arm_compute v20.08

Author: Jenkins

Android.bp

@@ -1,5 +1,5 @@
 //
-// Copyright © 2020 ARM Ltd. All rights reserved.
+// Copyright © 2020 Arm Ltd. All rights reserved.
 // SPDX-License-Identifier: MIT
 //
 
@@ -45,7 +45,8 @@ cc_library_static {
                          "arm_compute/core/NEON/kernels/assembly",
                          "arm_compute/core/NEON/kernels/convolution/common",
                          "arm_compute/core/NEON/kernels/convolution/depthwise",
-                         "arm_compute/core/NEON/kernels/convolution/winograd"],
+                         "src/core/NEON/kernels/assembly",
+                         "src/core/NEON/kernels/convolution/winograd"],
     export_include_dirs: [".", "./include"],
     srcs: [
         "src/core/AccessWindowAutoPadding.cpp",
@@ -55,6 +56,7 @@ cc_library_static {
         "src/core/CL/CLCoreRuntimeContext.cpp",
         "src/core/CL/CLHelpers.cpp",
         "src/core/CL/CLKernelLibrary.cpp",
+        "src/core/CL/CLUtils.cpp",
         "src/core/CL/ICLDistribution1D.cpp",
         "src/core/CL/ICLHOG.cpp",
         "src/core/CL/ICLKernel.cpp",
@@ -133,7 +135,6 @@ cc_library_static {
         "src/core/CL/kernels/CLGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel.cpp",
         "src/core/CL/kernels/CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.cpp",
         "src/core/CL/kernels/CLGEMMLowpReductionKernel.cpp",
-        "src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp",
         "src/core/CL/kernels/CLGEMMMatrixMultiplyKernel.cpp",
         "src/core/CL/kernels/CLGEMMMatrixMultiplyNativeKernel.cpp",
         "src/core/CL/kernels/CLGEMMMatrixMultiplyReshapedKernel.cpp",
@@ -158,6 +159,7 @@ cc_library_static {
         "src/core/CL/kernels/CLLKTrackerKernel.cpp",
         "src/core/CL/kernels/CLLocallyConnectedMatrixMultiplyKernel.cpp",
         "src/core/CL/kernels/CLMagnitudePhaseKernel.cpp",
+        "src/core/CL/kernels/CLMaxUnpoolingLayerKernel.cpp",
         "src/core/CL/kernels/CLMeanStdDevKernel.cpp",
         "src/core/CL/kernels/CLMeanStdDevNormalizationKernel.cpp",
         "src/core/CL/kernels/CLMedian3x3Kernel.cpp",
@@ -228,6 +230,7 @@ cc_library_static {
         "src/core/IDistribution1D.cpp",
         "src/core/IKernel.cpp",
         "src/core/ITensor.cpp",
+        "src/core/ITensorPack.cpp",
         "src/core/MultiImageInfo.cpp",
         "src/core/NEON/kernels/NEAbsoluteDifferenceKernel.cpp",
         "src/core/NEON/kernels/NEAccumulateKernel.cpp",
@@ -274,7 +277,6 @@ cc_library_static {
         "src/core/NEON/kernels/NEFastCornersKernel.cpp",
         "src/core/NEON/kernels/NEFillArrayKernel.cpp",
         "src/core/NEON/kernels/NEFillBorderKernel.cpp",
-        "src/core/NEON/kernels/NEFillInnerBorderKernel.cpp",
         "src/core/NEON/kernels/NEFlattenLayerKernel.cpp",
         "src/core/NEON/kernels/NEFloorKernel.cpp",
         "src/core/NEON/kernels/NEFuseBatchNormalizationKernel.cpp",
@@ -287,10 +289,8 @@ cc_library_static {
         "src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToInt8ScaleByFixedPointKernel.cpp",
         "src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.cpp",
         "src/core/NEON/kernels/NEGEMMLowpReductionKernel.cpp",
-        "src/core/NEON/kernels/NEGEMMMatrixAccumulateBiasesKernel.cpp",
         "src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp",
         "src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp",
-        "src/core/NEON/kernels/NEGEMMMatrixVectorMultiplyKernel.cpp",
         "src/core/NEON/kernels/NEGEMMTranspose1xWKernel.cpp",
         "src/core/NEON/kernels/NEGatherKernel.cpp",
         "src/core/NEON/kernels/NEGaussian3x3Kernel.cpp",
@@ -309,6 +309,7 @@ cc_library_static {
         "src/core/NEON/kernels/NELKTrackerKernel.cpp",
         "src/core/NEON/kernels/NELocallyConnectedMatrixMultiplyKernel.cpp",
         "src/core/NEON/kernels/NEMagnitudePhaseKernel.cpp",
+        "src/core/NEON/kernels/NEMaxUnpoolingLayerKernel.cpp",
         "src/core/NEON/kernels/NEMeanStdDevKernel.cpp",
         "src/core/NEON/kernels/NEMeanStdDevNormalizationKernel.cpp",
         "src/core/NEON/kernels/NEMedian3x3Kernel.cpp",
@@ -366,7 +367,6 @@ cc_library_static {
         "src/core/NEON/kernels/arm_gemm/mergeresults.cpp",
         "src/core/NEON/kernels/arm_gemm/misc.cpp",
         "src/core/NEON/kernels/arm_gemm/quantized.cpp",
-        "src/core/NEON/kernels/assembly/Helpers.cpp",
         "src/core/NEON/kernels/assembly/INEGEMMWrapperKernel.cpp",
         "src/core/NEON/kernels/convolution/common/padding.cpp",
         "src/core/NEON/kernels/convolution/common/qasymm8.cpp",
@@ -413,6 +413,7 @@ cc_library_static {
         "src/core/Utils.cpp",
         "src/core/Validate.cpp",
         "src/core/Version.cpp",
+        "src/core/utils/ScaleUtils.cpp",
         "src/core/utils/helpers/fft.cpp",
         "src/core/utils/helpers/tensor_transform.cpp",
         "src/core/utils/io/FileHandler.cpp",
@@ -435,6 +436,7 @@ cc_library_static {
         "src/runtime/CL/CLMemoryRegion.cpp",
         "src/runtime/CL/CLMultiHOG.cpp",
         "src/runtime/CL/CLMultiImage.cpp",
+        "src/runtime/CL/CLOperator.cpp",
         "src/runtime/CL/CLPyramid.cpp",
         "src/runtime/CL/CLRuntimeContext.cpp",
         "src/runtime/CL/CLScheduler.cpp",
@@ -519,6 +521,7 @@ cc_library_static {
         "src/runtime/CL/functions/CLLaplacianReconstruct.cpp",
         "src/runtime/CL/functions/CLLocallyConnectedLayer.cpp",
         "src/runtime/CL/functions/CLMagnitude.cpp",
+        "src/runtime/CL/functions/CLMaxUnpoolingLayer.cpp",
         "src/runtime/CL/functions/CLMeanStdDev.cpp",
         "src/runtime/CL/functions/CLMeanStdDevNormalizationLayer.cpp",
         "src/runtime/CL/functions/CLMedian3x3.cpp",
@@ -603,6 +606,7 @@ cc_library_static {
         "src/runtime/MemoryManagerOnDemand.cpp",
         "src/runtime/MultiHOG.cpp",
         "src/runtime/MultiImage.cpp",
+        "src/runtime/NEON/INEOperator.cpp",
         "src/runtime/NEON/INESimpleFunction.cpp",
         "src/runtime/NEON/INESimpleFunctionNoBorder.cpp",
         "src/runtime/NEON/functions/NEAbsoluteDifference.cpp",
@@ -685,6 +689,7 @@ cc_library_static {
         "src/runtime/NEON/functions/NELaplacianReconstruct.cpp",
         "src/runtime/NEON/functions/NELocallyConnectedLayer.cpp",
         "src/runtime/NEON/functions/NEMagnitude.cpp",
+        "src/runtime/NEON/functions/NEMaxUnpoolingLayer.cpp",
         "src/runtime/NEON/functions/NEMeanStdDev.cpp",
         "src/runtime/NEON/functions/NEMeanStdDevNormalizationLayer.cpp",
         "src/runtime/NEON/functions/NEMedian3x3.cpp",
@@ -740,6 +745,7 @@ cc_library_static {
         "src/runtime/OMP/OMPScheduler.cpp",
         "src/runtime/OffsetLifetimeManager.cpp",
         "src/runtime/OffsetMemoryPool.cpp",
+        "src/runtime/OperatorTensor.cpp",
         "src/runtime/PoolManager.cpp",
         "src/runtime/Pyramid.cpp",
         "src/runtime/RuntimeContext.cpp",
@@ -768,31 +774,35 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s16_12x8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_12x8/a55r1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_12x8/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_12x8/x1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_s8_4x4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u16_12x8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_12x8/a55r1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_12x8/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_12x8/x1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_gemm_u8_4x4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/a55r1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_hgemm_24x8/x1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_16x4/a55.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_16x4/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_16x4/x1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_fp32_mla_4x8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8s32_dot_16x4/a55.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_s8s32_dot_16x4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8u32_dot_16x4/a55.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8u32_dot_16x4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_dot_12x8/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_dot_12x8/x1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_mmla_12x8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_12x8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_u8u32_mmla_12x8/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/a64_native_fp32_mla_16x4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_12x8/a53.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_12x8/a55.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_12x8/a55r1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_12x8/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_12x8/x1.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemv_pretransposed/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemv_trans/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_fp32_mla_4x6/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_fp32_mla_4x8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_smallK_hybrid_s8s32_dot_4x6/a55.cpp",
@@ -809,21 +819,18 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_bf16fp32_mmla_8VLx2/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_fp16_mla_4VLx4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_fp32_mla_4VLx4/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_fp32_mmla_4VLx4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_s8s32_dot_4VLx4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_hybrid_u8u32_dot_4VLx4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_bf16fp32_dot_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_bf16fp32_mmla_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_fp16_mla_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_fp32_mla_3VLx8/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_fp32_mmla_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_s8s32_dot_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_s8s32_mmla_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_u8u32_dot_3VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_interleaved_u8u32_mmla_3VLx8/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_native_bf16fp32_dot_4VLx4/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_native_fp16_mla_4VLx4/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_native_fp32_mla_4VLx4/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_native_s8s32_dot_4VLx4/generic.cpp",
-                "src/core/NEON/kernels/arm_gemm/kernels/sve_native_u8u32_dot_4VLx4/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_fp32_mla_1VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_s8s32_dot_1VLx8/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/sve_smallK_hybrid_u8u32_dot_1VLx8/generic.cpp",

LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2017-2020 ARM Software
+Copyright (c) 2017-2020 Arm Limited
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

SConscript

@@ -1,4 +1,4 @@
-# Copyright (c) 2016, 2017 ARM Limited.
+# Copyright (c) 2016, 2017 Arm Limited.
 #
 # SPDX-License-Identifier: MIT
 #
@@ -24,9 +24,9 @@ import os.path
 import re
 import subprocess
 
-VERSION = "v20.05"
-LIBRARY_VERSION_MAJOR = 19
-LIBRARY_VERSION_MINOR =  1
+VERSION = "v20.08"
+LIBRARY_VERSION_MAJOR = 20
+LIBRARY_VERSION_MINOR =  0
 LIBRARY_VERSION_PATCH =  0
 SONAME_VERSION = str(LIBRARY_VERSION_MAJOR) + "." + str(LIBRARY_VERSION_MINOR) + "." + str(LIBRARY_VERSION_PATCH)
 
@@ -178,6 +178,7 @@ arm_compute_env.Append(LIBS = ['dl'])
 core_files = Glob('src/core/*.cpp')
 core_files += Glob('src/core/CPP/*.cpp')
 core_files += Glob('src/core/CPP/kernels/*.cpp')
+core_files += Glob('src/core/utils/*.cpp')
 core_files += Glob('src/core/utils/helpers/*.cpp')
 core_files += Glob('src/core/utils/io/*.cpp')
 core_files += Glob('src/core/utils/quantization/*.cpp')
@@ -230,6 +231,8 @@ if env['neon']:
     arm_compute_env.Append(CPPPATH = ["arm_compute/core/NEON/kernels/convolution/common/",
                                       "arm_compute/core/NEON/kernels/convolution/winograd/",
                                       "arm_compute/core/NEON/kernels/convolution/depthwise/",
+                                      "src/core/NEON/kernels/assembly/",
+                                      "src/core/NEON/kernels/convolution/winograd/",
                                       "arm_compute/core/NEON/kernels/assembly/"])
 
     graph_files += Glob('src/graph/backends/NEON/*.cpp')

SConstruct

@@ -1,4 +1,4 @@
-# Copyright (c) 2016, 2017 ARM Limited.
+# Copyright (c) 2016, 2017 Arm Limited.
 #
 # SPDX-License-Identifier: MIT
 #
@@ -204,7 +204,7 @@ elif 'v8' in env['arch']:
         env.Append(CXXFLAGS = ['-march=armv8-a'])
 
     if 'v8.6-a' in env['arch']:
-        env.Append(CPPDEFINES = ['V8P6', 'V8P6_BF', 'ARM_COMPUTE_FORCE_BF16'])
+        env.Append(CPPDEFINES = ['MMLA_INT8', 'MMLA_FP32', 'V8P6', 'V8P6_BF', 'ARM_COMPUTE_FORCE_BF16'])
 
 elif 'x86' in env['arch']:
     if env['estate'] == '32':

arm_compute/core/AccessWindowAutoPadding.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2019 ARM Limited.
+ * Copyright (c) 2017-2019 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *

arm_compute/core/AccessWindowStatic.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2019 ARM Limited.
+ * Copyright (c) 2017-2019 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *

arm_compute/core/AccessWindowTranspose.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017-2019 ARM Limited.
+ * Copyright (c) 2017-2019 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *

arm_compute/core/CL/CLCompileContext.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020 ARM Limited.
+ * Copyright (c) 2020 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *

arm_compute/core/CL/CLCoreRuntimeContext.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2019 ARM Limited.
+ * Copyright (c) 2019 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *

arm_compute/core/CL/CLDevice.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2020 ARM Limited.
+ * Copyright (c) 2020 Arm Limited.
  *
  * SPDX-License-Identifier: MIT
  *