame]	677	arm-linux-gnueabihf-g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute -larm_compute_core -o neon_convolution
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	678
				679	To cross compile a NEON example for Linux 64bit:
				680
Anthony Barbier	b2881fc	2017-09-29 17:12:12 +0100	[diff] [blame]	681	aarch64-linux-gnu-g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -L. -larm_compute -larm_compute_core -o neon_convolution
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	682
				683	(notice the only difference with the 32 bit command is that we don't need the -mfpu option and the compiler's name is different)
				684
				685	To cross compile an OpenCL example for Linux 32bit:
				686
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	687	arm-linux-gnueabihf-g++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute -larm_compute_core -o cl_convolution -DARM_COMPUTE_CL
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	688
				689	To cross compile an OpenCL example for Linux 64bit:
				690
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	691	aarch64-linux-gnu-g++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -L. -larm_compute -larm_compute_core -o cl_convolution -DARM_COMPUTE_CL
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	692
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	693	To cross compile a GLES example for Linux 32bit:
				694
				695	arm-linux-gnueabihf-g++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude/ -L. -larm_compute -larm_compute_core -std=c++11 -mfpu=neon -DARM_COMPUTE_GC -Iinclude/linux/ -o gc_absdiff
				696
				697	To cross compile a GLES example for Linux 64bit:
				698
				699	aarch64-linux-gnu-g++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude/ -L. -larm_compute -larm_compute_core -std=c++11 -DARM_COMPUTE_GC -Iinclude/linux/ -o gc_absdiff
				700
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	701	(notice the only difference with the 32 bit command is that we don't need the -mfpu option and the compiler's name is different)
				702
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	703	To cross compile the examples with the Graph API, such as graph_lenet.cpp, you need to link the examples against arm_compute_graph.so too.
				704
				705	@note The compute library must currently be built with both neon and opencl enabled - neon=1 and opencl=1
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	706
				707	i.e. to cross compile the "graph_lenet" example for Linux 32bit:
				708
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	709	arm-linux-gnueabihf-g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	710
				711	i.e. to cross compile the "graph_lenet" example for Linux 64bit:
				712
Isabella Gottardi	b28f29d	2017-11-09 17:05:07 +0000	[diff] [blame]	713	aarch64-linux-gnu-g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	714
				715	(notice the only difference with the 32 bit command is that we don't need the -mfpu option and the compiler's name is different)
				716
Anthony Barbier	e500747	2017-10-27 15:01:44 +0100	[diff] [blame]	717	@note If compiling using static libraries, this order must be followed when linking: arm_compute_graph_static, arm_compute, arm_compute_core
				718
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	719	To compile natively (i.e directly on an ARM device) for NEON for Linux 32bit:
				720
Anthony Barbier	b2881fc	2017-09-29 17:12:12 +0100	[diff] [blame]	721	g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -larm_compute -larm_compute_core -o neon_convolution
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	722
				723	To compile natively (i.e directly on an ARM device) for NEON for Linux 64bit:
				724
Anthony Barbier	b2881fc	2017-09-29 17:12:12 +0100	[diff] [blame]	725	g++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute -larm_compute_core -o neon_convolution
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	726
				727	(notice the only difference with the 32 bit command is that we don't need the -mfpu option)
				728
				729	To compile natively (i.e directly on an ARM device) for OpenCL for Linux 32bit or Linux 64bit:
				730
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	731	g++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute -larm_compute_core -o cl_convolution -DARM_COMPUTE_CL
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	732
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	733	To compile natively (i.e directly on an ARM device) for GLES for Linux 32bit or Linux 64bit:
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	734
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	735	g++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude/ -L. -larm_compute -larm_compute_core -std=c++11 -DARM_COMPUTE_GC -Iinclude/linux/ -o gc_absdiff
				736
				737	To compile natively the examples with the Graph API, such as graph_lenet.cpp, you need to link the examples against arm_compute_graph.so too.
				738	@note The compute library must currently be built with both neon and opencl enabled - neon=1 and opencl=1
				739
				740	i.e. to natively compile the "graph_lenet" example for Linux 32bit:
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	741
Isabella Gottardi	b28f29d	2017-11-09 17:05:07 +0000	[diff] [blame]	742	g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -mfpu=neon -L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	743
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	744	i.e. to natively compile the "graph_lenet" example for Linux 64bit:
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	745
Isabella Gottardi	b28f29d	2017-11-09 17:05:07 +0000	[diff] [blame]	746	g++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 L. -larm_compute_graph -larm_compute -larm_compute_core -Wl,--allow-shlib-undefined -o graph_lenet
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	747
				748	(notice the only difference with the 32 bit command is that we don't need the -mfpu option)
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	749
Anthony Barbier	e500747	2017-10-27 15:01:44 +0100	[diff] [blame]	750	@note If compiling using static libraries, this order must be followed when linking: arm_compute_graph_static, arm_compute, arm_compute_core
				751
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	752	@note These two commands assume libarm_compute.so is available in your library path, if not add the path to it using -L
				753
				754	To run the built executable simply run:
				755
				756	LD_LIBRARY_PATH=build ./neon_convolution
				757
				758	or
				759
				760	LD_LIBRARY_PATH=build ./cl_convolution
				761
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	762	@note Examples accept different types of arguments, to find out what they are run the example without any argument and the help will be displayed at the beginning of the run.
				763
				764	For example:
				765	LD_LIBRARY_PATH=. ./graph_lenet
				766
				767	./graph_lenet
				768
				769	Usage: ./graph_lenet [target] [path_to_data] [batches]
				770
				771	No data folder provided: using random values
				772
				773	Test passed
				774
				775	In this case the first argument of LeNet (like all the graph examples) is the target (i.e 0 to run on NEON, 1 to run on OpenCL if available, 2 to run on OpenCL using the CLTuner), the second argument is the path to the folder containing the npy files for the weights and finally the third argument is the number of batches to run.
				776
Moritz Pflanzer	07674de	2017-07-21 09:39:36 +0100	[diff] [blame]	777	@subsection S3_3_android Building for Android
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	778
				779	For Android, the library was successfully built and tested using Google's standalone toolchains:
Anthony Barbier	a8a28f6	2018-02-26 19:16:32 +0000	[diff] [blame]	780	- clang++ from NDK r16b for armv7a
				781	- clang++ from NDK r16b for arm64-v8a
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	782
				783	Here is a guide to <a href="https://developer.android.com/ndk/guides/standalone_toolchain.html">create your Android standalone toolchains from the NDK</a>
				784
Anthony Barbier	a8a28f6	2018-02-26 19:16:32 +0000	[diff] [blame]	785	- Download the NDK r16b from here: https://developer.android.com/ndk/downloads/index.html
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	786	- Make sure you have Python 2 installed on your machine.
				787	- Generate the 32 and/or 64 toolchains by running the following commands:
				788
Anthony Barbier	a8a28f6	2018-02-26 19:16:32 +0000	[diff] [blame]	789	$NDK/build/tools/make_standalone_toolchain.py --arch arm64 --install-dir $MY_TOOLCHAINS/aarch64-linux-android-ndk-r16b --stl gnustl --api 21
				790	$NDK/build/tools/make_standalone_toolchain.py --arch arm --install-dir $MY_TOOLCHAINS/arm-linux-android-ndk-r16b --stl gnustl --api 21
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	791
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	792	@attention Due to some NDK issues make sure you use clang++ & gnustl
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	793
				794	@note Make sure to add the toolchains to your PATH: export PATH=$PATH:$MY_TOOLCHAINS/aarch64-linux-android-4.9/bin:$MY_TOOLCHAINS/arm-linux-androideabi-4.9/bin
				795
				796	@subsubsection S3_3_1_library How to build the library ?
				797
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	798	To cross-compile the library in debug mode, with NEON only support, for Android 32bit:
				799
				800	CXX=clang++ CC=clang scons Werror=1 -j8 debug=1 neon=1 opencl=0 os=android arch=armv7a
				801
				802	To cross-compile the library in asserts mode, with OpenCL only support, for Android 64bit:
				803
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	804	CXX=clang++ CC=clang scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=1 embed_kernels=1 os=android arch=arm64-v8a
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	805
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	806	To cross-compile the library in asserts mode, with GLES_COMPUTE only support, for Android 64bit:
				807
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	808	CXX=clang++ CC=clang scons Werror=1 -j8 debug=0 asserts=1 neon=0 opencl=0 gles_compute=1 embed_kernels=1 os=android arch=arm64-v8a
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	809
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	810	@subsubsection S3_3_2_examples How to manually build the examples ?
				811
				812	The examples get automatically built by scons as part of the build process of the library described above. This section just describes how you can build and link your own application against our library.
				813
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	814	@note The following command lines assume the arm_compute binaries are present in the current directory or in the system library path. If this is not the case you can specify the location of the pre-built library with the compiler option -L. When building the OpenCL example the commands below assume that the CL headers are located in the include folder where the command is executed.
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	815
				816	Once you've got your Android standalone toolchain built and added to your path you can do the following:
				817
				818	To cross compile a NEON example:
				819
				820	#32 bit:
Georgios Pinitas	9873ea3	2017-12-05 15:28:55 +0000	[diff] [blame]	821	arm-linux-androideabi-clang++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o neon_convolution_arm -static-libstdc++ -pie
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	822	#64 bit:
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	823	aarch64-linux-android-clang++ examples/neon_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o neon_convolution_aarch64 -static-libstdc++ -pie
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	824
				825	To cross compile an OpenCL example:
				826
				827	#32 bit:
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	828	arm-linux-androideabi-clang++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o cl_convolution_arm -static-libstdc++ -pie -DARM_COMPUTE_CL
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	829	#64 bit:
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	830	aarch64-linux-android-clang++ examples/cl_convolution.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o cl_convolution_aarch64 -static-libstdc++ -pie -DARM_COMPUTE_CL
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	831
				832	To cross compile a GLES example:
Anthony Barbier	cc0a80b	2017-12-15 11:37:29 +0000	[diff] [blame]	833
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	834	#32 bit:
				835	arm-linux-androideabi-clang++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o gc_absdiff_arm -static-libstdc++ -pie -DARM_COMPUTE_GC
				836	#64 bit:
				837	aarch64-linux-android-clang++ examples/gc_absdiff.cpp utils/Utils.cpp -I. -Iinclude -std=c++11 -larm_compute-static -larm_compute_core-static -L. -o gc_absdiff_aarch64 -static-libstdc++ -pie -DARM_COMPUTE_GC
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	838
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	839	To cross compile the examples with the Graph API, such as graph_lenet.cpp, you need to link the library arm_compute_graph also.
				840	(notice the compute library has to be built with both neon and opencl enabled - neon=1 and opencl=1)
				841
				842	#32 bit:
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	843	arm-linux-androideabi-clang++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -Wl,--whole-archive -larm_compute_graph-static -Wl,--no-whole-archive -larm_compute-static -larm_compute_core-static -L. -o graph_lenet_arm -static-libstdc++ -pie -DARM_COMPUTE_CL
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	844	#64 bit:
Georgios Pinitas	d9eb275	2018-04-03 13:44:29 +0100	[diff] [blame]	845	aarch64-linux-android-clang++ examples/graph_lenet.cpp utils/Utils.cpp utils/GraphUtils.cpp -I. -Iinclude -std=c++11 -Wl,--whole-archive -larm_compute_graph-static -Wl,--no-whole-archive -larm_compute-static -larm_compute_core-static -L. -o graph_lenet_aarch64 -static-libstdc++ -pie -DARM_COMPUTE_CL
Gian Marco Iodice	daec1aa	2017-09-29 12:03:18 +0100	[diff] [blame]	846
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	847	@note Due to some issues in older versions of the Mali OpenCL DDK (<= r13p0), we recommend to link arm_compute statically on Android.
Anthony Barbier	20dbb82	2017-12-13 21:19:39 +0000	[diff] [blame]	848	@note When linked statically the arm_compute_graph library currently needs the --whole-archive linker flag in order to work properly
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	849
				850	Then you need to do is upload the executable and the shared library to the device using ADB:
				851
				852	adb push neon_convolution_arm /data/local/tmp/
				853	adb push cl_convolution_arm /data/local/tmp/
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	854	adb push gc_absdiff_arm /data/local/tmp/
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	855	adb shell chmod 777 -R /data/local/tmp/
				856
				857	And finally to run the example:
				858
				859	adb shell /data/local/tmp/neon_convolution_arm
				860	adb shell /data/local/tmp/cl_convolution_arm
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	861	adb shell /data/local/tmp/gc_absdiff_arm
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	862
				863	For 64bit:
				864
				865	adb push neon_convolution_aarch64 /data/local/tmp/
				866	adb push cl_convolution_aarch64 /data/local/tmp/
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	867	adb push gc_absdiff_aarch64 /data/local/tmp/
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	868	adb shell chmod 777 -R /data/local/tmp/
				869
				870	And finally to run the example:
				871
				872	adb shell /data/local/tmp/neon_convolution_aarch64
				873	adb shell /data/local/tmp/cl_convolution_aarch64
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	874	adb shell /data/local/tmp/gc_absdiff_aarch64
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	875
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	876	@note Examples accept different types of arguments, to find out what they are run the example without any argument and the help will be displayed at the beginning of the run.
				877
				878	For example:
				879	adb shell /data/local/tmp/graph_lenet
				880
				881	/data/local/tmp/graph_lenet
				882
				883	Usage: /data/local/tmp/graph_lenet [target] [path_to_data] [batches]
				884
				885	No data folder provided: using random values
				886
				887	Test passed
				888
				889	In this case the first argument of LeNet (like all the graph examples) is the target (i.e 0 to run on NEON, 1 to run on OpenCL if available, 2 to run on OpenCL using the CLTuner), the second argument is the path to the folder containing the npy files for the weights and finally the third argument is the number of batches to run.
				890
Michalis Spyrou	6e52ba3	2017-10-04 15:40:38 +0100	[diff] [blame]	891	@subsection S3_4_bare_metal Building for bare metal
				892
				893	For bare metal, the library was successfully built using linaros's latest (gcc-linaro-6.3.1-2017.05) bare metal toolchains:
				894	- arm-eabi for armv7a
				895	- aarch64-elf for arm64-v8a
				896
				897	Download linaro for <a href="https://releases.linaro.org/components/toolchain/binaries/6.3-2017.05/arm-eabi/">armv7a</a> and <a href="https://releases.linaro.org/components/toolchain/binaries/6.3-2017.05/aarch64-elf/">arm64-v8a</a>.
				898
				899	@note Make sure to add the toolchains to your PATH: export PATH=$PATH:$MY_TOOLCHAINS/gcc-linaro-6.3.1-2017.05-x86_64_aarch64-elf/bin:$MY_TOOLCHAINS/gcc-linaro-6.3.1-2017.05-x86_64_arm-eabi/bin
				900
				901	@subsubsection S3_4_1_library How to build the library ?
				902
				903	To cross-compile the library with NEON support for baremetal arm64-v8a:
				904
				905	scons Werror=1 -j8 debug=0 neon=1 opencl=0 os=bare_metal arch=arm64-v8a build=cross_compile cppthreads=0 openmp=0 standalone=1
				906
				907	@subsubsection S3_4_2_examples How to manually build the examples ?
				908
				909	Examples are disabled when building for bare metal. If you want to build the examples you need to provide a custom bootcode depending on the target architecture and link against the compute library. More information about bare metal bootcode can be found <a href="http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dai0527a/index.html">here</a>.
				910
				911	@subsection S3_5_windows_host Building on a Windows host system
Moritz Pflanzer	07674de	2017-07-21 09:39:36 +0100	[diff] [blame]	912
				913	Using `scons` directly from the Windows command line is known to cause
				914	problems. The reason seems to be that if `scons` is setup for cross-compilation
				915	it gets confused about Windows style paths (using backslashes). Thus it is
				916	recommended to follow one of the options outlined below.
				917
Michalis Spyrou	6e52ba3	2017-10-04 15:40:38 +0100	[diff] [blame]	918	@subsubsection S3_5_1_ubuntu_on_windows Bash on Ubuntu on Windows
Moritz Pflanzer	07674de	2017-07-21 09:39:36 +0100	[diff] [blame]	919
				920	The best and easiest option is to use
				921	<a href="https://msdn.microsoft.com/en-gb/commandline/wsl/about">Ubuntu on Windows</a>.
				922	This feature is still marked as beta and thus might not be available.
				923	However, if it is building the library is as simple as opening a *Bash on
				924	Ubuntu on Windows* shell and following the general guidelines given above.
				925
Michalis Spyrou	6e52ba3	2017-10-04 15:40:38 +0100	[diff] [blame]	926	@subsubsection S3_5_2_cygwin Cygwin
Moritz Pflanzer	07674de	2017-07-21 09:39:36 +0100	[diff] [blame]	927
				928	If the Windows subsystem for Linux is not available <a href="https://www.cygwin.com/">Cygwin</a>
				929	can be used to install and run `scons`. In addition to the default packages
				930	installed by Cygwin `scons` has to be selected in the installer. (`git` might
				931	also be useful but is not strictly required if you already have got the source
				932	code of the library.) Linaro provides pre-built versions of
				933	<a href="http://releases.linaro.org/components/toolchain/binaries/">GCC cross-compilers</a>
				934	that can be used from the Cygwin terminal. When building for Android the
				935	compiler is included in the Android standalone toolchain. After everything has
				936	been set up in the Cygwin terminal the general guide on building the library
				937	can be followed.
				938
Michalis Spyrou	6e52ba3	2017-10-04 15:40:38 +0100	[diff] [blame]	939	@subsection S3_6_cl_stub_library The OpenCL stub library
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	940
				941	In the opencl-1.2-stubs folder you will find the sources to build a stub OpenCL library which then can be used to link your application or arm_compute against.
				942
				943	If you preferred you could retrieve the OpenCL library from your device and link against this one but often this library will have dependencies on a range of system libraries forcing you to link your application against those too even though it is not using them.
				944
				945	@warning This OpenCL library provided is a stub and not a real implementation. You can use it to resolve OpenCL's symbols in arm_compute while building the example but you must make sure the real libOpenCL.so is in your PATH when running the example or it will not work.
				946
				947	To cross-compile the stub OpenCL library simply run:
				948
				949	<target-prefix>-gcc -o libOpenCL.so -Iinclude opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
				950
				951	For example:
				952
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	953	#Linux 32bit
				954	arm-linux-gnueabihf-gcc -o libOpenCL.so -Iinclude opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
				955	#Linux 64bit
				956	aarch64-linux-gnu-gcc -o libOpenCL.so -Iinclude -shared opencl-1.2-stubs/opencl_stubs.c -fPIC
				957	#Android 32bit
				958	arm-linux-androideabi-clang -o libOpenCL.so -Iinclude -shared opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
				959	#Android 64bit
Anthony Barbier	14c86a9	2017-12-14 16:27:41 +0000	[diff] [blame]	960	aarch64-linux-android-clang -o libOpenCL.so -Iinclude -shared opencl-1.2-stubs/opencl_stubs.c -fPIC -shared
				961
				962	@subsection S3_7_gles_stub_library The Linux OpenGLES and EGL stub libraries
				963
				964	In the opengles-3.1-stubs folder you will find the sources to build stub EGL and OpenGLES libraries which then can be used to link your Linux application of arm_compute against.
				965
				966	@note The stub libraries are only needed on Linux. For Android, the NDK toolchains already provide the meta-EGL and meta-GLES libraries.
				967
				968	To cross-compile the stub OpenGLES and EGL libraries simply run:
				969
				970	<target-prefix>-gcc -o libEGL.so -Iinclude/linux opengles-3.1-stubs/EGL.c -fPIC -shared
				971	<target-prefix>-gcc -o libGLESv2.so -Iinclude/linux opengles-3.1-stubs/GLESv2.c -fPIC -shared
				972
				973	#Linux 32bit
				974	arm-linux-gnueabihf-gcc -o libEGL.so -Iinclude/linux opengles-3.1-stubs/EGL.c -fPIC -shared
				975	arm-linux-gnueabihf-gcc -o libGLESv2.so -Iinclude/linux opengles-3.1-stubs/GLESv2.c -fPIC -shared
				976
				977	#Linux 64bit
				978	aarch64-linux-gnu-gcc -o libEGL.so -Iinclude/linux opengles-3.1-stubs/EGL.c -fPIC -shared
				979	aarch64-linux-gnu-gcc -o libGLESv2.so -Iinclude/linux opengles-3.1-stubs/GLESv2.c -fPIC -shared
Anthony Barbier	6ff3b19	2017-09-04 18:44:23 +0100	[diff] [blame]	980	*/
Anthony Barbier	3762e74	2018-03-02 11:49:33 +0000	[diff] [blame]	981	} // namespace arm_compute