Gitiles
Code Review
Sign In
review.mlplatform.org
/
ml
/
ComputeLibrary
/
25f45a4a0158a709d28a5207a867a9b5ce390621
/
src
/
runtime
/
NEON
/
functions
25f45a4
COMPMID-1060 LSTM FP32 NEON
by Michalis Spyrou
· 6 years ago
26b2216
COMPMID-1480 Add support for NHWC QASYMM8/FP32(non-optimized) to NEON DepthwiseConvolution
by Giorgio Arena
· 6 years ago
1d38259
COMPMID-1366 Implement NECopy
by Michalis Spyrou
· 6 years ago
916d1bc
COMPMID-1498 - Enable grouping in CLGEMMConvolutionLayer
by Gian Marco Iodice
· 6 years ago
e3f1bd1
COMPMID-1246 Remove unused window iterator from NERNNLayer.
by Michalis Spyrou
· 6 years ago
db9d46d
COMPMID-1485 - Add support for NHWC when running NEGEMMConvolutionLayer with FP16/QASYMM8
by Gian Marco Iodice
· 6 years ago
0f17039
COMPMID-1342 Add grouping support to CLIm2ColKernel
by Giorgio Arena
· 6 years ago
d8b03dd
MLCE-13: Sanitizing matrix argument in the Warp.
by Pablo Tello
· 6 years ago
f2bd261
COMPMID-1188 - Removed the multiplication by 4 in NEGEMMInterleavedWrapper
by Gian Marco Iodice
· 6 years ago
215b4ea
COMPMID-1277 - Optimizing CLIm2ColKernel for NHWC.
by Gian Marco Iodice
· 6 years ago
20394d5
COMPMID-1248 Enabled memory manager in NEWinogradConvolutionLayer
by Anthony Barbier
· 6 years ago
3d677cc
COMPMID-1406: Refactor gemm_interleaved to use our own types and scheduler
by Anthony Barbier
· 6 years ago
597a856
COMPMID-872 - Rework NEGEMMConvolutionLayer to use NEGEMM
by Gian Marco Iodice
· 6 years ago
195b0ba
MLCE-36: FC tranpose weights
by Georgios Pinitas
· 6 years ago
ebbb7f2
COMPMID-1188 - Fixed CLGEMMConvolutionLayer/NEGEMMConvolutionLayer for NHWC
by Gian Marco Iodice
· 6 years ago
9036749
COMPMID-1445: Fixed NEON GEMM assembly dispatch for QASYMM8
by Anthony Barbier
· 6 years ago
40ed6d8
COMPMID-1440: Access original B in gemm assembly when not pretransposed.
by Georgios Pinitas
· 6 years ago
eb84d6b
COMPMID-1434: Fix NEWinograd for NHWC and sub-tensors
by Georgios Pinitas
· 6 years ago
ef776a8
COMPMID-1386: Add FC convert weights on NEON
by Georgios Pinitas
· 6 years ago
a855af1
COMPMID-1401 Implement NEFullyConnectedLayer for QASYMM8
by Giorgio Arena
· 6 years ago
eaefd00
COMPMID-1419: Make NEGEMMAssemblyDispatch dynamically typed instead of templated
by Anthony Barbier
· 6 years ago
c8e84b5
COMPMID-1405: Create our own gemm_native kernel / function.
by Anthony Barbier
· 6 years ago
7d66a8e
COMPMID-1386: Add support for converting weights for CL.
by Georgios Pinitas
· 6 years ago
ae54e02
COMPMID-1364: Add support for NHWC in NEDepthConcatenateLayer
by Georgios Pinitas
· 6 years ago
e155337
COMPMID-1357: Stop passing around raw pointers in NEWinogradConvolution
by Anthony Barbier
· 6 years ago
578225e
COMPMID-1357: Port NEWinogradConvolution to use NEGEMMAssemblyDispatch
by Anthony Barbier
· 6 years ago
e29acf1
COMPMID-1365: Add support for NHWC in CLDepthConcatenateLayer
by Georgios Pinitas
· 6 years ago
67fd4e8
COMPMID-1381: Fix nightly build failure on armv7
by Anthony Barbier
· 6 years ago
71d9b57
COMPMID-1381: Cleaned up the AssemblyHelper interface
by Anthony Barbier
· 6 years ago
7784c83
COMPMID-1167: Validation for NEDepthwiseConvolutionLayer
by Abe Mbise
· 6 years ago
7485d5a
COMPMID-970 : Remove QS8 / QS16 support
by Vidhya Sudhan Loganathan
· 6 years ago
17d6d3c
COMPMID-1352: Disable support for 4D softmax layer.
by Georgios Pinitas
· 6 years ago
f3139dd
COMPMID-1334 (Nightly) Fix std::bad_alloc error in 32-bit NEON runs
by Giorgio Arena
· 6 years ago
bb71fe5
COMPMID-1253: Nightly: Fix Canny Edge NEON failing
by Michele Di Giorgio
· 6 years ago
7282d56
COMPMID-1287: Extending NEWinogradLayer test suite
by Pablo Tello
· 6 years ago
542e92d
COMPMID-1067 NEON RNN FP32 / FP16
by Michalis Spyrou
· 6 years ago
7221933
COMPMID-1145: (API) Introduce prepare() stage (NEON/CL/GLES)
by Georgios Pinitas
· 6 years ago
7df2786
COMPMID-1162: Enable NHWC data layout support for NEWinogradConvolutionLayer - part1
by Pablo Tello
· 6 years ago
e083771
COMPMID-1160 Turn Graph hints into heuristics
by Giorgio Arena
· 6 years ago
b95e210
COMPMID-1209: Enable memory manager for the GEMM workspace buffer
by Georgios Pinitas
· 6 years ago
1b99338
COMPMID-568: Implement Canny edge function for CL/NEON
by Abe Mbise
· 7 years ago
2345f43
COMPMID-1168: enabling memory manager for _gemm_output gives wrong results
by Georgios Pinitas
· 6 years ago
fa7ad56
COMPMID-1163: NEON Scale NHWC failures
by Georgios Pinitas
· 6 years ago
96b86a9
COMPMID-959 - Fix Validate in Convolution Layer
by Isabella Gottardi
· 6 years ago
a3221e6
COMPMID-1106 Add fast math support in NEWinogradConvolutionLayer
by Giorgio Arena
· 6 years ago
20b4313
COMPMID-814: Add validate method to scale.
by Georgios Pinitas
· 6 years ago
4d1e8a2
COMPMID-1166: Fixed alpha/beta mixup for some merges.
by David Mansell
· 6 years ago
a8aef29
COMPMID-792 - Re-enabled Winograd on NEON in all graph examples.
by Gian Marco Iodice
· 6 years ago
cb0010b
COMPMID-1102 : Enable the use of 4x4 tile sizes in neon implementation of winograd conv.
by Vidhya Sudhan Loganathan
· 6 years ago
73d4aef
COMPMID-948: Add validation to NEL2NormalizeLayer
by John Richardson
· 6 years ago
393fa4c
COMPMID-814: NEScale NHWC support
by Georgios Pinitas
· 6 years ago
657bdb3
COMPMID-1050 CL/NEON: Create a function to convert the 2D weights of FC layer from NHWC to NCHW and viceversa
by Giorgio Arena
· 6 years ago
e250389
COMPMID-810 Add NHWC data format support for NEON convolution
by Michalis Spyrou
· 6 years ago
d8cde85
COMPMID-959: Assembly kernel marks weights as unused when pretranspose
by Georgios Pinitas
· 6 years ago
932b561
COMPMID-959: Perform pretranspose if allowed on NEON assembly
by Georgios Pinitas
· 6 years ago
c0f5443
COMPMID-808 Add NHWC data format support for NEON direct convolution
by Giorgio Arena
· 6 years ago
7657224
COMPMID-926 Add depth multiplier support to NEON/CL/GLES depthwise convolution
by Giorgio Arena
· 6 years ago
57dac84
COMPMID-806 Add NHWC data format support format for NEON pooling
by Michalis Spyrou
· 6 years ago
9fb1159
COMPMID-1074: Rename WinograLayer.cpp to WinogradConvolutionLayer.cpp
by Georgios Pinitas
· 6 years ago
2b3129e
COMPMID-1041 NEON Winograd: update function to use GEMM function
by Michalis Spyrou
· 6 years ago
84ce1f9
COMPMID-718 : Winograd: add validate method and tests
by Vidhya Sudhan Loganathan
· 6 years ago
3ca9786
COMPMID-718 : Winograd: add validate method and tests
by Vidhya Sudhan Loganathan
· 6 years ago
e52a300
COMPMID-1026 - Add support for 4x4 output tile in CLWinogradConvolutionLayer
by Gian Marco Iodice
· 6 years ago
bb54e4e
COMPMID-797 Integrate Mobilenet QASYMM8 with new graph.
by Giorgio Arena
· 6 years ago
1ed1fc6
COMPMID-812 Add NHWC data format support for NEON depthwise convolution (optimized case).
by Giorgio Arena
· 6 years ago
a72300a
COMPMID-1051 - Fix validate method in NEGEMMConvolutionLayer
by Gian Marco Iodice
· 6 years ago
95a3f55
COMPMID-959: Disabled winograd on Arm Cortex-A53 while we're working on a fix
by Anthony Barbier
· 6 years ago
156fcf3
COMPMID-802 Add NHWC data format support for NEON im2col.
by Giorgio Arena
· 6 years ago
4d9379a
COMPMID-1040: Added support for nullptr bias tensor in NEWinogradLayer
by Andrew Mundy
· 6 years ago
3f217ec
COMPMID-908 - Merge Activation layer with Convolution Layer (NEON. CL, GLES)
by Isabella Gottardi
· 6 years ago
d9eb275
COMPMID-797: Switch to new graph.
by Georgios Pinitas
· 6 years ago
7fad9b1
COMPMID-1021: CPUInfo refactoring.
by Pablo Tello
· 6 years ago
1562be3
COMPMID-998: Release unused trainable parameters.
by Georgios Pinitas
· 6 years ago
7da29b6
COMPMID-1017: Implement dilated convolution in NEON, OpenCL, and GC
by Alex Gilday
· 6 years ago
61b4fca
COMPMID-959 Fix clang-tidy issue
by Anthony Barbier
· 6 years ago
27c08ab
COMPMID-754: Add validation to LocallyConnected and NEDeconv layers
by Alex Gilday
· 6 years ago
60954c6
COMPMID-754: Add validation to (De)QuantizationLayers
by Alex Gilday
· 6 years ago
66c656a
COMPMID-881: Addded missing DataType::QASYMM8 in GEMMLowp function.
by Pablo Tello
· 6 years ago
eb82fd2
COMPMID-881: RSH new arm_gemm interface.
by Pablo Tello
· 6 years ago
d8734b5
COMPMID-793 : Add graph intermediate representation
by Georgios Pinitas
· 7 years ago
33a6990
COMPMID-540 Replace NEDeconvolutionLayerUpsampleKernel with NEScaleKernel
by Michalis Spyrou
· 6 years ago
861f0db
COMPMID-941 Add NEON broadcast multiply support
by Michalis Spyrou
· 6 years ago
88627fb
COMPMID-959: Update valid region in DepthConcatenate
by Georgios Pinitas
· 6 years ago
b4e3e1c
COMPMID-617: Add validate support for NEON FullyConnectedLayer
by Ioan-Cristian Szabo
· 7 years ago
6acc6ad
COMPMID-846: Create a ConvolutionLayer for NEON
by Isabella Gottardi
· 6 years ago
9be0c5a
COMPMID-936: Convolution failure in NEON Convolution Layer.
by Georgios Pinitas
· 6 years ago
dde9ec9
COMPMID-909: Enabling in-place computation for batchnormalization and activation at graph level
by Michele Di Giorgio
· 6 years ago
21f67d6
COMPMID-934: Return an error in Validate when we don't support asymmetric padding
by Anthony Barbier
· 6 years ago
f6c572c
COMPMID-784: Productise Winograd.
by Pablo Tello
· 6 years ago
57c033b
COMPMID-906: Use fused activation in NEON Batch normalization
by Georgios Pinitas
· 6 years ago
02ee429
COMPMID-765: Fix CPPPermute error when permuting the strides.
by Georgios Pinitas
· 6 years ago
056b5d9
COMPMID-765: Switch 1x1 DeconvolutionLayer to use the ConvolutionLayer
by Georgios Pinitas
· 6 years ago
284cfe2
COMPMID-903: Implements NEPermute for NHWC conversions
by Georgios Pinitas
· 6 years ago
ced7a8d
COMPMID-875: Deconvolution 4x4 not working
by Georgios Pinitas
· 6 years ago
1167487
COMPMID-897 Merge batch normalization with bounded relu
by Giorgio Arena
· 6 years ago
4074c99
COMPMID-873: Integrate RSH NEON Depthwise Convolution routine
by Georgios Pinitas
· 6 years ago
6e464c3
COMPMID-828 - Add support for non square pool size - Part1
by Isabella Gottardi
· 6 years ago
604afc0
COMPMID-894: Segfault: neon_cnn on S5 neo
by Georgios Pinitas
· 6 years ago
679463a
COMPMID-784: Added support for PADDING = SAME in Winograd layer.
by Pablo Tello
· 6 years ago
b7e3028
COMPMID-890: Valgrind: NEON Convolution Layer validation fails
by Isabella Gottardi
· 6 years ago
Next »