Compute Library v22.02

ARM-software · Feb 26, 2022 · 8f587de · 8f587de
1 parent 91ee4d0
commit 8f587de
Show file tree

Hide file tree

Showing 400 changed files with 14,362 additions and 11,090 deletions.
diff --git a/Android.bp b/Android.bp
@@ -40,6 +40,7 @@ opencl_srcs = [
         "src/core/CL/cl_kernels/common/floor.cl",
         "src/core/CL/cl_kernels/common/gather.cl",
         "src/core/CL/cl_kernels/common/gemm.cl",
+        "src/core/CL/cl_kernels/common/gemm_utils.cl",
         "src/core/CL/cl_kernels/common/gemmlowp.cl",
         "src/core/CL/cl_kernels/common/gemv.cl",
         "src/core/CL/cl_kernels/common/generate_proposals.cl",
@@ -80,17 +81,13 @@ opencl_srcs = [
         "src/core/CL/cl_kernels/nchw/channel_shuffle.cl",
         "src/core/CL/cl_kernels/nchw/depth_to_space.cl",
         "src/core/CL/cl_kernels/nchw/dequantization_layer.cl",
-        "src/core/CL/cl_kernels/nchw/direct_convolution1x1.cl",
-        "src/core/CL/cl_kernels/nchw/direct_convolution3x3.cl",
-        "src/core/CL/cl_kernels/nchw/direct_convolution5x5.cl",
-        "src/core/CL/cl_kernels/nchw/direct_convolution_quantized.cl",
+        "src/core/CL/cl_kernels/nchw/direct_convolution.cl",
         "src/core/CL/cl_kernels/nchw/im2col.cl",
         "src/core/CL/cl_kernels/nchw/normalization_layer.cl",
         "src/core/CL/cl_kernels/nchw/normalize_planar_yuv_layer.cl",
         "src/core/CL/cl_kernels/nchw/normalize_planar_yuv_layer_quantized.cl",
         "src/core/CL/cl_kernels/nchw/pooling_layer.cl",
         "src/core/CL/cl_kernels/nchw/prior_box_layer.cl",
-        "src/core/CL/cl_kernels/nchw/remap.cl",
         "src/core/CL/cl_kernels/nchw/reorg_layer.cl",
         "src/core/CL/cl_kernels/nchw/scale.cl",
         "src/core/CL/cl_kernels/nchw/space_to_batch.cl",
@@ -114,7 +111,6 @@ opencl_srcs = [
         "src/core/CL/cl_kernels/nhwc/normalize_planar_yuv_layer_quantized.cl",
         "src/core/CL/cl_kernels/nhwc/pooling_layer.cl",
         "src/core/CL/cl_kernels/nhwc/pooling_layer_quantized.cl",
-        "src/core/CL/cl_kernels/nhwc/remap.cl",
         "src/core/CL/cl_kernels/nhwc/reorg_layer.cl",
         "src/core/CL/cl_kernels/nhwc/scale.cl",
         "src/core/CL/cl_kernels/nhwc/space_to_batch.cl",
@@ -241,7 +237,6 @@ cc_library_static {
         "src/core/CL/kernels/CLROIPoolingLayerKernel.cpp",
         "src/core/CL/kernels/CLRangeKernel.cpp",
         "src/core/CL/kernels/CLReductionOperationKernel.cpp",
-        "src/core/CL/kernels/CLRemapKernel.cpp",
         "src/core/CL/kernels/CLReorgLayerKernel.cpp",
         "src/core/CL/kernels/CLReverseKernel.cpp",
         "src/core/CL/kernels/CLSelectKernel.cpp",
@@ -293,7 +288,6 @@ cc_library_static {
         "src/core/NEON/kernels/NEROIPoolingLayerKernel.cpp",
         "src/core/NEON/kernels/NERangeKernel.cpp",
         "src/core/NEON/kernels/NEReductionOperationKernel.cpp",
-        "src/core/NEON/kernels/NERemapKernel.cpp",
         "src/core/NEON/kernels/NEReorgLayerKernel.cpp",
         "src/core/NEON/kernels/NEReverseKernel.cpp",
         "src/core/NEON/kernels/NESelectKernel.cpp",
@@ -336,8 +330,6 @@ cc_library_static {
         "src/core/NEON/kernels/arm_gemm/transform.cpp",
         "src/core/NEON/kernels/batchnormalization/impl/NEON/fp16.cpp",
         "src/core/NEON/kernels/batchnormalization/impl/NEON/fp32.cpp",
-        "src/core/NEON/kernels/batchnormalization/impl/SVE/fp16.cpp",
-        "src/core/NEON/kernels/batchnormalization/impl/SVE/fp32.cpp",
         "src/core/NEON/kernels/convolution/common/padding.cpp",
         "src/core/NEON/kernels/convolution/common/qasymm8.cpp",
         "src/core/NEON/kernels/convolution/common/qsymm8.cpp",
@@ -431,44 +423,79 @@ cc_library_static {
         "src/cpu/kernels/CpuTransposeKernel.cpp",
         "src/cpu/kernels/CpuWeightsReshapeKernel.cpp",
         "src/cpu/kernels/CpuWinogradConv2dKernel.cpp",
-        "src/cpu/kernels/activation/neon/fp16.cpp",
-        "src/cpu/kernels/activation/neon/fp32.cpp",
-        "src/cpu/kernels/activation/neon/qasymm8.cpp",
-        "src/cpu/kernels/activation/neon/qasymm8_signed.cpp",
-        "src/cpu/kernels/activation/neon/qsymm16.cpp",
-        "src/cpu/kernels/activation/sve/fp16.cpp",
-        "src/cpu/kernels/activation/sve/fp32.cpp",
-        "src/cpu/kernels/activation/sve/qasymm8.cpp",
-        "src/cpu/kernels/activation/sve/qasymm8_signed.cpp",
-        "src/cpu/kernels/activation/sve/qsymm16.cpp",
-        "src/cpu/kernels/add/neon/qasymm8.cpp",
-        "src/cpu/kernels/add/neon/qasymm8_signed.cpp",
-        "src/cpu/kernels/add/neon/qsymm16.cpp",
-        "src/cpu/kernels/add/sve/impl.cpp",
-        "src/cpu/kernels/add/sve/qasymm8.cpp",
-        "src/cpu/kernels/add/sve/qasymm8_signed.cpp",
-        "src/cpu/kernels/add/sve/qsymm16.cpp",
-        "src/cpu/kernels/elementwise/sve/elementwise.cpp",
-        "src/cpu/kernels/elementwise/sve/elementwise_unary.cpp",
+        "src/cpu/kernels/activation/generic/neon/fp16.cpp",
+        "src/cpu/kernels/activation/generic/neon/fp32.cpp",
+        "src/cpu/kernels/activation/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/activation/generic/neon/qasymm8_signed.cpp",
+        "src/cpu/kernels/activation/generic/neon/qsymm16.cpp",
+        "src/cpu/kernels/add/generic/neon/fp16.cpp",
+        "src/cpu/kernels/add/generic/neon/fp32.cpp",
+        "src/cpu/kernels/add/generic/neon/impl.cpp",
+        "src/cpu/kernels/add/generic/neon/integer.cpp",
+        "src/cpu/kernels/add/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/add/generic/neon/qasymm8_signed.cpp",
+        "src/cpu/kernels/add/generic/neon/qsymm16.cpp",
+        "src/cpu/kernels/boundingboxtransform/generic/neon/fp16.cpp",
+        "src/cpu/kernels/boundingboxtransform/generic/neon/fp32.cpp",
+        "src/cpu/kernels/boundingboxtransform/generic/neon/impl.cpp",
+        "src/cpu/kernels/boundingboxtransform/generic/neon/qsymm16.cpp",
+        "src/cpu/kernels/crop/generic/neon/fp16.cpp",
+        "src/cpu/kernels/crop/generic/neon/fp32.cpp",
+        "src/cpu/kernels/crop/generic/neon/impl.cpp",
+        "src/cpu/kernels/crop/generic/neon/integer.cpp",
+        "src/cpu/kernels/elementwise_binary/generic/neon/fp16.cpp",
+        "src/cpu/kernels/elementwise_binary/generic/neon/fp32.cpp",
+        "src/cpu/kernels/elementwise_binary/generic/neon/integer.cpp",
+        "src/cpu/kernels/elementwise_binary/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/elementwise_binary/generic/neon/qasymm8_signed.cpp",
+        "src/cpu/kernels/elementwise_unary/generic/neon/fp16.cpp",
+        "src/cpu/kernels/elementwise_unary/generic/neon/fp32.cpp",
+        "src/cpu/kernels/elementwise_unary/generic/neon/integer.cpp",
         "src/cpu/kernels/floor/neon/fp16.cpp",
         "src/cpu/kernels/floor/neon/fp32.cpp",
+        "src/cpu/kernels/genproposals/generic/neon/fp16.cpp",
+        "src/cpu/kernels/genproposals/generic/neon/fp32.cpp",
+        "src/cpu/kernels/genproposals/generic/neon/impl.cpp",
+        "src/cpu/kernels/genproposals/generic/neon/qsymm16.cpp",
+        "src/cpu/kernels/instancenorm/generic/neon/fp16.cpp",
+        "src/cpu/kernels/instancenorm/generic/neon/fp32.cpp",
+        "src/cpu/kernels/instancenorm/generic/neon/impl.cpp",
         "src/cpu/kernels/internal/CpuDepthwiseConv2dAssemblyWrapperKernel.cpp",
         "src/cpu/kernels/internal/CpuPool2dAssemblyWrapperKernel.cpp",
+        "src/cpu/kernels/maxunpool/generic/neon/fp16.cpp",
+        "src/cpu/kernels/maxunpool/generic/neon/fp32.cpp",
+        "src/cpu/kernels/maxunpool/generic/neon/impl.cpp",
+        "src/cpu/kernels/maxunpool/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/maxunpool/generic/neon/qasymm8_signed.cpp",
+        "src/cpu/kernels/meanstddevnorm/generic/neon/fp16.cpp",
+        "src/cpu/kernels/meanstddevnorm/generic/neon/fp32.cpp",
+        "src/cpu/kernels/meanstddevnorm/generic/neon/impl.cpp",
         "src/cpu/kernels/pool2d/neon/fp16.cpp",
         "src/cpu/kernels/pool2d/neon/fp32.cpp",
         "src/cpu/kernels/pool2d/neon/nchw/all.cpp",
         "src/cpu/kernels/pool2d/neon/qasymm8.cpp",
         "src/cpu/kernels/pool2d/neon/qasymm8_signed.cpp",
+        "src/cpu/kernels/range/generic/neon/fp16.cpp",
+        "src/cpu/kernels/range/generic/neon/fp32.cpp",
+        "src/cpu/kernels/range/generic/neon/impl.cpp",
+        "src/cpu/kernels/range/generic/neon/integer.cpp",
+        "src/cpu/kernels/roialign/generic/neon/fp16.cpp",
+        "src/cpu/kernels/roialign/generic/neon/fp32.cpp",
+        "src/cpu/kernels/roialign/generic/neon/impl.cpp",
+        "src/cpu/kernels/roialign/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/roialign/generic/neon/qasymm8_signed.cpp",
         "src/cpu/kernels/scale/neon/fp16.cpp",
         "src/cpu/kernels/scale/neon/integer.cpp",
         "src/cpu/kernels/scale/neon/qasymm8.cpp",
         "src/cpu/kernels/scale/neon/qasymm8_signed.cpp",
-        "src/cpu/kernels/scale/sve/fp16.cpp",
-        "src/cpu/kernels/scale/sve/fp32.cpp",
-        "src/cpu/kernels/scale/sve/integer.cpp",
-        "src/cpu/kernels/scale/sve/qasymm8.cpp",
-        "src/cpu/kernels/scale/sve/qasymm8_signed.cpp",
-        "src/cpu/kernels/softmax/impl/sve/impl.cpp",
+        "src/cpu/kernels/select/generic/neon/fp16.cpp",
+        "src/cpu/kernels/select/generic/neon/fp32.cpp",
+        "src/cpu/kernels/select/generic/neon/impl.cpp",
+        "src/cpu/kernels/select/generic/neon/integer.cpp",
+        "src/cpu/kernels/softmax/generic/neon/fp16.cpp",
+        "src/cpu/kernels/softmax/generic/neon/fp32.cpp",
+        "src/cpu/kernels/softmax/generic/neon/qasymm8.cpp",
+        "src/cpu/kernels/softmax/generic/neon/qasymm8_signed.cpp",
         "src/cpu/kernels/sub/neon/qasymm8.cpp",
         "src/cpu/kernels/sub/neon/qasymm8_signed.cpp",
         "src/cpu/kernels/sub/neon/qsymm16.cpp",
@@ -683,7 +710,6 @@ cc_library_static {
         "src/runtime/CL/functions/CLRange.cpp",
         "src/runtime/CL/functions/CLReduceMean.cpp",
         "src/runtime/CL/functions/CLReductionOperation.cpp",
-        "src/runtime/CL/functions/CLRemap.cpp",
         "src/runtime/CL/functions/CLReorgLayer.cpp",
         "src/runtime/CL/functions/CLReshapeLayer.cpp",
         "src/runtime/CL/functions/CLReverse.cpp",
@@ -794,7 +820,6 @@ cc_library_static {
         "src/runtime/NEON/functions/NERange.cpp",
         "src/runtime/NEON/functions/NEReduceMean.cpp",
         "src/runtime/NEON/functions/NEReductionOperation.cpp",
-        "src/runtime/NEON/functions/NERemap.cpp",
         "src/runtime/NEON/functions/NEReorgLayer.cpp",
         "src/runtime/NEON/functions/NEReshapeLayer.cpp",
         "src/runtime/NEON/functions/NEReverse.cpp",
@@ -1013,8 +1038,11 @@ cc_library_static {
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_hybrid_u8u32_mmla_6x16/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_dot_8x12/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_dot_8x12/x1.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_mmla_8x12/a510.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_bf16fp32_mmla_8x12/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_8x12/a510.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_s8s32_mmla_8x12/generic.cpp",
+                "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_u8u32_mmla_8x12/a510.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_interleaved_u8u32_mmla_8x12/generic.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/a53.cpp",
                 "src/core/NEON/kernels/arm_gemm/kernels/a64_sgemm_8x12/a55.cpp",

diff --git a/LICENSE b/LICENSE
@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2017-2021 Arm Limited
+Copyright (c) 2017-2022 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