Gitiles
Code Review
Sign In
review.mlplatform.org
/
ml
/
ComputeLibrary
/
748a7c81245ae81d04607b3a762cf65cd39026f2
/
src
« Previous
3d8fe49
INFPRF-609:Performance Issue of the Latest ArmCL
by giuros01
· 5 years ago
2aa7fd0
COMPMID-2601 [CL] add mixed precision support to PoolingLayer
by Sang-Hoon Park
· 5 years ago
5c4a8e9
COMPMID-2592 Create a new kernel for CLPad with SYMMETRIC and REFLECT mode
by Giorgio Arena
· 5 years ago
04ea4e8
COMPMID-2581: Fuse batch normalization with convolution and depthwise convolution at graph level for OpenCL - FP16
by giuros01
· 5 years ago
ae0fc86
COMPMID-2452: Dot product optimizations on merge/transforms
by Georgios Pinitas
· 5 years ago
9637b2e
COMPMID-2671: Change ArgMinMax NEON/CL output type to Signed32
by Michele Di Giorgio
· 5 years ago
cecb0a7
COMPMID-2637 [CL] fix broadcast pixel-wise multiplication with 5D tensors
by Sang-Hoon Park
· 5 years ago
879d131
COMPMID-2452: Fix compilation issues on armclang
by Georgios Pinitas
· 5 years ago
0c17aa2
COMPMID-2571: Add mixed-precision support in CLGEMMReshaped for FP16
by Gian Marco Iodice
· 5 years ago
40958ad
COMPMID-2663: Interface to unbound mappings.
by Georgios Pinitas
· 5 years ago
6b612f5
COMPMID-2310: CLGenerateProposalsLayer: support for QASYMM8
by Michele Di Giorgio
· 5 years ago
bc88c62
COMPMID-2452: Replace DOT assemble macro with the corresponding word directives.
by Georgios Pinitas
· 5 years ago
05639f6
COMPMID-2571: Add support for FP16 in CLGEMMReshaped - part 1
by Gian Marco Iodice
· 5 years ago
1a569a3
COMPMID-2161 [NEON] Create IWeightManager class
by Michalis Spyrou
· 5 years ago
572576a
COMPMID-2674: Buffer overrun in GEMM assembly tranform kernels.
by Georgios Pinitas
· 5 years ago
2725197
MLCE-140: Workaround multi-threading problem with NEON backend
by Pablo Tello
· 5 years ago
b785dd4
COMPMID-2670: [CL/GC] Create a test case for dynamic tensor support
by Georgios Pinitas
· 5 years ago
ee6454a
COMPMID-2668: Add GEMM/GEMMLowp heuristic for Mali-G51
by Gian Marco Iodice
· 5 years ago
4c268b9
COMPMID-2681: Use NE/CL/Pad in NE/CL/GenerateProposalsLayer
by Michele Di Giorgio
· 5 years ago
c9564cb
COMPMID-2257: Implement NEGenerateProposals.
by Pablo Tello
· 5 years ago
7b9d7ca
COMPMID-2675: Fix arguments passed at compile time for GEMM - OpenCL
by Gian Marco Iodice
· 5 years ago
769c638
COMPMID-2314: Implement NEON INSTANCE_NORMALIZATION function
by Manuel Bottini
· 5 years ago
2568c6b
COMPMID-2665: CLGEMMMatrixMultiplyReshaped nightly tests are failing due to cl error -45
by Michele Di Giorgio
· 5 years ago
26014cf
COMPMID-2649: Generalize MemoryGroup.
by Georgios Pinitas
· 5 years ago
caa7dee
COMPMID-2641 [NEON] Create a test case for dynamic tensor support
by Michalis Spyrou
· 5 years ago
d87a7b2
COMPMID-2650: Add support for QASYMM16 in CLQuantizationLayer
by Michele Di Giorgio
· 5 years ago
e874ef9
COMPMID-2646: Wrap thread-related structure to ease bare-metal support
by Georgios Pinitas
· 5 years ago
8d5dd86
COMPMID-2659: Activation.SQRT Failure on NEON
by Michele Di Giorgio
· 5 years ago
9285adb
COMPMID-2599: Implement a new and generic depthwise convolution on OpenCL (Fp32/FP16-NHWC)
by Gian Marco Iodice
· 5 years ago
70ad619
COMPMID-2635: Add support for QASYMM8 in CPPBoxWithNonMaximaSuppressionLimit
by Michele Di Giorgio
· 5 years ago
9032ee3
MLCE-129: NEPad 30x slower than TensorFlow's implementation
by Manuel Bottini
· 5 years ago
ffd31de
COMPMID-2246: NEBoundingBoxTransform support QASYMM16
by Pablo Tello
· 5 years ago
12833d0
COMPMID-2204: RuntimeContext interface for NEON functions.
by Georgios Pinitas
· 5 years ago
ed0e35b
COMPMID-2246: Implement NEBoundingBoxTransform.
by Pablo Tello
· 5 years ago
4aff98f
COMPMID-2247: Extend support of CLBoundingBoxTransform for QUANT16_ASYMM
by Michele Di Giorgio
· 5 years ago
49be2e3
COMPMID-2644: Add is_dynamic flag to ITensorInfo.
by Georgios Pinitas
· 5 years ago
ae99b6e
COMPMID-1965 Extend CLGEMMMatrixMultiplyReshapedKernel to support transposed LHS (t) and not-transpose RHS
by Giorgio Arena
· 5 years ago
c2a6059
COMPMID-2639: CLPadLayer support for 4D padding
by Matthew Jackson
· 5 years ago
b19c55d
COMPMID-2640: Fix performance regression for Resnet101 Int8 on NEON
by Gian Marco Iodice
· 5 years ago
3f632f3
COMPMID-2418: CLDequantizationLayer support for QASYMM8_PER_CHANNEL
by Michalis Spyrou
· 5 years ago
351bd13
compmid-2573: Investigate FP16 Winograd reference implementations
by giuros01
· 5 years ago
ebe2e8c
COMPMID-2318: Implement NEROIAlignLayer
by Pablo Tello
· 5 years ago
d75f9e9
COMPMID-2318: Implement NEROIAlignLayer
by Pablo Tello
· 5 years ago
578a9fc
COMPMID-2317: Implement CLROIAlignLayer
by Michele Di Giorgio
· 5 years ago
b9070a4
COMPMID-2605: Add asymmetric padding support for Deconvolution layer
by Matthew Jackson
· 5 years ago
275f99c
COMPMID-2451: Use kernel lws_hint() on enqueue
by Georgios Pinitas
· 5 years ago
601ba3f
COMPMID-2336: Validate multiplier and offset calculation in LSTMQuantized and DepthwiseConvolution functions
by Michele Di Giorgio
· 5 years ago
3e183d9
COMPMID-2336: Update GEMM code
by Michalis Spyrou
· 5 years ago
35ea9a7
COMPMID-2317: Add new QASYMM16 data type
by Michele Di Giorgio
· 5 years ago
29a01c9
COMPMID-2417: NEDequantizationLayer support for QASYMM8_PER_CHANNEL
by Michalis Spyrou
· 5 years ago
fb0fdcd
COMPMID-2460: Add ELU activation on NEON,CL,GLES
by Georgios Pinitas
· 5 years ago
c853021
COMPMID-2417: Add new QASYMM8_PER_CHANNEL data type
by Michalis Spyrou
· 5 years ago
57a8961
COMPMID-2631: Deprecate CLGEMMMatrixAdditionKernel
by Gian Marco Iodice
· 5 years ago
ec8cce8
COMPMID-2602: FullyConnected VTS failure on CL and Neon TENSOR_QUANT8_ASYMM
by Gian Marco Iodice
· 5 years ago
205eed8
COMPMID-2080 Create a new kernel for CLPad with CONSTANT mode
by Giorgio Arena
· 5 years ago
400abc8
COMPMID-2596 MobilenetSSD produce wrong results
by Michalis Spyrou
· 5 years ago
f512580
COMPMID-2590: Avg_Pooling2d Fails on CL NHWC FP16
by Michele Di Giorgio
· 5 years ago
ecb1427
COMPMID-2492 Fix Segfault in NEDeconvolutionLayer for NHWC
by SiCong Li
· 5 years ago
1e6c915
COMPMID-2604: (Nightly) Failing 'GC/ConvolutionLayer/Configuration
by Gian Marco Iodice
· 5 years ago
9f0d5ec
COMPMID-2597: Checking bias type in NEDeconvolutionLayer::validate even when bias == nullptr
by Manuel Bottini
· 5 years ago
81d7e78
COMPMID-2595: (Nightly) Mismatches in NEArgMinMax for FP16
by Michele Di Giorgio
· 5 years ago
0c5191c
COMPMID-2336: Fix call to validate in NEDeconvolutionLayer
by Michele Di Giorgio
· 5 years ago
70e890e
COMPMID-2541 Remove DepthwiseSeparableConvolutionLayer for NEON/CL
by Giorgio Arena
· 5 years ago
f44fe3d
COMPMID-2568: NEON Convolution layer failure
by giuros01
· 5 years ago
f948b10
COMPMID-2582: Disable fuse batch normalization on OpenCL for FP16
by Gian Marco Iodice
· 5 years ago
9d3e7f9
COMPMID-2569: Fix access window in NEGEMMTranspose1xWKernel and NEGEMMLowpMatrixMultiplyKernel
by Michele Di Giorgio
· 5 years ago
3ae323f
COMPMID-2580: Fix out of bound read in Depthwise Convolution layer (OpenCL)
by Gian Marco Iodice
· 5 years ago
93c70b8
COMPMID-2547: CLSpaceToBatchLayer causes NN Test Failures on QUANT8_ASYMM Data Type
by Michele Di Giorgio
· 5 years ago
1720358
COMPMID-2336 Extend validation for depthwise native and fix same pad calculator
by Giorgio Arena
· 5 years ago
7786f25
COMPMID-2526: Disabled Winograd when padding > half_kernel_size
by Pablo Tello
· 5 years ago
a7acb3c
COMPMID-1849: Implement CPPDetectionPostProcessLayer
by Isabella Gottardi
· 6 years ago
4746326
COMPMID-2336 Fix valgrind error for BatchNormalizationLayer on NEON with NHWC layout
by Giorgio Arena
· 5 years ago
d820db6
COMPMID-2545: Reduce tests required by GEMM (OpenCL)
by Gian Marco Iodice
· 5 years ago
b238f5f
COMPMID-2539: Add bias addition check in CLGEMM validation
by Gian Marco Iodice
· 5 years ago
607339f
COMPMID-2527: Remove activation function validation in CLWinogradConvolutionLayer
by Gian Marco Iodice
· 5 years ago
1c29ffc
COMPMID-2336: Fix build issues.
by Georgios Pinitas
· 5 years ago
169cda3
COMPMID-2055: Fusion of ConvolutionLayer with BatchNormalization at graph only for CL
by Manuel Bottini
· 5 years ago
f3622be
COMPMID-1979: Fuse Activation Function in CLGEMM - part 4
by Gian Marco Iodice
· 5 years ago
c4d5136
COMPMID-2493: Update qs8 in Depthwise assembly
by Georgios Pinitas
· 5 years ago
4d600c7
COMPMID-2500: Report error in Int8 Conv2d if multiplier > 1.
by Georgios Pinitas
· 5 years ago
a231ea6
COMPMID-2336: Fix GEMMLowp 32bit check
by Georgios Pinitas
· 5 years ago
5d42b46
COMPMID-2467 Include IP information to the tuning config
by Giorgio Arena
· 5 years ago
cf3abde
COMPMID-2430: Adding FusedDepthwiseConvolutionBatchNormalizationLayer condition in NEON function factory
by Manuel Bottini
· 5 years ago
f33484f
COMPMID-2336: Guard quint8 gemm files for aarch64 only.
by Georgios Pinitas
· 5 years ago
883bad7
COMPMID-1849: Add DetectorPostProcess operator
by Isabella Gottardi
· 5 years ago
bd9097d
COMPMID-2336: Rename the new generic depthwise convolution on NEON
by Gian Marco Iodice
· 5 years ago
44f5572
COMPMID-2179 New generic depthwise convolution for NEON F32 NHWC
by Giorgio Arena
· 5 years ago
c050e0c
COMPMID-2490: (Nightly) Valgrind error in NEON Scale
by Michalis Spyrou
· 5 years ago
cfa2bba
COMPMID-2178: Update GEMM assembly code.
by Georgios Pinitas
· 5 years ago
f59b16f
COMPMID-2336: Set the correct valid region to the concatenation layer output
by Isabella Gottardi
· 5 years ago
7faa0ec
Use .inst directives instead of .int directives.
by Ramana Radhakrishnan
· 5 years ago
cb6858f
Update copyright years
by Ramana Radhakrishnan
· 5 years ago
d1f5476
COMPMID-1979: Fuse Activation Function in CLGEMM - part 3
by Gian Marco Iodice
· 5 years ago
10c53f1
COMPMID-2307: QUANTIZED_16BIT_LSTM operator for CL
by Manuel Bottini
· 5 years ago
d25af67
COMPMID-2456: NEDeconvolutionLayer.cpp, NHWC is not supported
by Manuel Bottini
· 5 years ago
06be6f8
COMPMID-2096: Refactor the CLGEMMLowp function selection (heuristic)
by Gian Marco Iodice
· 5 years ago
ff27192
COMPMID-2336: Fix validation for quantized NEDepthwiseConvolutionLayer
by Michele Di Giorgio
· 5 years ago
ca1f460
COMPMID-1979: Fuse Activation Function in CLGEMM - part 2
by Gian Marco Iodice
· 5 years ago
c95988a
COMPMID-2336: Change GPUTarget regex
by Michalis Spyrou
· 5 years ago
ed75326
COMPMID-2283: Implement SIN operator for NEON
by Manuel Bottini
· 5 years ago
Next »