diff options
| author | Michalis Spyrou <michalis.spyrou@arm.com> | 2019-04-26 14:54:54 +0100 |
|---|---|---|
| committer | Michalis Spyrou <michalis.spyrou@arm.com> | 2019-05-01 10:06:58 +0000 |
| commit | a4f378dcd39addd4a63db1c0848f2c120804f4eb (patch) | |
| tree | 6fa8a0071bef32d2bdef0e5469678a7cfecea348 | |
| parent | 8ec0bb6d9027bb7505d6fa0eada42a52c6e1073b (diff) | |
| download | ComputeLibrary-a4f378dcd39addd4a63db1c0848f2c120804f4eb.tar.gz | |
COMPMID-1995: Fix clang-tidy warnings
- Remove VirtualCall checks
- Fix some unused variables errors
- Use std::array insted of C style arrays
- Various fixes
Change-Id: Ife6170b7102de42b8f04e298dcf8476bf90779f0
Signed-off-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1049
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
77 files changed, 379 insertions, 352 deletions
diff --git a/.clang-tidy b/.clang-tidy index 4d1313bd6f..ad2dbb7a19 100644 --- a/.clang-tidy +++ b/.clang-tidy @@ -1,5 +1,5 @@ --- -Checks: 'clang-diagnostic-*,clang-analyzer-*,*,-abseil-*,-fuchsia-*,-bugprone-*,-hicpp-*,-cppcoreguidelines-pro-bounds-pointer-arithmetic,-cppcoreguidelines-pro-bounds-array-to-pointer-decay,-cppcoreguidelines-pro-bounds-constant-array-index,-cert-err58-cpp,-cppcoreguidelines-pro-type-reinterpret-cast,-google-runtime-references,-google-build-using-namespace,-readability-redundant-member-init,-readability-redundant-declaration,-readability-else-after-return,-performance-type-promotion-in-math-fn,-cert-err60-cpp,-cppcoreguidelines-narrowing-conversions,-readability-magic-numbers,-cppcoreguidelines-avoid-magic-numbers,-readability-named-parameter,-readability-implicit-bool-conversion,-readability-uppercase-literal-suffix' +Checks: 'clang-diagnostic-*,clang-analyzer-*,*,-abseil-*,-fuchsia-*,-bugprone-*,-hicpp-*,-cppcoreguidelines-pro-bounds-pointer-arithmetic,-cppcoreguidelines-pro-bounds-array-to-pointer-decay,-cppcoreguidelines-pro-bounds-constant-array-index,-cert-err58-cpp,-cppcoreguidelines-pro-type-reinterpret-cast,-google-runtime-references,-google-build-using-namespace,-readability-redundant-member-init,-readability-redundant-declaration,-readability-else-after-return,-performance-type-promotion-in-math-fn,-cert-err60-cpp,-cppcoreguidelines-narrowing-conversions,-readability-magic-numbers,-cppcoreguidelines-avoid-magic-numbers,-readability-named-parameter,-readability-implicit-bool-conversion,-readability-uppercase-literal-suffix,-clang-analyzer-optin.cplusplus.VirtualCall' WarningsAsErrors: '' HeaderFilterRegex: '' AnalyzeTemporaryDtors: false diff --git a/arm_compute/runtime/CL/functions/CLCropResize.h b/arm_compute/runtime/CL/functions/CLCropResize.h index d6c9fed928..9ec2071a4f 100644 --- a/arm_compute/runtime/CL/functions/CLCropResize.h +++ b/arm_compute/runtime/CL/functions/CLCropResize.h @@ -105,10 +105,10 @@ public: InterpolationPolicy _method; float _extrapolation_value; - std::unique_ptr<CLScale[]> _scale; - std::unique_ptr<CLCopyKernel[]> _copy; - std::unique_ptr<CLTensor[]> _crop_results{ nullptr }; - std::unique_ptr<CLTensor[]> _scaled_results{ nullptr }; + std::vector<std::unique_ptr<CLScale>> _scale; + std::vector<std::unique_ptr<CLCopyKernel>> _copy; + std::vector<std::unique_ptr<CLTensor>> _crop_results; + std::vector<std::unique_ptr<CLTensor>> _scaled_results; }; } // namespace arm_compute #endif /* __ARM_COMPUTE_CL_CROP_RESIZE_H__ */ diff --git a/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h b/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h index ad37ba51ab..1ddbacf327 100644 --- a/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h +++ b/arm_compute/runtime/NEON/functions/NEWinogradConvolutionLayer.h @@ -53,7 +53,7 @@ class NEWinogradConvolutionLayer : public IFunction { public: /** Constructor */ - NEWinogradConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager = nullptr); + NEWinogradConvolutionLayer(const std::shared_ptr<IMemoryManager> &memory_manager = nullptr); /** Set the input and output tensors. * diff --git a/examples/cl_convolution.cpp b/examples/cl_convolution.cpp index b15bbb6cb4..f2d19ef0cc 100644 --- a/examples/cl_convolution.cpp +++ b/examples/cl_convolution.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -36,7 +36,7 @@ using namespace utils; /** Gaussian 3x3 matrix */ -const int16_t gaussian3x3[] = +const std::array<int16_t, 9> gaussian3x3 = { 1, 2, 1, 2, 4, 2, @@ -45,7 +45,7 @@ const int16_t gaussian3x3[] = /** Gaussian 5x5 matrix */ -const int16_t gaussian5x5[] = +const std::array<int16_t, 25> gaussian5x5 = { 1, 4, 6, 4, 1, 4, 16, 24, 16, 4, @@ -82,8 +82,8 @@ public: dst.allocator()->init(*src.info()); // Apply a Gaussian 3x3 filter to the source image followed by a Gaussian 5x5: - conv3x3.configure(&src, &tmp, gaussian3x3, 0 /* Let arm_compute calculate the scale */, BorderMode::UNDEFINED); - conv5x5.configure(&tmp, &dst, gaussian5x5, 0 /* Let arm_compute calculate the scale */, BorderMode::UNDEFINED); + conv3x3.configure(&src, &tmp, gaussian3x3.data(), 0 /* Let arm_compute calculate the scale */, BorderMode::UNDEFINED); + conv5x5.configure(&tmp, &dst, gaussian5x5.data(), 0 /* Let arm_compute calculate the scale */, BorderMode::UNDEFINED); // Allocate all the images src.allocator()->allocate(); @@ -115,7 +115,11 @@ public: save_to_ppm(dst, output_filename); // save_to_ppm maps and unmaps the image to store as PPM } } - CLImage src{}, tmp{}, dst{}; + +private: + CLImage src{}; + CLImage tmp{}; + CLImage dst{}; CLConvolution3x3 conv3x3{}; CLConvolution5x5 conv5x5{}; std::string output_filename{}; diff --git a/examples/cl_sgemm.cpp b/examples/cl_sgemm.cpp index 805aec1cf3..8e0263dde2 100644 --- a/examples/cl_sgemm.cpp +++ b/examples/cl_sgemm.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -41,7 +41,9 @@ class CLSGEMMExample : public Example public: bool do_setup(int argc, char **argv) override { - NPYLoader npy0, npy1, npy2; + NPYLoader npy0; + NPYLoader npy1; + NPYLoader npy2; alpha = 1.0f; beta = 0.0f; @@ -184,7 +186,10 @@ public: } private: - CLTensor src0{}, src1{}, src2{}, dst{}; + CLTensor src0{}; + CLTensor src1{}; + CLTensor src2{}; + CLTensor dst{}; CLGEMM sgemm{}; CLTuner tuner{}; float alpha{}, beta{}; diff --git a/scripts/clang_tidy_rules.py b/scripts/clang_tidy_rules.py index a7c5051638..d35273d545 100755 --- a/scripts/clang_tidy_rules.py +++ b/scripts/clang_tidy_rules.py @@ -4,7 +4,7 @@ import re import sys def get_list_includes(): - return "include/linux include . 3rdparty/include kernels".split() + return "arm_compute/core/NEON/kernels/assembly arm_compute/core/NEON/kernels/convolution/common include/linux include . 3rdparty/include kernels".split() def get_list_flags( filename, arch): assert arch in ["armv7", "aarch64"] diff --git a/src/core/CL/kernels/CLPixelWiseMultiplicationKernel.cpp b/src/core/CL/kernels/CLPixelWiseMultiplicationKernel.cpp index 9fa92bde75..dda9b16083 100644 --- a/src/core/CL/kernels/CLPixelWiseMultiplicationKernel.cpp +++ b/src/core/CL/kernels/CLPixelWiseMultiplicationKernel.cpp @@ -274,7 +274,7 @@ BorderSize CLPixelWiseMultiplicationKernel::border_size() const { const unsigned int replicateSize = _output->info()->dimension(0) - std::min(_input1->info()->dimension(0), _input2->info()->dimension(0)); const unsigned int border = std::min<unsigned int>(num_elems_processed_per_iteration - 1U, replicateSize); - return BorderSize(0, border, 0, 0); + return BorderSize{ 0, border, 0, 0 }; } namespace @@ -409,6 +409,6 @@ BorderSize CLComplexPixelWiseMultiplicationKernel::border_size() const { const unsigned int replicateSize = _output->info()->dimension(0) - std::min(_input1->info()->dimension(0), _input2->info()->dimension(0)); const unsigned int border = std::min<unsigned int>(num_elems_processed_per_iteration_complex - 1U, replicateSize); - return BorderSize(0, border, 0, 0); + return BorderSize{ 0, border, 0, 0 }; } } // namespace arm_compute
\ No newline at end of file diff --git a/src/core/CL/kernels/CLSobel5x5Kernel.cpp b/src/core/CL/kernels/CLSobel5x5Kernel.cpp index 46aa074d61..be2a44b079 100644 --- a/src/core/CL/kernels/CLSobel5x5Kernel.cpp +++ b/src/core/CL/kernels/CLSobel5x5Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -138,7 +138,7 @@ CLSobel5x5VertKernel::CLSobel5x5VertKernel() BorderSize CLSobel5x5VertKernel::border_size() const { - return BorderSize(2, 0); + return BorderSize{ 2, 0 }; } void CLSobel5x5VertKernel::configure(const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined) diff --git a/src/core/CL/kernels/CLSobel7x7Kernel.cpp b/src/core/CL/kernels/CLSobel7x7Kernel.cpp index 0c94e88acf..a4a20c1863 100644 --- a/src/core/CL/kernels/CLSobel7x7Kernel.cpp +++ b/src/core/CL/kernels/CLSobel7x7Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -141,7 +141,7 @@ CLSobel7x7VertKernel::CLSobel7x7VertKernel() BorderSize CLSobel7x7VertKernel::border_size() const { - return BorderSize(3, 0); + return BorderSize{ 3, 0 }; } void CLSobel7x7VertKernel::configure(const ICLTensor *input_x, const ICLTensor *input_y, ICLTensor *output_x, ICLTensor *output_y, bool border_undefined) diff --git a/src/core/CL/kernels/CLSoftmaxLayerKernel.cpp b/src/core/CL/kernels/CLSoftmaxLayerKernel.cpp index 403256baae..e2d988103c 100644 --- a/src/core/CL/kernels/CLSoftmaxLayerKernel.cpp +++ b/src/core/CL/kernels/CLSoftmaxLayerKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -62,11 +62,12 @@ CLBuildOptions prepare_quantized_softmax_build_options(float input_scale, float const double beta_multiplier = std::min( 1.0 * beta * input_scale * (1 << (31 - scaled_diff_int_bits)), - (1ll << 31) - 1.0); - int input_beta_multiplier, input_beta_left_shift; + (1LL << 31) - 1.0); + int input_beta_multiplier; + int input_beta_left_shift; quantization::calculate_quantized_multiplier_greater_than_one(beta_multiplier, &input_beta_multiplier, &input_beta_left_shift); - const double max_input_rescaled = 1.0 * ((1 << scaled_diff_int_bits) - 1) * (1ll << (31 - scaled_diff_int_bits)) / (1ll << input_beta_left_shift); + const double max_input_rescaled = 1.0 * ((1 << scaled_diff_int_bits) - 1) * (1LL << (31 - scaled_diff_int_bits)) / (1LL << input_beta_left_shift); const int diff_min = -1.f * std::floor(max_input_rescaled); CLBuildOptions build_opts; @@ -337,7 +338,7 @@ void CLLogits1DNormKernel::configure(const ICLTensor *input, const ICLTensor *su ARM_COMPUTE_ERROR_ON_NULLPTR(input, sum, output); // Note: output should always have a scale of 1/256 and offset 0 - const QuantizationInfo allowed_quantization_info = QuantizationInfo(1.f / 256, 0); + const QuantizationInfo allowed_quantization_info = QuantizationInfo(1.F / 256, 0); const bool is_quantized_asymmetric = (input->info()->data_type() == DataType::S32); const DataType output_data_type = is_quantized_asymmetric ? DataType::QASYMM8 : input->info()->data_type(); diff --git a/src/core/CL/kernels/CLStridedSliceKernel.cpp b/src/core/CL/kernels/CLStridedSliceKernel.cpp index c40f3c9f0b..5a6b958719 100644 --- a/src/core/CL/kernels/CLStridedSliceKernel.cpp +++ b/src/core/CL/kernels/CLStridedSliceKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2018 ARM Limited. + * Copyright (c) 2018-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -115,7 +115,9 @@ void CLStridedSliceKernel::configure(const ICLTensor *input, ICLTensor *output, const TensorShape &input_shape = input->info()->tensor_shape(); - Coordinates starts_abs, ends_abs, final_strides; + Coordinates starts_abs; + Coordinates ends_abs; + Coordinates final_strides; std::tie(starts_abs, ends_abs, final_strides) = arm_compute::helpers::tensor_transform::calculate_strided_slice_coords( input_shape, starts, ends, strides, diff --git a/src/core/HOGInfo.cpp b/src/core/HOGInfo.cpp index 4f99455b56..bfee12c04f 100644 --- a/src/core/HOGInfo.cpp +++ b/src/core/HOGInfo.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -68,24 +68,24 @@ Size2D HOGInfo::num_cells_per_block() const { ARM_COMPUTE_ERROR_ON(_cell_size.width == 0 || _cell_size.height == 0); - return Size2D(_block_size.width / _cell_size.width, - _block_size.height / _cell_size.height); + return Size2D{ _block_size.width / _cell_size.width, + _block_size.height / _cell_size.height }; } Size2D HOGInfo::num_cells_per_block_stride() const { ARM_COMPUTE_ERROR_ON(_cell_size.width == 0 || _cell_size.height == 0); - return Size2D(_block_stride.width / _cell_size.width, - _block_stride.height / _cell_size.height); + return Size2D{ _block_stride.width / _cell_size.width, + _block_stride.height / _cell_size.height }; } Size2D HOGInfo::num_block_positions_per_image(const Size2D &image_size) const { ARM_COMPUTE_ERROR_ON(_block_stride.width == 0 || _block_stride.height == 0); - return Size2D(((image_size.width - _block_size.width) / _block_stride.width) + 1, - ((image_size.height - _block_size.height) / _block_stride.height) + 1); + return Size2D{ ((image_size.width - _block_size.width) / _block_stride.width) + 1, + ((image_size.height - _block_size.height) / _block_stride.height) + 1 }; } const Size2D &HOGInfo::cell_size() const diff --git a/src/core/ITensor.cpp b/src/core/ITensor.cpp index e6c80e8879..7cf04b51c4 100644 --- a/src/core/ITensor.cpp +++ b/src/core/ITensor.cpp @@ -64,7 +64,7 @@ void ITensor::copy_from(const ITensor &src) const size_t line_size = src_info->element_size() * src_info->dimension(0); - execute_window_loop(win_src, [&](const Coordinates & id) + execute_window_loop(win_src, [&](const Coordinates &) { memcpy(dst_it.ptr(), src_it.ptr(), line_size); }, diff --git a/src/core/NEON/kernels/NEAbsoluteDifferenceKernel.cpp b/src/core/NEON/kernels/NEAbsoluteDifferenceKernel.cpp index e0c2891592..62285e0578 100644 --- a/src/core/NEON/kernels/NEAbsoluteDifferenceKernel.cpp +++ b/src/core/NEON/kernels/NEAbsoluteDifferenceKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -49,7 +49,7 @@ void abs_diff_U8_U8_U8(const ITensor *in1, const ITensor *in2, ITensor *out, con Iterator input2(in2, window); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t input1_val = vld1q_u8(input1.ptr()); const uint8x16_t input2_val = vld1q_u8(input2.ptr()); @@ -78,7 +78,7 @@ void abs_diff_S16_S16_S16(const ITensor *in1, const ITensor *in2, ITensor *out, Iterator input2(in2, window); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { int16x8x2_t input1_val = vld2q_s16(reinterpret_cast<const int16_t *>(input1.ptr())); int16x8x2_t input2_val = vld2q_s16(reinterpret_cast<const int16_t *>(input2.ptr())); @@ -93,7 +93,7 @@ void abs_diff_U8_S16_S16(const ITensor *in1, const ITensor *in2, ITensor *out, c Iterator input2(in2, window); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t input1_val = vld1q_u8(input1.ptr()); const int16x8x2_t input2_val = diff --git a/src/core/NEON/kernels/NEAccumulateKernel.cpp b/src/core/NEON/kernels/NEAccumulateKernel.cpp index dae08008fd..d601adc1af 100644 --- a/src/core/NEON/kernels/NEAccumulateKernel.cpp +++ b/src/core/NEON/kernels/NEAccumulateKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -126,7 +126,7 @@ void NEAccumulateWeightedFP16Kernel::run(const Window &window, const ThreadInfo const float16x8_t scale_val = vdupq_n_f16(1.f - _alpha); const float16x8_t scale_val2 = vdupq_n_f16(_alpha); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { fp16::acc_we_v16_u8(input.ptr(), accum.ptr(), scale_val, scale_val2); }, @@ -271,7 +271,7 @@ void NEAccumulateKernel::run(const Window &window, const ThreadInfo &info) Iterator input(_input, window); Iterator accum(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { acc_v16_u8(input.ptr(), accum.ptr()); }, @@ -314,7 +314,7 @@ void NEAccumulateWeightedKernel::run(const Window &window, const ThreadInfo &inf const float32x4_t scale_val = vdupq_n_f32(1.f - _alpha); const float32x4_t scale_val2 = vdupq_n_f32(_alpha); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { acc_we_v16_u8(input.ptr(), accum.ptr(), scale_val, scale_val2); }, @@ -353,7 +353,7 @@ void NEAccumulateSquaredKernel::run(const Window &window, const ThreadInfo &info Iterator input(_input, window); Iterator accum(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { acc_sq_v16_u8(input.ptr(), _shift, accum.ptr()); }, diff --git a/src/core/NEON/kernels/NEActivationLayerKernel.cpp b/src/core/NEON/kernels/NEActivationLayerKernel.cpp index cf31cb841a..8de8db9ad9 100644 --- a/src/core/NEON/kernels/NEActivationLayerKernel.cpp +++ b/src/core/NEON/kernels/NEActivationLayerKernel.cpp @@ -195,7 +195,7 @@ NEActivationLayerKernel::activation(const Window &window) const auto a = static_cast<T>(_act_info.a()); const auto b = static_cast<T>(_act_info.b()); - execute_window_loop(win_collapsed, [&](const Coordinates & id) + execute_window_loop(win_collapsed, [&](const Coordinates &) { const auto input_ptr = reinterpret_cast<const T *>(input.ptr()); const auto output_ptr = reinterpret_cast<T *>(output.ptr()); @@ -327,7 +327,7 @@ typename std::enable_if<std::is_same<T, qasymm8_t>::value, void>::type NEActivat float32x4_t vs = vdupq_n_f32(s); float32x4_t vo = vdupq_n_f32(o); - execute_window_loop(win_collapsed, [&](const Coordinates & id) + execute_window_loop(win_collapsed, [&](const Coordinates &) { const auto input_ptr = reinterpret_cast<const T *>(input.ptr()); const auto output_ptr = reinterpret_cast<T *>(output.ptr()); diff --git a/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp b/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp index ffa578f40e..ca79a0a419 100644 --- a/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp +++ b/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp @@ -84,7 +84,7 @@ void add_same(const ITensor *in1, const ITensor *in2, ITensor *out, ConvertPolic Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto non_broadcast_input_ptr = reinterpret_cast<const T *>(non_broadcast_input.ptr()); const auto output_ptr = reinterpret_cast<T *>(output.ptr()); @@ -120,7 +120,7 @@ void add_same(const ITensor *in1, const ITensor *in2, ITensor *out, ConvertPolic Iterator input2(in2, input2_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto input1_ptr = reinterpret_cast<const T *>(input1.ptr()); const auto input2_ptr = reinterpret_cast<const T *>(input2.ptr()); @@ -165,8 +165,8 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor const auto window_end_x = static_cast<int>(window.x().end()); const bool is_broadcast_across_x = (input1_win.x().step() == 0) || (input2_win.x().step() == 0); - const float output_scale = out->info()->quantization_info().scale; - const int output_offset = out->info()->quantization_info().offset; + const float output_scale = out->info()->quantization_info().scale; + const int output_offset = out->info()->quantization_info().offset; const float32x4_t vscale1 = vdupq_n_f32(in1->info()->quantization_info().scale); const float32x4_t vscale2 = vdupq_n_f32(in2->info()->quantization_info().scale); @@ -192,7 +192,7 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto non_broadcast_input_ptr = reinterpret_cast<const uint8_t *>(non_broadcast_input.ptr()); const auto output_ptr = reinterpret_cast<uint8_t *>(output.ptr()); @@ -234,7 +234,7 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[3], bf.val[3]), invvscaleo)), -#else //__aarch64__ +#else //__aarch64__ vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[0], bf.val[0]), invvscaleo)), vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), @@ -252,7 +252,7 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor for(; x < window_end_x; ++x) { const float afs = static_cast<int32_t>(*(non_broadcast_input_ptr + x) - non_broadcast_qinfo.offset) * non_broadcast_qinfo.scale; - *(output_ptr + x) = out->info()->quantization_info().quantize((afs + bfs),RoundingPolicy::TO_NEAREST_UP); + *(output_ptr + x) = out->info()->quantization_info().quantize((afs + bfs), RoundingPolicy::TO_NEAREST_UP); } }, broadcast_input, non_broadcast_input, output); @@ -270,7 +270,7 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor Iterator input2(in2, input2_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto input1_ptr = reinterpret_cast<const uint8_t *>(input1.ptr()); const auto input2_ptr = reinterpret_cast<const uint8_t *>(input2.ptr()); @@ -311,7 +311,7 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[3], bf.val[3]), invvscaleo)), -#else //__aarch64__ +#else //__aarch64__ vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[0], bf.val[0]), invvscaleo)), vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), @@ -330,7 +330,7 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor { const float afs = static_cast<int32_t>((*(input1_ptr + x)) - input1_qinfo.offset) * input1_qinfo.scale; const float bfs = static_cast<int32_t>((*(input2_ptr + x)) - input2_qinfo.offset) * input2_qinfo.scale; - *(output_ptr + x) = out->info()->quantization_info().quantize((afs + bfs),RoundingPolicy::TO_NEAREST_UP); + *(output_ptr + x) = out->info()->quantization_info().quantize((afs + bfs), RoundingPolicy::TO_NEAREST_UP); } }, input1, input2, output); @@ -357,7 +357,7 @@ void add_S16_U8_S16(const ITensor *in1, const ITensor *in2, ITensor *out, Conver const auto window_start_x = static_cast<int>(window.x().start()); const auto window_end_x = static_cast<int>(window.x().end()); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto input1_ptr = reinterpret_cast<const int16_t *>(input1.ptr()); const auto input2_ptr = reinterpret_cast<const uint8_t *>(input2.ptr()); @@ -427,7 +427,7 @@ void add_U8_U8_S16(const ITensor *in1, const ITensor *in2, ITensor *out, Convert const auto window_start_x = static_cast<int>(window.x().start()); const auto window_end_x = static_cast<int>(window.x().end()); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto input1_ptr = reinterpret_cast<const uint8_t *>(input1.ptr()); const auto input2_ptr = reinterpret_cast<const uint8_t *>(input2.ptr()); diff --git a/src/core/NEON/kernels/NEArithmeticSubtractionKernel.cpp b/src/core/NEON/kernels/NEArithmeticSubtractionKernel.cpp index ff5893de96..45e1562d8d 100644 --- a/src/core/NEON/kernels/NEArithmeticSubtractionKernel.cpp +++ b/src/core/NEON/kernels/NEArithmeticSubtractionKernel.cpp @@ -55,7 +55,7 @@ void sub_wrap_U8_U8_U8(const ITensor *in1, const ITensor *in2, ITensor *out, con Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t ta1 = vld1q_u8(input1.ptr()); const uint8x16_t ta2 = vld1q_u8(input2.ptr()); @@ -71,7 +71,7 @@ void sub_saturate_U8_U8_U8(const ITensor *in1, const ITensor *in2, ITensor *out, Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t ta1 = vld1q_u8(input1.ptr()); const uint8x16_t ta2 = vld1q_u8(input2.ptr()); @@ -87,7 +87,7 @@ void sub_saturate_QAYSMM8_QAYSMM8_QAYSMM8(const ITensor *in1, const ITensor *in2 Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float32x4x4_t ta1 = vdequantize(vld1q_u8(reinterpret_cast<const qasymm8_t *>(input1.ptr())), in1->info()->quantization_info()); const float32x4x4_t ta2 = vdequantize(vld1q_u8(reinterpret_cast<const qasymm8_t *>(input2.ptr())), in2->info()->quantization_info()); @@ -115,7 +115,7 @@ void sub_wrap_S16_S16_S16(const ITensor *in1, const ITensor *in2, ITensor *out, Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t ta1 = vld2q_s16(reinterpret_cast<const int16_t *>(input1.ptr())); const int16x8x2_t ta2 = vld2q_s16(reinterpret_cast<const int16_t *>(input2.ptr())); @@ -139,7 +139,7 @@ void sub_saturate_S16_S16_S16(const ITensor *in1, const ITensor *in2, ITensor *o Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t ta1 = vld2q_s16(reinterpret_cast<const int16_t *>(input1.ptr())); const int16x8x2_t ta2 = vld2q_s16(reinterpret_cast<const int16_t *>(input2.ptr())); @@ -179,7 +179,7 @@ void sub_F16_F16_F16(const ITensor *in1, const ITensor *in2, ITensor *out, const Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float16x8x2_t a = vld2q_f16(reinterpret_cast<const float16_t *>(input1.ptr())); const float16x8x2_t b = vld2q_f16(reinterpret_cast<const float16_t *>(input2.ptr())); @@ -202,7 +202,7 @@ void sub_F32_F32_F32(const ITensor *in1, const ITensor *in2, ITensor *out, const Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float32x4x4_t ta1 = vld4q_f32(reinterpret_cast<const float *>(input1.ptr())); const float32x4x4_t ta2 = vld4q_f32(reinterpret_cast<const float *>(input2.ptr())); @@ -227,7 +227,7 @@ void sub_wrap_S16_U8_S16(const ITensor *in1, const ITensor *in2, ITensor *out, c Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t bv_0 = vld1q_u8(input2.ptr()); int16x8_t a1_0 = vld1q_s16(reinterpret_cast<const int16_t *>(input1.ptr())); @@ -248,7 +248,7 @@ void sub_saturate_S16_U8_S16(const ITensor *in1, const ITensor *in2, ITensor *ou Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t bv_0 = vld1q_u8(input2.ptr()); int16x8_t a1_0 = vld1q_s16(reinterpret_cast<const int16_t *>(input1.ptr())); @@ -269,7 +269,7 @@ void sub_wrap_U8_S16_S16(const ITensor *in1, const ITensor *in2, ITensor *out, c Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t bv_0 = vld1q_u8(input1.ptr()); int16x8_t a1_0 = vld1q_s16(reinterpret_cast<const int16_t *>(input2.ptr())); @@ -290,7 +290,7 @@ void sub_saturate_U8_S16_S16(const ITensor *in1, const ITensor *in2, ITensor *ou Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t bv_0 = vld1q_u8(input1.ptr()); int16x8_t a1_0 = vld1q_s16(reinterpret_cast<const int16_t *>(input2.ptr())); @@ -311,7 +311,7 @@ void sub_wrap_U8_U8_S16(const ITensor *in1, const ITensor *in2, ITensor *out, co Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t av_0 = vld1q_u8(input1.ptr()); const uint8x16_t bv_0 = vld1q_u8(input2.ptr()); @@ -333,7 +333,7 @@ void sub_saturate_U8_U8_S16(const ITensor *in1, const ITensor *in2, ITensor *out Iterator input2(in2, window.broadcast_if_dimension_le_one(in2->info()->tensor_shape())); Iterator output(out, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t av_0 = vld1q_u8(input1.ptr()); const uint8x16_t bv_0 = vld1q_u8(input2.ptr()); @@ -515,5 +515,5 @@ BorderSize NEArithmeticSubtractionKernel::border_size() const { const unsigned int replicateSize = _output->info()->dimension(0) - std::min(_input1->info()->dimension(0), _input2->info()->dimension(0)); const unsigned int border = std::min<unsigned int>(num_elems_processed_per_iteration - 1U, replicateSize); - return BorderSize(0, border, 0, 0); + return BorderSize{ 0, border, 0, 0 }; }
\ No newline at end of file diff --git a/src/core/NEON/kernels/NEBitwiseAndKernel.cpp b/src/core/NEON/kernels/NEBitwiseAndKernel.cpp index ed83286acf..71312a923d 100644 --- a/src/core/NEON/kernels/NEBitwiseAndKernel.cpp +++ b/src/core/NEON/kernels/NEBitwiseAndKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -106,7 +106,7 @@ void NEBitwiseAndKernel::run(const Window &window, const ThreadInfo &info) Iterator input2(_input2, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { bitwise_and<uint8_t>(input1.ptr(), input2.ptr(), output.ptr()); }, diff --git a/src/core/NEON/kernels/NEBitwiseNotKernel.cpp b/src/core/NEON/kernels/NEBitwiseNotKernel.cpp index 08d7fe2610..5791dcc704 100644 --- a/src/core/NEON/kernels/NEBitwiseNotKernel.cpp +++ b/src/core/NEON/kernels/NEBitwiseNotKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -89,7 +89,7 @@ void NEBitwiseNotKernel::run(const Window &window, const ThreadInfo &info) Iterator input(_input, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { bitwise_not_U8_U8(input.ptr(), output.ptr()); }, diff --git a/src/core/NEON/kernels/NEBitwiseOrKernel.cpp b/src/core/NEON/kernels/NEBitwiseOrKernel.cpp index 1b17cc283c..8aed9bb6da 100644 --- a/src/core/NEON/kernels/NEBitwiseOrKernel.cpp +++ b/src/core/NEON/kernels/NEBitwiseOrKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -102,7 +102,7 @@ void NEBitwiseOrKernel::run(const Window &window, const ThreadInfo &info) Iterator input2(_input2, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { bitwise_or_U8_U8_U8(input1.ptr(), input2.ptr(), output.ptr()); }, diff --git a/src/core/NEON/kernels/NEBitwiseXorKernel.cpp b/src/core/NEON/kernels/NEBitwiseXorKernel.cpp index 9451e8a08d..e2dcb95f64 100644 --- a/src/core/NEON/kernels/NEBitwiseXorKernel.cpp +++ b/src/core/NEON/kernels/NEBitwiseXorKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -98,7 +98,7 @@ void NEBitwiseXorKernel::run(const Window &window, const ThreadInfo &info) Iterator input2(_input2, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { bitwise_xor_U8_U8_U8(input1.ptr(), input2.ptr(), output.ptr()); }, diff --git a/src/core/NEON/kernels/NEBox3x3Kernel.cpp b/src/core/NEON/kernels/NEBox3x3Kernel.cpp index 0c9700526b..7a53f93b86 100644 --- a/src/core/NEON/kernels/NEBox3x3Kernel.cpp +++ b/src/core/NEON/kernels/NEBox3x3Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -49,7 +49,7 @@ void NEBox3x3FP16Kernel::run(const Window &window, const ThreadInfo &info) const float16x8_t oneovernine = vdupq_n_f16(1.0f / 9.0f); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); @@ -160,7 +160,7 @@ void NEBox3x3Kernel::run(const Window &window, const ThreadInfo &info) const float32x4_t oneovernine = vdupq_n_f32(1.0f / 9.0f); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); diff --git a/src/core/NEON/kernels/NECannyEdgeKernel.cpp b/src/core/NEON/kernels/NECannyEdgeKernel.cpp index fa51a7bb0b..8d822bd701 100644 --- a/src/core/NEON/kernels/NECannyEdgeKernel.cpp +++ b/src/core/NEON/kernels/NECannyEdgeKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -950,7 +950,7 @@ void NEGradientKernel::run(const Window &window, const ThreadInfo &info) Iterator magnitude(_magnitude, window); Iterator phase(_phase, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { (*_func)(gx.ptr(), gy.ptr(), magnitude.ptr(), phase.ptr()); }, @@ -1034,7 +1034,7 @@ void NEEdgeNonMaxSuppressionKernel::run(const Window &window, const ThreadInfo & const size_t input1_stride = _magnitude->info()->strides_in_bytes()[1]; const size_t input1_stride_ushort = input1_stride / data_size_from_type(_magnitude->info()->data_type()); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { (*_func)(magnitude.ptr(), phase.ptr(), output.ptr(), input1_stride_ushort, _lower_thr, _upper_thr); }, @@ -1113,7 +1113,7 @@ void NEEdgeTraceKernel::run(const Window &window, const ThreadInfo &info) const size_t input_stride = _input->info()->strides_in_bytes()[1]; const size_t output_stride = _output->info()->strides_in_bytes()[1]; - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { edge_trace_U8_U8(input.ptr(), output.ptr(), input_stride, output_stride); }, diff --git a/src/core/NEON/kernels/NEChannelCombineKernel.cpp b/src/core/NEON/kernels/NEChannelCombineKernel.cpp index 28fb4bdb10..539154d671 100644 --- a/src/core/NEON/kernels/NEChannelCombineKernel.cpp +++ b/src/core/NEON/kernels/NEChannelCombineKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -286,7 +286,7 @@ void NEChannelCombineKernel::combine_3C(const Window &win) Iterator p2(_planes[2], win); Iterator out(_output, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto p0_ptr = static_cast<uint8_t *>(p0.ptr()); const auto p1_ptr = static_cast<uint8_t *>(p1.ptr()); @@ -315,7 +315,7 @@ void NEChannelCombineKernel::combine_4C(const Window &win) Iterator p3(_planes[3], win); Iterator out(_output, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto p0_ptr = static_cast<uint8_t *>(p0.ptr()); const auto p1_ptr = static_cast<uint8_t *>(p1.ptr()); @@ -353,7 +353,7 @@ void NEChannelCombineKernel::combine_YUV_1p(const Window &win) constexpr auto shift = is_uyvy ? 1 : 0; - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto p0_ptr = static_cast<uint8_t *>(p0.ptr()); const auto p1_ptr = static_cast<uint8_t *>(p1.ptr()); @@ -409,7 +409,7 @@ void NEChannelCombineKernel::combine_YUV_2p(const Window &win) // Increase step size after iterator is created to calculate stride correctly for multi channel format out_win.set_dimension_step(Window::DimX, out_win.x().step() * _x_subsampling[1]); - execute_window_loop(out_win, [&](const Coordinates & id) + execute_window_loop(out_win, [&](const Coordinates &) { const uint8x8x2_t pixels = { @@ -444,7 +444,7 @@ void NEChannelCombineKernel::copy_plane(const Window &win, uint32_t plane_id) Iterator in(_planes[plane_id], tmp_win); Iterator out(_output_multi->plane(plane_id), tmp_win); - execute_window_loop(tmp_win, [&](const Coordinates & id) + execute_window_loop(tmp_win, [&](const Coordinates &) { const uint8x8_t pixels = vld1_u8(in.ptr()); diff --git a/src/core/NEON/kernels/NEChannelExtractKernel.cpp b/src/core/NEON/kernels/NEChannelExtractKernel.cpp index 98b2f280d4..61e1304963 100644 --- a/src/core/NEON/kernels/NEChannelExtractKernel.cpp +++ b/src/core/NEON/kernels/NEChannelExtractKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -192,7 +192,7 @@ void NEChannelExtractKernel::extract_1C_from_2C_img(const Window &win) Iterator in(_input, win); Iterator out(_output, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto in_ptr = static_cast<uint8_t *>(in.ptr()); const auto out_ptr = static_cast<uint8_t *>(out.ptr()); @@ -207,7 +207,7 @@ void NEChannelExtractKernel::extract_1C_from_3C_img(const Window &win) Iterator in(_input, win); Iterator out(_output, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto in_ptr = static_cast<uint8_t *>(in.ptr()); const auto out_ptr = static_cast<uint8_t *>(out.ptr()); @@ -222,7 +222,7 @@ void NEChannelExtractKernel::extract_1C_from_4C_img(const Window &win) Iterator in(_input, win); Iterator out(_output, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto in_ptr = static_cast<uint8_t *>(in.ptr()); const auto out_ptr = static_cast<uint8_t *>(out.ptr()); @@ -242,7 +242,7 @@ void NEChannelExtractKernel::extract_YUYV_uv(const Window &win) Iterator in(_input, win); Iterator out(_output, win_out); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto in_ptr = static_cast<uint8_t *>(in.ptr()); const auto out_ptr = static_cast<uint8_t *>(out.ptr()); diff --git a/src/core/NEON/kernels/NEConvolutionKernel.cpp b/src/core/NEON/kernels/NEConvolutionKernel.cpp index 0a10546b7b..b154340bee 100644 --- a/src/core/NEON/kernels/NEConvolutionKernel.cpp +++ b/src/core/NEON/kernels/NEConvolutionKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -317,7 +317,7 @@ NEConvolutionKernel<matrix_size>::NEConvolutionKernel() template <unsigned int matrix_size> BorderSize NEConvolutionKernel<matrix_size>::border_size() const { - return BorderSize(matrix_size / 2); + return BorderSize{ matrix_size / 2 }; } template <unsigned int matrix_size> @@ -388,7 +388,7 @@ void NEConvolutionKernel<3>::convolution(const Window &win) const unsigned char *input_mid_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-1, 0)); const unsigned char *input_low_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-1, 1)); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int32x4_t out = vdupq_n_s32(0); int32x4_t out2 = vdupq_n_s32(0); @@ -437,7 +437,7 @@ void NEConvolutionKernel<5>::convolution(const Window &win) const unsigned char *input_low1_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-2, 1)); const unsigned char *input_low2_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-2, 2)); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int32x4_t out = vdupq_n_s32(0); int32x4_t out2 = vdupq_n_s32(0); @@ -496,7 +496,7 @@ void NEConvolutionKernel<7>::convolution(const Window &win) const unsigned char *input_low2_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-3, 2)); const unsigned char *input_low3_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-3, 3)); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int32x4_t out = vdupq_n_s32(0); int32x4_t out2 = vdupq_n_s32(0); @@ -565,7 +565,7 @@ void NEConvolutionKernel<9>::convolution(const Window &win) const unsigned char *input_low3_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-4, 3)); const unsigned char *input_low4_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-4, 4)); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int32x4_t out = vdupq_n_s32(0); int32x4_t out2 = vdupq_n_s32(0); @@ -728,7 +728,7 @@ inline void NESeparableConvolutionHorKernel<5>::convolve<uint16_t>(const Window Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -761,7 +761,7 @@ inline void NESeparableConvolutionHorKernel<5>::convolve<int16_t>(const Window & Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -794,7 +794,7 @@ void NESeparableConvolutionHorKernel<5>::convolve<int32_t>(const Window &window) Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -840,7 +840,7 @@ inline void NESeparableConvolutionHorKernel<7>::convolve<uint16_t>(const Window Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -875,7 +875,7 @@ inline void NESeparableConvolutionHorKernel<7>::convolve<int16_t>(const Window & Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -910,7 +910,7 @@ void NESeparableConvolutionHorKernel<7>::convolve<int32_t>(const Window &window) Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -962,7 +962,7 @@ inline void NESeparableConvolutionHorKernel<9>::convolve<uint16_t>(const Window Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -999,7 +999,7 @@ inline void NESeparableConvolutionHorKernel<9>::convolve<int16_t>(const Window & Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -1036,7 +1036,7 @@ void NESeparableConvolutionHorKernel<9>::convolve<int32_t>(const Window &window) Iterator input(_input, win_in); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -1096,7 +1096,7 @@ NESeparableConvolutionVertKernel<matrix_size>::NESeparableConvolutionVertKernel( template <unsigned int matrix_size> BorderSize NESeparableConvolutionVertKernel<matrix_size>::border_size() const { - return BorderSize(matrix_size / 2, 0); + return BorderSize{ matrix_size / 2, 0 }; } template <unsigned int matrix_size> @@ -1209,7 +1209,7 @@ void NESeparableConvolutionVertKernel<matrix_size>::convolution_u16(const Window input_ptrs[k_half + i] = _input->ptr_to_element(Coordinates(0, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { uint16x8_t out0 = vdupq_n_u16(0); uint16x8_t out1 = vdupq_n_u16(0); @@ -1275,7 +1275,7 @@ void NESeparableConvolutionVertKernel<matrix_size>::convolution_s16(const Window input_ptrs[k_half + i] = _input->ptr_to_element(Coordinates(0, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int16x8_t out0 = vdupq_n_s16(0); int16x8_t out1 = vdupq_n_s16(0); @@ -1343,7 +1343,7 @@ void NESeparableConvolutionVertKernel<matrix_size>::convolution_s32(const Window const int32x4_t zero = vdupq_n_s32(0); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int32x4x2_t out0 = { @@ -1576,7 +1576,7 @@ void NEConvolutionRectangleKernel::convolution(const Window &win) input_ptrs[k_row_half + i] = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-k_col_half, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int32x4_t out = vdupq_n_s32(0); int32x4_t out2 = vdupq_n_s32(0); diff --git a/src/core/NEON/kernels/NECopyKernel.cpp b/src/core/NEON/kernels/NECopyKernel.cpp index c5103f07c8..4722c05507 100644 --- a/src/core/NEON/kernels/NECopyKernel.cpp +++ b/src/core/NEON/kernels/NECopyKernel.cpp @@ -70,7 +70,7 @@ void NECopyKernel::run(const Window &window, const ThreadInfo &info) Iterator input_it(_input, out_slice); Iterator output_it(_output, out_slice); - execute_window_loop(out_slice, [&](const Coordinates & id) + execute_window_loop(out_slice, [&](const Coordinates &) { memcpy(output_it.ptr(), input_it.ptr(), _output->info()->dimension(0) * _output->info()->element_size()); }, diff --git a/src/core/NEON/kernels/NEDepthConvertLayerKernel.cpp b/src/core/NEON/kernels/NEDepthConvertLayerKernel.cpp index 54337551a7..cbc90a058f 100644 --- a/src/core/NEON/kernels/NEDepthConvertLayerKernel.cpp +++ b/src/core/NEON/kernels/NEDepthConvertLayerKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -148,7 +148,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info const float32x4_t scale = vdupq_n_f32(_input->info()->quantization_info().scale); const int32x4_t offset = vdupq_n_s32(_input->info()->quantization_info().offset); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t texels_u8 = vld1q_u8(input.ptr()); const uint16x8x2_t texels_u16 = @@ -184,7 +184,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info const float16x8_t scale = vdupq_n_f16(static_cast<float16_t>(_input->info()->quantization_info().scale)); const int16x8_t offset = vdupq_n_s16(static_cast<int16_t>(_input->info()->quantization_info().offset)); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t texels_u8 = vld1q_u8(input.ptr()); const int16x8x2_t texels_s16 = @@ -216,7 +216,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info case DataType::S16: { /* Up-conversion U8 -> S16 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t texels_u8 = vld1q_u8(input.ptr()); @@ -237,7 +237,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info case DataType::S32: { /* Up-conversion U8 -> S32 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t texels_u8 = vld1q_u8(input.ptr()); @@ -260,7 +260,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info case DataType::U16: { /* Up-conversion U8 -> U16 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t texels_u8 = vld1q_u8(input.ptr()); @@ -294,7 +294,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info /* Down-conversion S16 -> U8 */ if(ConvertPolicy::SATURATE == _policy) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t texels = { @@ -310,7 +310,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info } else { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t texels = { @@ -332,7 +332,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info const int32x4_t b = vdupq_n_s32(_shift); /* Up-conversion S16 -> S32 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t texels = { @@ -376,7 +376,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info /* Down-conversion U16 -> U8 */ if(ConvertPolicy::SATURATE == _policy) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint16x8x2_t texels = { @@ -392,7 +392,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info } else { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint16x8x2_t texels = { @@ -413,7 +413,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info const int32x4_t b = vdupq_n_s32(_shift); /* Up-conversion U16 -> U32 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint16x8x2_t texels = { @@ -504,7 +504,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info const int32x4_t zero_val_vec = vdupq_n_s32(0); /* Down-conversion F32 -> QASYMM8 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float32x4x4_t texels = { @@ -535,7 +535,7 @@ void NEDepthConvertLayerKernel::run(const Window &window, const ThreadInfo &info const float32x4_t scale = vdupq_n_f32(1.f / (1 << _shift)); /* Down-conversion F32 -> F16 */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float32x4x4_t texels = { diff --git a/src/core/NEON/kernels/NEDequantizationLayerKernel.cpp b/src/core/NEON/kernels/NEDequantizationLayerKernel.cpp index 119aa4ad9a..1520225249 100644 --- a/src/core/NEON/kernels/NEDequantizationLayerKernel.cpp +++ b/src/core/NEON/kernels/NEDequantizationLayerKernel.cpp @@ -111,7 +111,7 @@ void run_dequantization(const ITensor *input, ITensor *output, const Window &win Iterator in(input, win_collapsed); Iterator out(output, win_collapsed); - execute_window_loop(win_collapsed, [&](const Coordinates & id) + execute_window_loop(win_collapsed, [&](const Coordinates &) { const auto in_ptr = reinterpret_cast<const uint8_t *>(in.ptr()); const auto out_ptr = reinterpret_cast<T *>(out.ptr()); diff --git a/src/core/NEON/kernels/NEDerivativeKernel.cpp b/src/core/NEON/kernels/NEDerivativeKernel.cpp index cfed324773..1d7237a5b4 100644 --- a/src/core/NEON/kernels/NEDerivativeKernel.cpp +++ b/src/core/NEON/kernels/NEDerivativeKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -124,7 +124,7 @@ void NEDerivativeKernel::derivative_x(const Window &window) Iterator out_x(_output_x, window); /* Apply 1-D centered point discrete derivative mask ([-1 0 1]) along the X direction */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { /* Load left and right data */ const uint8x16_t l_data = vld1q_u8(in.ptr() - 1); @@ -153,7 +153,7 @@ void NEDerivativeKernel::derivative_y(const Window &window) const size_t stride = _input->info()->strides_in_bytes()[1]; /* Apply 1-D centered point discrete derivative mask ([-1 0 1]^T) along the Y direction */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { /* Load top and bottom data */ const uint8x16_t t_data = vld1q_u8(in.ptr() - stride); @@ -183,7 +183,7 @@ void NEDerivativeKernel::derivative_xy(const Window &window) const size_t stride = _input->info()->strides_in_bytes()[1]; /* Apply 1-D centered point discrete derivative masks ([-1 0 1] and [-1 0 1]^T) along the X and Y directions */ - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { /* Load top, bottom, left and right data */ const uint8x16_t t_data = vld1q_u8(in.ptr() - stride); diff --git a/src/core/NEON/kernels/NEDirectConvolutionLayerOutputStageKernel.cpp b/src/core/NEON/kernels/NEDirectConvolutionLayerOutputStageKernel.cpp index 09836f1d6b..7e113935c5 100644 --- a/src/core/NEON/kernels/NEDirectConvolutionLayerOutputStageKernel.cpp +++ b/src/core/NEON/kernels/NEDirectConvolutionLayerOutputStageKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -266,7 +266,7 @@ void output_stage_nhwc(ITensor *input, const ITensor *bias, const Window &window if(in_place) // In place accumulate { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Get bias and pointer to input const auto in_ptr = reinterpret_cast<T1 *>(in.ptr()); @@ -287,7 +287,7 @@ void output_stage_nhwc(ITensor *input, const ITensor *bias, const Window &window else // Out of place accumulate { Iterator out(output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Get bias and pointer to input const auto in_ptr = reinterpret_cast<T1 *>(in.ptr()); @@ -363,7 +363,7 @@ void output_stage_nchw<int32_t, uint8_t, false, false>(ITensor *input, const ITe Iterator in(input, window); Iterator out(output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Get bias and pointer to input const auto in_ptr = reinterpret_cast<int32_t *>(in.ptr()); @@ -399,7 +399,7 @@ void output_stage_nhwc<int32_t, uint8_t, false, true>(ITensor *input, const ITen Iterator bi(bias, window_bias); Iterator out(output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Get bias and pointer to input const auto in_ptr = reinterpret_cast<int32_t *>(in.ptr()); @@ -433,7 +433,7 @@ void output_stage_nhwc<int32_t, uint8_t, false, false>(ITensor *input, const ITe Iterator in(input, window); Iterator out(output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Get pointer to input const auto in_ptr = reinterpret_cast<int32_t *>(in.ptr()); diff --git a/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp b/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp index aa458c2119..6b87ea017b 100644 --- a/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp +++ b/src/core/NEON/kernels/NEElementwiseOperationKernel.cpp @@ -506,7 +506,7 @@ void elementwise_op(const ITensor *in1, const ITensor *in2, ITensor *out, const Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { auto output_ptr = reinterpret_cast<OutputScalarType *>(output.ptr()); const auto non_broadcast_input_ptr = reinterpret_cast<const InputScalarType *>(non_broadcast_input.ptr()); @@ -531,7 +531,7 @@ void elementwise_op(const ITensor *in1, const ITensor *in2, ITensor *out, const Iterator input2(in2, input2_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { auto output_ptr = reinterpret_cast<OutputScalarType *>(output.ptr()); const auto input1_ptr = reinterpret_cast<const InputScalarType *>(input1.ptr()); @@ -599,7 +599,7 @@ void elementwise_op_quantized(const ITensor *in1, const ITensor *in2, ITensor *o Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto non_broadcast_input_ptr = reinterpret_cast<const uint8_t *>(non_broadcast_input.ptr()); const auto output_ptr = reinterpret_cast<uint8_t *>(output.ptr()); @@ -640,7 +640,7 @@ void elementwise_op_quantized(const ITensor *in1, const ITensor *in2, ITensor *o Iterator input2(in2, input2_win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const auto input1_ptr = reinterpret_cast<const uint8_t *>(input1.ptr()); const auto input2_ptr = reinterpret_cast<const uint8_t *>(input2.ptr()); diff --git a/src/core/NEON/kernels/NEElementwiseUnaryKernel.cpp b/src/core/NEON/kernels/NEElementwiseUnaryKernel.cpp index 7ecc4d1c44..34696d872a 100644 --- a/src/core/NEON/kernels/NEElementwiseUnaryKernel.cpp +++ b/src/core/NEON/kernels/NEElementwiseUnaryKernel.cpp @@ -87,7 +87,7 @@ void elementwise_op(const ITensor *in, ITensor *out, const Window &window) Iterator input(in, win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { auto output_ptr = reinterpret_cast<ScalarType *>(output.ptr()); const auto input_ptr = reinterpret_cast<const ScalarType *>(input.ptr()); diff --git a/src/core/NEON/kernels/NEFastCornersKernel.cpp b/src/core/NEON/kernels/NEFastCornersKernel.cpp index 919efd2ae2..81bcc8bc3c 100644 --- a/src/core/NEON/kernels/NEFastCornersKernel.cpp +++ b/src/core/NEON/kernels/NEFastCornersKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -49,32 +49,30 @@ inline uint8x8x2_t create_permutation_index(size_t k) { ARM_COMPUTE_ERROR_ON(k >= PERMUTATIONS); - static const uint8_t permutations_table[PERMUTATIONS][PERM_SIZE] - { - { 0, 1, 2, 3, 4, 5, 6, 7, 8, 255, 255, 255, 255, 255, 255, 255 }, - { 15, 0, 1, 2, 3, 4, 5, 6, 7, 255, 255, 255, 255, 255, 255, 255 }, - { 14, 15, 0, 1, 2, 3, 4, 5, 6, 255, 255, 255, 255, 255, 255, 255 }, - { 13, 14, 15, 0, 1, 2, 3, 4, 5, 255, 255, 255, 255, 255, 255, 255 }, - { 12, 13, 14, 15, 0, 1, 2, 3, 4, 255, 255, 255, 255, 255, 255, 255 }, - { 11, 12, 13, 14, 15, 0, 1, 2, 3, 255, 255, 255, 255, 255, 255, 255 }, - { 10, 11, 12, 13, 14, 15, 0, 1, 2, 255, 255, 255, 255, 255, 255, 255 }, - { 9, 10, 11, 12, 13, 14, 15, 0, 1, 255, 255, 255, 255, 255, 255, 255 }, - { 8, 9, 10, 11, 12, 13, 14, 15, 0, 255, 255, 255, 255, 255, 255, 255 }, - { 7, 8, 9, 10, 11, 12, 13, 14, 15, 255, 255, 255, 255, 255, 255, 255 }, - { 6, 7, 8, 9, 10, 11, 12, 13, 14, 255, 255, 255, 255, 255, 255, 255 }, - { 5, 6, 7, 8, 9, 10, 11, 12, 13, 255, 255, 255, 255, 255, 255, 255 }, - { 4, 5, 6, 7, 8, 9, 10, 11, 12, 255, 255, 255, 255, 255, 255, 255 }, - { 3, 4, 5, 6, 7, 8, 9, 10, 11, 255, 255, 255, 255, 255, 255, 255 }, - { 2, 3, 4, 5, 6, 7, 8, 9, 10, 255, 255, 255, 255, 255, 255, 255 }, - { 1, 2, 3, 4, 5, 6, 7, 8, 9, 255, 255, 255, 255, 255, 255, 255 } - - }; + static const std::array<std::array<uint8_t, PERMUTATIONS>, PERM_SIZE> permutations_table{ { { 0, 1, 2, 3, 4, 5, 6, 7, 8, 255, 255, 255, 255, 255, 255, 255 }, + { 15, 0, 1, 2, 3, 4, 5, 6, 7, 255, 255, 255, 255, 255, 255, 255 }, + { 14, 15, 0, 1, 2, 3, 4, 5, 6, 255, 255, 255, 255, 255, 255, 255 }, + { 13, 14, 15, 0, 1, 2, 3, 4, 5, 255, 255, 255, 255, 255, 255, 255 }, + { 12, 13, 14, 15, 0, 1, 2, 3, 4, 255, 255, 255, 255, 255, 255, 255 }, + { 11, 12, 13, 14, 15, 0, 1, 2, 3, 255, 255, 255, 255, 255, 255, 255 }, + { 10, 11, 12, 13, 14, 15, 0, 1, 2, 255, 255, 255, 255, 255, 255, 255 }, + { 9, 10, 11, 12, 13, 14, 15, 0, 1, 255, 255, 255, 255, 255, 255, 255 }, + { 8, 9, 10, 11, 12, 13, 14, 15, 0, 255, 255, 255, 255, 255, 255, 255 }, + { 7, 8, 9, 10, 11, 12, 13, 14, 15, 255, 255, 255, 255, 255, 255, 255 }, + { 6, 7, 8, 9, 10, 11, 12, 13, 14, 255, 255, 255, 255, 255, 255, 255 }, + { 5, 6, 7, 8, 9, 10, 11, 12, 13, 255, 255, 255, 255, 255, 255, 255 }, + { 4, 5, 6, 7, 8, 9, 10, 11, 12, 255, 255, 255, 255, 255, 255, 255 }, + { 3, 4, 5, 6, 7, 8, 9, 10, 11, 255, 255, 255, 255, 255, 255, 255 }, + { 2, 3, 4, 5, 6, 7, 8, 9, 10, 255, 255, 255, 255, 255, 255, 255 }, + { 1, 2, 3, 4, 5, 6, 7, 8, 9, 255, 255, 255, 255, 255, 255, 255 } + + } }; const uint8x8x2_t index = { { - vld1_u8(permutations_table[k]), - vld1_u8(permutations_table[k] + 8) + vld1_u8(permutations_table[k].data()), + vld1_u8(permutations_table[k].data() + 8) } }; @@ -112,7 +110,7 @@ inline uint8x8x4_t create_circle_index_register() . . 9 8 7 . . . */ - static const uint8_t top_right[8] = + static const std::array<uint8_t, 8> top_right = { /* The register r.val[0] will be used to retrieve these texels: . . . 0 1 . . . @@ -130,7 +128,7 @@ inline uint8x8x4_t create_circle_index_register() 255 }; - static const uint8_t bottom_right[8] = + static const std::array<uint8_t, 8> bottom_right = { /* The register r.val[1] will be used to retrieve these texels: . . . . . . 5 . @@ -147,7 +145,7 @@ inline uint8x8x4_t create_circle_index_register() 20 /* low table, third row, elem 5, value 7 in the diagram above*/ }; - static const uint8_t top_left[8] = + static const std::array<uint8_t, 8> top_left = { /* The register r.val[2] will be used to retrieve these texels: . . F . . . . . @@ -165,7 +163,7 @@ inline uint8x8x4_t create_circle_index_register() 2 /* top table, first row, elem 3, value F in the diagram above*/ }; - static const uint8_t bottom_left[8] = + static const std::array<uint8_t, 8> bottom_left = { /* The register r.val[3] will be used to retrieve these texels: B . . . . . . . @@ -185,10 +183,10 @@ inline uint8x8x4_t create_circle_index_register() const uint8x8x4_t reg = { { - vld1_u8(top_right), - vld1_u8(bottom_right), - vld1_u8(top_left), - vld1_u8(bottom_left) + vld1_u8(top_right.data()), + vld1_u8(bottom_right.data()), + vld1_u8(top_left.data()), + vld1_u8(bottom_left.data()) } }; @@ -268,7 +266,7 @@ inline bool is_permutation_corner(const uint8x16_t &permutation, const uint8x16_ return is_permutation_brighter(permutation, pg) || is_permutation_darker(permutation, pl); } -inline bool point_is_fast_corner(uint8_t p, uint8_t threshold, const uint8x8x2_t &tbl_circle_texels, uint8x8x2_t perm_indices[PERMUTATIONS]) +inline bool point_is_fast_corner(uint8_t p, uint8_t threshold, const uint8x8x2_t &tbl_circle_texels, std::array<uint8x8x2_t, PERMUTATIONS> &perm_indices) { /* This function determines whether the point 'p' is a corner. @@ -287,7 +285,7 @@ inline bool point_is_fast_corner(uint8_t p, uint8_t threshold, const uint8x8x2_t return corner_detected; } -inline uint8x8x2_t create_circle_tbl(const uint8_t *const __restrict buffer[7], size_t in_offset, const uint8x8x4_t &circle_index_r) +inline uint8x8x2_t create_circle_tbl(const std::array<uint8_t *const __restrict, 7> &buffer, size_t in_offset, const uint8x8x4_t &circle_index_r) { /* This function builds a LUT holding the 16 texels in the Brensenham circle radius 3. @@ -329,7 +327,7 @@ inline uint8x8x2_t create_circle_tbl(const uint8_t *const __restrict buffer[7], return tbl_circle_texels; } -inline uint8_t get_point_score(uint8_t p, uint8_t tolerance, const uint8x8x2_t &tbl_circle, uint8x8x2_t perm_indices[PERMUTATIONS]) +inline uint8_t get_point_score(uint8_t p, uint8_t tolerance, const uint8x8x2_t &tbl_circle, std::array<uint8x8x2_t, PERMUTATIONS> &perm_indices) { uint8_t b = 255; uint8_t a = tolerance; @@ -411,7 +409,7 @@ void NEFastCornersKernel::run(const Window &window, const ThreadInfo &info) Iterator in(_input, window); Iterator out(_output, window); - const uint8_t *const __restrict in_row[7] = + const std::array<uint8_t *const __restrict, 7> in_row { _input->ptr_to_element(Coordinates(-3, -3)), _input->ptr_to_element(Coordinates(-3, -2)), @@ -429,7 +427,7 @@ void NEFastCornersKernel::run(const Window &window, const ThreadInfo &info) return p_is_in_ab && q_is_in_ab; }; - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const size_t in_offset = in.offset(); const uint8_t p0 = *in.ptr(); @@ -455,11 +453,11 @@ void NEFastCornersKernel::run(const Window &window, const ThreadInfo &info) /* at this stage we use the full test with the 16 permutations to classify the point as corner or not */ const uint8x8x2_t tbl_circle_texel = create_circle_tbl(in_row, in_offset, circle_index_r); - if(point_is_fast_corner(p0, _threshold, tbl_circle_texel, perm_index.data())) + if(point_is_fast_corner(p0, _threshold, tbl_circle_texel, perm_index)) { if(_non_max_suppression) { - score = get_point_score(p0, _threshold, tbl_circle_texel, perm_index.data()); + score = get_point_score(p0, _threshold, tbl_circle_texel, perm_index); } else { diff --git a/src/core/NEON/kernels/NEFillInnerBorderKernel.cpp b/src/core/NEON/kernels/NEFillInnerBorderKernel.cpp index d1cff6f62d..50060b2376 100644 --- a/src/core/NEON/kernels/NEFillInnerBorderKernel.cpp +++ b/src/core/NEON/kernels/NEFillInnerBorderKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -111,7 +111,7 @@ void NEFillInnerBorderKernel::fill_value_single_channel(const Window &window) Iterator vertical_it(_tensor, vertical); - execute_window_loop(vertical, [&](const Coordinates & id) + execute_window_loop(vertical, [&](const Coordinates &) { std::fill_n(reinterpret_cast<T *>(vertical_it.ptr()), _border_size.left, constant_border_value); std::fill_n(reinterpret_cast<T *>(vertical_it.ptr()) + width - _border_size.right, _border_size.right, constant_border_value); @@ -122,7 +122,7 @@ void NEFillInnerBorderKernel::fill_value_single_channel(const Window &window) // All values are set at once Iterator horizontal_it(_tensor, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { for(size_t i = 0; i < _border_size.top; ++i) { diff --git a/src/core/NEON/kernels/NEFloorKernel.cpp b/src/core/NEON/kernels/NEFloorKernel.cpp index 6551d9ed57..43554a097c 100644 --- a/src/core/NEON/kernels/NEFloorKernel.cpp +++ b/src/core/NEON/kernels/NEFloorKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -111,7 +111,7 @@ void NEFloorKernel::run(const Window &window, const ThreadInfo &info) if(data_type == DataType::F32) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float32x4_t res = vfloorq_f32(vld1q_f32(reinterpret_cast<const float *>(input.ptr()))); vst1q_f32(reinterpret_cast<float *>(output.ptr()), res); @@ -121,7 +121,7 @@ void NEFloorKernel::run(const Window &window, const ThreadInfo &info) #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC else if(data_type == DataType::F16) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float16x8_t res = vfloorq_f16(vld1q_f16(reinterpret_cast<const float16_t *>(input.ptr()))); vst1q_f16(reinterpret_cast<float16_t *>(output.ptr()), res); diff --git a/src/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.cpp b/src/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.cpp index a100cd2bf6..b561d1e3f4 100644 --- a/src/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMLowpMatrixMultiplyKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017, 2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -479,7 +479,7 @@ void inline vector_matrix_multiply_s8(Iterator &ina, Iterator &inb, Iterator &ou void inline matrix_multiply_u8(Iterator &ina, Iterator &inb, Iterator &out, int width_b, size_t out_stride, const Window &window) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8_t *mtx_a0 = ina.ptr(); const uint8_t *mtx_b0 = inb.ptr(); @@ -599,7 +599,7 @@ void inline matrix_multiply_s8(Iterator &ina, Iterator &inb, Iterator &out, int // The implementation assumes that the matrix A and Matrix B have been reshaped respectively with NEGEMMInterleave4x4 and NEGEMMTranspose1xW // The reshaping of the matrices helps to have a cache friendly implementation and helps to avoid the data re-arrangements needed for computing 16x4 elements per iteration // All the values needed for computing a single 4x4 block will be read from consecutive memory positions - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { auto *mtx_a0 = reinterpret_cast<const int8_t *>(ina.ptr()); auto *mtx_b0 = reinterpret_cast<const int8_t *>(inb.ptr()); diff --git a/src/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.cpp b/src/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.cpp index 16a10a8463..46e53cec12 100644 --- a/src/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMLowpOffsetContributionOutputStageKernel.cpp @@ -444,7 +444,7 @@ void run_offset_contribution_output_stage(const Window &window, if(bias != nullptr) { Iterator bias_it = get_bias_it(collapsed_window, bias); - execute_window_loop(collapsed_window, [&](const Coordinates & id) + execute_window_loop(collapsed_window, [&](const Coordinates &) { run_offset_contribution_output_stage_window<false, false, true, is_bounded_relu, is_fixed_point>(nullptr, nullptr, reinterpret_cast<const int32_t *>(bias_it.ptr()), mm_result_it, out_it, result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset, @@ -454,7 +454,7 @@ void run_offset_contribution_output_stage(const Window &window, } else { - execute_window_loop(collapsed_window, [&](const Coordinates & id) + execute_window_loop(collapsed_window, [&](const Coordinates &) { run_offset_contribution_output_stage_window<false, false, false, is_bounded_relu, is_fixed_point>(nullptr, nullptr, nullptr, mm_result_it, out_it, result_offset_s32, result_shift_s32, min_u8, max_u8, a_offset, b_offset, k_offset, diff --git a/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.cpp b/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.cpp index d3cfc7a8fa..4906e6a987 100644 --- a/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel.cpp @@ -114,7 +114,7 @@ void NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel::run(const Window win_biases.set(Window::DimY, Window::Dimension(0, 1, 1)); Iterator bias(_bias, win_biases); - execute_window_loop(win_collapsed, [&](const Coordinates & id) + execute_window_loop(win_collapsed, [&](const Coordinates &) { // Compute 16 elements per iteration int x = window_start_x; @@ -165,7 +165,7 @@ void NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPointKernel::run(const Window } else { - execute_window_loop(win_collapsed, [&](const Coordinates & id) + execute_window_loop(win_collapsed, [&](const Coordinates &) { // Compute 16 elements per iteration int x = window_start_x; diff --git a/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel.cpp b/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel.cpp index 573373f6b1..a221bd7925 100644 --- a/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -183,7 +183,7 @@ void NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel::run(const Window &window) win_biases.set(Window::DimY, Window::Dimension(0, 1, 1)); Iterator bias(_bias, win_biases); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { // Compute 16 elements per iteration int x = window_start_x; @@ -245,7 +245,7 @@ void NEGEMMLowpQuantizeDownInt32ToUint8ScaleKernel::run(const Window &window) } else { - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { // Compute 16 elements per iteration int x = window_start_x; diff --git a/src/core/NEON/kernels/NEGEMMMatrixAccumulateBiasesKernel.cpp b/src/core/NEON/kernels/NEGEMMMatrixAccumulateBiasesKernel.cpp index 42353ed0eb..5ac2323896 100644 --- a/src/core/NEON/kernels/NEGEMMMatrixAccumulateBiasesKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMMatrixAccumulateBiasesKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -122,7 +122,7 @@ void NEGEMMMatrixAccumulateBiasesKernel::run(const Window &window, const ThreadI { case DataType::F32: { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float32x4x4_t accum = vld4q_f32(reinterpret_cast<const float *>(in0_out.ptr())); const float32x4x4_t biases = vld4q_f32(reinterpret_cast<const float *>(in1.ptr())); @@ -144,7 +144,7 @@ void NEGEMMMatrixAccumulateBiasesKernel::run(const Window &window, const ThreadI #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const float16x8x2_t accum = vld2q_f16(reinterpret_cast<const float16_t *>(in0_out.ptr())); const float16x8x2_t biases = vld2q_f16(reinterpret_cast<const float16_t *>(in1.ptr())); diff --git a/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp b/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp index 757dbbc399..86bea849e4 100644 --- a/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMMatrixAdditionKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -60,7 +60,7 @@ void matrix_addition_f32(const ITensor *input, ITensor *output, const Window &wi Iterator in(input, window); Iterator out(output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const auto in_ptr = reinterpret_cast<const float *>(in.ptr()); const auto out_ptr = reinterpret_cast<float *>(out.ptr()); @@ -87,7 +87,7 @@ void matrix_addition_f16(const ITensor *input, ITensor *output, const Window &wi Iterator in(input, window); Iterator out(output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const auto in_ptr = reinterpret_cast<const float16_t *>(in.ptr()); const auto out_ptr = reinterpret_cast<float16_t *>(out.ptr()); diff --git a/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp b/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp index f182fb24c3..a82fae7521 100644 --- a/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp +++ b/src/core/NEON/kernels/NEGEMMMatrixMultiplyKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -388,7 +388,7 @@ void matrix_matrix_multiply_f32(const ITensor *input0, const ITensor *input1, IT // The implementation assumes that the matrix A and Matrix B have been reshaped respectively with NEGEMMInterleave4x4 and NEGEMMTranspose1xW // The reshaping of the matrices helps to have a cache friendly implementation and helps to avoid the data re-arrangements needed for computing 16x4 elements per iteration // All the values needed for computing a single 4x4 block will be read from consecutive memory positions - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { auto mtx_a0 = reinterpret_cast<const float *>(ina.ptr()); auto mtx_b0 = reinterpret_cast<const float *>(inb.ptr()); @@ -687,7 +687,7 @@ void matrix_matrix_multiply_f16(const ITensor *input0, const ITensor *input1, IT const float16x8_t alpha_f16 = vdupq_n_f16(alpha); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const auto *mtx_a0 = reinterpret_cast<const float16_t *>(ina.ptr()); const auto *mtx_b0 = reinterpret_cast<const float16_t *>(inb.ptr()); diff --git a/src/core/NEON/kernels/NEGaussian3x3Kernel.cpp b/src/core/NEON/kernels/NEGaussian3x3Kernel.cpp index 048c22933c..f41298067f 100644 --- a/src/core/NEON/kernels/NEGaussian3x3Kernel.cpp +++ b/src/core/NEON/kernels/NEGaussian3x3Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -80,7 +80,7 @@ void NEGaussian3x3Kernel::run(const Window &window, const ThreadInfo &info) static const int16x8_t two = vdupq_n_s16(2); static const int16x8_t four = vdupq_n_s16(4); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); diff --git a/src/core/NEON/kernels/NEGaussian5x5Kernel.cpp b/src/core/NEON/kernels/NEGaussian5x5Kernel.cpp index b62e2816c0..0e4549e640 100644 --- a/src/core/NEON/kernels/NEGaussian5x5Kernel.cpp +++ b/src/core/NEON/kernels/NEGaussian5x5Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -88,7 +88,7 @@ void NEGaussian5x5HorKernel::run(const Window &window, const ThreadInfo &info) static const int16x8_t six = vdupq_n_s16(6); static const int16x8_t four = vdupq_n_s16(4); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { uint8x16_t data = vld1q_u8(input.ptr()); @@ -112,7 +112,7 @@ void NEGaussian5x5HorKernel::run(const Window &window, const ThreadInfo &info) BorderSize NEGaussian5x5VertKernel::border_size() const { - return BorderSize(2, 0); + return BorderSize{ 2, 0 }; } void NEGaussian5x5VertKernel::configure(const ITensor *input, ITensor *output, bool border_undefined) @@ -159,7 +159,7 @@ void NEGaussian5x5VertKernel::run(const Window &window, const ThreadInfo &info) const uint16x8_t six = vdupq_n_u16(6); const uint16x8_t four = vdupq_n_u16(4); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const size_t input_offset_high_s16 = input.offset(); const size_t input_offset_low_s16 = input.offset() + 16; diff --git a/src/core/NEON/kernels/NEGaussianPyramidKernel.cpp b/src/core/NEON/kernels/NEGaussianPyramidKernel.cpp index 7a123e2f57..13cee19f5c 100644 --- a/src/core/NEON/kernels/NEGaussianPyramidKernel.cpp +++ b/src/core/NEON/kernels/NEGaussianPyramidKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -47,7 +47,7 @@ NEGaussianPyramidHorKernel::NEGaussianPyramidHorKernel() BorderSize NEGaussianPyramidHorKernel::border_size() const { - return BorderSize(0, 2); + return BorderSize{ 0, 2 }; } void NEGaussianPyramidHorKernel::configure(const ITensor *input, ITensor *output) @@ -126,7 +126,7 @@ void NEGaussianPyramidHorKernel::run(const Window &window, const ThreadInfo &inf Iterator out(_output, win_out); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16x2_t data_2q = vld2q_u8(in.ptr()); const uint8x16_t &data_even = data_2q.val[0]; @@ -155,7 +155,7 @@ NEGaussianPyramidVertKernel::NEGaussianPyramidVertKernel() BorderSize NEGaussianPyramidVertKernel::border_size() const { - return BorderSize(2, 0); + return BorderSize{ 2, 0 }; } void NEGaussianPyramidVertKernel::configure(const ITensor *input, ITensor *output) @@ -236,7 +236,7 @@ void NEGaussianPyramidVertKernel::run(const Window &window, const ThreadInfo &in const uint8_t *input_low_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(0, 3)); const uint8_t *input_low2_ptr = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(0, 4)); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Low data const uint16x8_t data_low_t2 = vreinterpretq_u16_s16(vld1q_s16(reinterpret_cast<const int16_t *>(input_top2_ptr + in.offset()))); diff --git a/src/core/NEON/kernels/NEHOGDescriptorKernel.cpp b/src/core/NEON/kernels/NEHOGDescriptorKernel.cpp index c204395586..c58b1c024a 100644 --- a/src/core/NEON/kernels/NEHOGDescriptorKernel.cpp +++ b/src/core/NEON/kernels/NEHOGDescriptorKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -695,7 +695,7 @@ void NEHOGOrientationBinningKernel::run(const Window &window, const ThreadInfo & Iterator phase(_input_phase, win_phase); Iterator out(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const auto mag_row_ptr = reinterpret_cast<const int16_t *>(mag.ptr()); const auto phase_row_ptr = reinterpret_cast<const uint8_t *>(phase.ptr()); diff --git a/src/core/NEON/kernels/NEHarrisCornersKernel.cpp b/src/core/NEON/kernels/NEHarrisCornersKernel.cpp index 61221c1070..34e68e71cb 100644 --- a/src/core/NEON/kernels/NEHarrisCornersKernel.cpp +++ b/src/core/NEON/kernels/NEHarrisCornersKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -719,7 +719,7 @@ void NEHarrisScoreKernel<block_size>::run(const Window &window, const ThreadInfo const size_t input_stride = _input1->info()->strides_in_bytes()[1] / element_size_from_data_type(_input1->info()->data_type()); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { (*_func)(input1.ptr(), input2.ptr(), output.ptr(), input_stride, _norm_factor, _sensitivity, _strength_thresh); }, diff --git a/src/core/NEON/kernels/NEIntegralImageKernel.cpp b/src/core/NEON/kernels/NEIntegralImageKernel.cpp index 16a3cf7f07..b6db5f0e6a 100644 --- a/src/core/NEON/kernels/NEIntegralImageKernel.cpp +++ b/src/core/NEON/kernels/NEIntegralImageKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -63,7 +63,7 @@ void NEIntegralImageKernel::configure(const ITensor *input, ITensor *output) BorderSize NEIntegralImageKernel::border_size() const { - return BorderSize(1, 0, 0, 1); + return BorderSize{ 1, 0, 0, 1 }; } bool NEIntegralImageKernel::is_parallelisable() const @@ -83,7 +83,7 @@ void NEIntegralImageKernel::run(const Window &window, const ThreadInfo &info) const auto output_top_left = reinterpret_cast<const uint32_t *>(_output->ptr_to_element(Coordinates(-1, -1))); const auto output_top_mid = reinterpret_cast<const uint32_t *>(_output->ptr_to_element(Coordinates(0, -1))); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t input_pixels = vld1q_u8(input.ptr()); diff --git a/src/core/NEON/kernels/NEL2NormalizeLayerKernel.cpp b/src/core/NEON/kernels/NEL2NormalizeLayerKernel.cpp index cda041de66..efdcc44e0e 100644 --- a/src/core/NEON/kernels/NEL2NormalizeLayerKernel.cpp +++ b/src/core/NEON/kernels/NEL2NormalizeLayerKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -61,7 +61,7 @@ void l2_normalize_X(const ITensor *in, const ITensor *sum, ITensor *out, float e const auto sum_value = *reinterpret_cast<const T *>(sum_it.ptr()); const auto vec_normalize_value = wrapper::vdup_n(static_cast<T>(1.f / std::sqrt(std::max(sum_value, static_cast<T>(epsilon)))), ExactTagType{}); - execute_window_loop(in_slice, [&](const Coordinates & id) + execute_window_loop(in_slice, [&](const Coordinates &) { const auto in_ptr = reinterpret_cast<const T *>(input_it.ptr()); const auto out_ptr = reinterpret_cast<T *>(output_it.ptr()); @@ -93,7 +93,7 @@ void l2_normalize_Y(const ITensor *in, const ITensor *sum, ITensor *out, float e auto eps = wrapper::vdup_n(static_cast<T>(epsilon), ExactTagType{}); - execute_window_loop(in_slice, [&](const Coordinates & id) + execute_window_loop(in_slice, [&](const Coordinates &) { const auto in_ptr = reinterpret_cast<const T *>(input_it.ptr()); const auto sum_ptr = reinterpret_cast<const T *>(sum_it.ptr()); @@ -127,7 +127,7 @@ void l2_normalize_Z(const ITensor *in, const ITensor *sum, ITensor *out, float e auto eps = wrapper::vdup_n(static_cast<T>(epsilon), ExactTagType{}); - execute_window_loop(in_slice, [&](const Coordinates & id) + execute_window_loop(in_slice, [&](const Coordinates &) { const auto in_ptr = reinterpret_cast<const T *>(input_it.ptr()); const auto sum_ptr = reinterpret_cast<const T *>(sum_it.ptr()); diff --git a/src/core/NEON/kernels/NEMagnitudePhaseKernel.cpp b/src/core/NEON/kernels/NEMagnitudePhaseKernel.cpp index 4a318f02c1..8c09898403 100644 --- a/src/core/NEON/kernels/NEMagnitudePhaseKernel.cpp +++ b/src/core/NEON/kernels/NEMagnitudePhaseKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -323,7 +323,7 @@ void NEMagnitudePhaseKernel<mag_type, phase_type>::magnitude(const Window &windo Iterator gy(_gy, window); Iterator magnitude(_magnitude, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t input1 = { @@ -369,7 +369,7 @@ void NEMagnitudePhaseKernel<mag_type, phase_type>::phase(const Window &window) Iterator gy(_gy, window); Iterator phase(_phase, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t input1 = { @@ -415,7 +415,7 @@ void NEMagnitudePhaseKernel<mag_type, phase_type>::magnitude_phase(const Window Iterator magnitude(_magnitude, window); Iterator phase(_phase, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const int16x8x2_t input1 = { diff --git a/src/core/NEON/kernels/NEMeanStdDevKernel.cpp b/src/core/NEON/kernels/NEMeanStdDevKernel.cpp index 7895b009d6..0af63059fb 100644 --- a/src/core/NEON/kernels/NEMeanStdDevKernel.cpp +++ b/src/core/NEON/kernels/NEMeanStdDevKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -51,7 +51,7 @@ std::pair<uint64x1_t, uint64x1_t> accumulate(const Window &window, Iterator &ite uint64x1_t sum_squared = vdup_n_u64(0); // Calculate sum - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t in_data = vld1q_u8(iterator.ptr()); diff --git a/src/core/NEON/kernels/NEMedian3x3Kernel.cpp b/src/core/NEON/kernels/NEMedian3x3Kernel.cpp index 5bcdc7bc7a..9dc1bc9a36 100644 --- a/src/core/NEON/kernels/NEMedian3x3Kernel.cpp +++ b/src/core/NEON/kernels/NEMedian3x3Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -87,7 +87,7 @@ void NEMedian3x3Kernel::run(const Window &window, const ThreadInfo &info) Iterator input(_input, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); diff --git a/src/core/NEON/kernels/NEMemsetKernel.cpp b/src/core/NEON/kernels/NEMemsetKernel.cpp index 2b57b1595b..a0fab99c55 100644 --- a/src/core/NEON/kernels/NEMemsetKernel.cpp +++ b/src/core/NEON/kernels/NEMemsetKernel.cpp @@ -67,7 +67,7 @@ void NEMemsetKernel::run(const Window &window, const ThreadInfo &info) collapsed.set(Window::DimX, Window::Dimension(0, 1, 1)); Iterator tensor_it(_tensor, collapsed); - execute_window_loop(collapsed, [&](const Coordinates & id) + execute_window_loop(collapsed, [&](const Coordinates &) { uint8_t *base_addr = start_valid_region + tensor_it.offset(); // Set memory diff --git a/src/core/NEON/kernels/NEMinMaxLayerKernel.cpp b/src/core/NEON/kernels/NEMinMaxLayerKernel.cpp index 5d1b4b3aa4..fe3af0b44f 100644 --- a/src/core/NEON/kernels/NEMinMaxLayerKernel.cpp +++ b/src/core/NEON/kernels/NEMinMaxLayerKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2017-2018 ARM Limited. + * Copyright (c) 2017-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -144,7 +144,7 @@ void NEMinMaxLayerKernel::run(const Window &window, const ThreadInfo &info) float carry_min_scalar = std::numeric_limits<float>::max(); float carry_max_scalar = std::numeric_limits<float>::lowest(); - execute_window_loop(window_input, [&](const Coordinates & id) + execute_window_loop(window_input, [&](const Coordinates &) { int x = x_start; const auto in_ptr = reinterpret_cast<const float *>(input.ptr() + id_batch[1] * _input->info()->strides_in_bytes()[3]); @@ -203,7 +203,7 @@ void NEMinMaxLayerKernel::reset() Iterator output(_output, window_output); - execute_window_loop(window_output, [&](const Coordinates & id) + execute_window_loop(window_output, [&](const Coordinates &) { vst1_f32(reinterpret_cast<float *>(output.ptr()), reset_values); }, diff --git a/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp b/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp index befece2741..08b27e319e 100644 --- a/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp +++ b/src/core/NEON/kernels/NEMinMaxLocationKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -152,7 +152,7 @@ void NEMinMaxKernel::minmax_U8(Window win) Iterator input(_input, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int x = x_start; @@ -209,7 +209,7 @@ void NEMinMaxKernel::minmax_S16(Window win) Iterator input(_input, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int x = x_start; const auto in_ptr = reinterpret_cast<const int16_t *>(input.ptr()); @@ -268,7 +268,7 @@ void NEMinMaxKernel::minmax_F32(Window win) Iterator input(_input, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { int x = x_start; const auto in_ptr = reinterpret_cast<const float *>(input.ptr()); @@ -323,11 +323,11 @@ bool NEMinMaxLocationKernel::is_parallelisable() const template <class T, std::size_t... N> struct NEMinMaxLocationKernel::create_func_table<T, utility::index_sequence<N...>> { - static const NEMinMaxLocationKernel::MinMaxLocFunction func_table[sizeof...(N)]; + static const std::array<NEMinMaxLocationKernel::MinMaxLocFunction, sizeof...(N)> func_table; }; template <class T, std::size_t... N> -const NEMinMaxLocationKernel::MinMaxLocFunction NEMinMaxLocationKernel::create_func_table<T, utility::index_sequence<N...>>::func_table[sizeof...(N)] = +const std::array<NEMinMaxLocationKernel::MinMaxLocFunction, sizeof...(N)> NEMinMaxLocationKernel::create_func_table<T, utility::index_sequence<N...>>::func_table { &NEMinMaxLocationKernel::minmax_loc<T, bool(N & 8), bool(N & 4), bool(N & 2), bool(N & 1)>... }; diff --git a/src/core/NEON/kernels/NENonLinearFilterKernel.cpp b/src/core/NEON/kernels/NENonLinearFilterKernel.cpp index 52dbe26f42..00536f092c 100644 --- a/src/core/NEON/kernels/NENonLinearFilterKernel.cpp +++ b/src/core/NEON/kernels/NENonLinearFilterKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016-2018 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -117,7 +117,7 @@ inline void sort9(uint8x8_t &p0, uint8x8_t &p1, uint8x8_t &p2, sort(p4, p2); } -inline void sort21(uint8x8_t p[21]) +inline void sort21(std::array<uint8x8_t, 21> &p) { sort(p[0], p[1]); sort(p[2], p[3]); @@ -222,7 +222,7 @@ inline void sort21(uint8x8_t p[21]) sort(p[10], p[16]); } -inline void sort25(uint8x8_t p[25]) +inline void sort25(std::array<uint8x8_t, 25> &p) { sort(p[1], p[2]); sort(p[0], p[1]); @@ -429,7 +429,7 @@ void NENonLinearFilterKernel::median_filter_box<3, 3>(const Window &win) const auto input_mid_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-1, 0))); const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-1, 1))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); @@ -463,7 +463,7 @@ void NENonLinearFilterKernel::median_filter_box<5, 5>(const Window &win) const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 1))); const auto input_bot2_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 2))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x16_t top2_data = vld1q_u8(input_top2_ptr + input.offset()); const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); @@ -471,7 +471,7 @@ void NENonLinearFilterKernel::median_filter_box<5, 5>(const Window &win) const uint8x16_t bot_data = vld1q_u8(input_bot_ptr + input.offset()); const uint8x16_t bot2_data = vld1q_u8(input_bot2_ptr + input.offset()); - const uint8x8_t d[] = + const std::array<uint8x8_t, 10> d = { vget_low_u8(top2_data), vget_high_u8(top2_data), @@ -485,7 +485,7 @@ void NENonLinearFilterKernel::median_filter_box<5, 5>(const Window &win) vget_high_u8(bot2_data) }; - uint8x8_t p[25]; + std::array<uint8x8_t, 25> p{ 0 }; for(unsigned int i = 0; i < 5; ++i) { const unsigned int idx_d = i * 2; @@ -524,7 +524,7 @@ void NENonLinearFilterKernel::min_filter_box(const Window &win) input_ptrs[k_row_half + i] = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-k_col_half, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { // Get min of rows uint8x16_t rows_min = vld1q_u8(input_ptrs[0] + input.offset()); @@ -563,7 +563,7 @@ void NENonLinearFilterKernel::max_filter_box(const Window &win) input_ptrs[k_row_half + i] = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(-k_col_half, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { uint8x16_t rows_max = vld1q_u8(input_ptrs[0] + input.offset()); @@ -593,7 +593,7 @@ void NENonLinearFilterKernel::median_filter_cross<3, 3>(const Window &win) const auto input_mid_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-1, 0))); const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(0, 1))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x8_t top_data = vld1_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); @@ -624,7 +624,7 @@ void NENonLinearFilterKernel::median_filter_cross<5, 5>(const Window &win) const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(0, 1))); const auto input_bot2_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(0, 2))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x8_t top2_data = vld1_u8(input_top2_ptr + input.offset()); const uint8x8_t top_data = vld1_u8(input_top_ptr + input.offset()); @@ -671,7 +671,7 @@ void NENonLinearFilterKernel::min_filter_cross(const Window &win) input_ptrs[k_row_half + i] = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(0, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { uint8x8_t rows_min = vld1_u8(input_ptrs[0] + input.offset()); @@ -717,7 +717,7 @@ void NENonLinearFilterKernel::max_filter_cross(const Window &win) input_ptrs[k_row_half + i] = _input->buffer() + _input->info()->offset_element_in_bytes(Coordinates(0, i)); } - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { uint8x8_t rows_max = vld1_u8(input_ptrs[0] + input.offset()); @@ -754,7 +754,7 @@ void NENonLinearFilterKernel::median_filter_disk<5, 5>(const Window &win) const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 1))); const auto input_bot2_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 2))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x16_t top2_data = vextq_u8(vld1q_u8(input_top2_ptr + input.offset()), zero, 1); const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); @@ -762,7 +762,7 @@ void NENonLinearFilterKernel::median_filter_disk<5, 5>(const Window &win) const uint8x16_t bot_data = vld1q_u8(input_bot_ptr + input.offset()); const uint8x16_t bot2_data = vextq_u8(vld1q_u8(input_bot2_ptr + input.offset()), zero, 1); - uint8x8_t d[] = + std::array<uint8x8_t, 10> d = { vget_low_u8(top2_data), vget_high_u8(top2_data), @@ -776,7 +776,7 @@ void NENonLinearFilterKernel::median_filter_disk<5, 5>(const Window &win) vget_high_u8(bot2_data) }; - uint8x8_t p[21]; + std::array<uint8x8_t, 21> p{ 0 }; p[0] = d[0]; p[1] = vext_u8(d[0], d[1], 1); p[2] = vext_u8(d[0], d[1], 2); @@ -816,7 +816,7 @@ void NENonLinearFilterKernel::min_filter_disk<5, 5>(const Window &win) const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 1))); const auto input_bot2_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 2))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x16_t top2_data = vextq_u8(vld1q_u8(input_top2_ptr + input.offset()), zero, 1); const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); @@ -849,7 +849,7 @@ void NENonLinearFilterKernel::max_filter_disk<5, 5>(const Window &win) const auto input_bot_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 1))); const auto input_bot2_ptr = static_cast<const unsigned char *>(_input->ptr_to_element(Coordinates(-2, 2))); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { const uint8x16_t top2_data = vextq_u8(vld1q_u8(input_top2_ptr + input.offset()), zero, 1); const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); @@ -889,7 +889,7 @@ void NENonLinearFilterKernel::non_linear_filter_generic(const Window &win) std::array<uint8_t, mask_size> vals{ {} }; - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { // Clear array std::fill(std::begin(vals), std::end(vals), 0); diff --git a/src/core/NEON/kernels/NENonMaximaSuppression3x3Kernel.cpp b/src/core/NEON/kernels/NENonMaximaSuppression3x3Kernel.cpp index 8f97e6ac16..674a7c88cc 100644 --- a/src/core/NEON/kernels/NENonMaximaSuppression3x3Kernel.cpp +++ b/src/core/NEON/kernels/NENonMaximaSuppression3x3Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -506,7 +506,7 @@ void NENonMaximaSuppression3x3Kernel::run(const Window &window, const ThreadInfo const size_t input_stride = _input->info()->strides_in_bytes()[1] / element_size_from_data_type(_input->info()->data_type()); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { _func(input.ptr(), output.ptr(), input_stride); }, diff --git a/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp b/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp index e3166e02b6..b565300906 100644 --- a/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp +++ b/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp @@ -627,7 +627,7 @@ void NEPixelWiseMultiplicationKernel::run(const Window &window, const ThreadInfo if(_func_qasymm8 != nullptr) { - execute_window_loop(collapsed, [&](const Coordinates & id) + execute_window_loop(collapsed, [&](const Coordinates &) { (*_func_qasymm8)(input1.ptr(), input2.ptr(), output.ptr(), _scale, _input1->info()->quantization_info(), _input2->info()->quantization_info(), _output->info()->quantization_info()); @@ -638,7 +638,7 @@ void NEPixelWiseMultiplicationKernel::run(const Window &window, const ThreadInfo } else if(_func_int != nullptr) { - execute_window_loop(collapsed, [&](const Coordinates & id) + execute_window_loop(collapsed, [&](const Coordinates &) { (*_func_int)(input1.ptr(), input2.ptr(), output.ptr(), _scale_exponent); collapsed.slide_window_slice_3D(slice_input1); @@ -649,7 +649,7 @@ void NEPixelWiseMultiplicationKernel::run(const Window &window, const ThreadInfo else { ARM_COMPUTE_ERROR_ON(_func_float == nullptr); - execute_window_loop(collapsed, [&](const Coordinates & id) + execute_window_loop(collapsed, [&](const Coordinates &) { (*_func_float)(input1.ptr(), input2.ptr(), output.ptr(), _scale); collapsed.slide_window_slice_3D(slice_input1); @@ -663,6 +663,6 @@ BorderSize NEPixelWiseMultiplicationKernel::border_size() const { const unsigned int replicateSize = _output->info()->dimension(0) - std::min(_input1->info()->dimension(0), _input2->info()->dimension(0)); const unsigned int border = std::min<unsigned int>(num_elems_processed_per_iteration - 1U, replicateSize); - return BorderSize(0, border, 0, 0); + return BorderSize{ 0, border, 0, 0 }; } } // namespace arm_compute diff --git a/src/core/NEON/kernels/NEQuantizationLayerKernel.cpp b/src/core/NEON/kernels/NEQuantizationLayerKernel.cpp index 136457c34e..4deeb1c7cc 100644 --- a/src/core/NEON/kernels/NEQuantizationLayerKernel.cpp +++ b/src/core/NEON/kernels/NEQuantizationLayerKernel.cpp @@ -51,7 +51,7 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output) return Status{}; } -inline const float32x4x4_t load_value(const float *input_ptr) +inline float32x4x4_t load_value(const float *input_ptr) { return { wrapper::vloadq(input_ptr), wrapper::vloadq(input_ptr + 4), @@ -119,7 +119,7 @@ void NEQuantizationLayerKernel::quantize(const Window &window, const Quantizatio Iterator input(_input, win_collapsed); Iterator output(_output, win_collapsed); - execute_window_loop(win_collapsed, [&](const Coordinates & id) + execute_window_loop(win_collapsed, [&](const Coordinates &) { auto input_ptr = reinterpret_cast<const T *>(input.ptr()); auto output_ptr = reinterpret_cast<uint8_t *>(output.ptr()); diff --git a/src/core/NEON/kernels/NEScharr3x3Kernel.cpp b/src/core/NEON/kernels/NEScharr3x3Kernel.cpp index f23c31bc0a..3add699f62 100644 --- a/src/core/NEON/kernels/NEScharr3x3Kernel.cpp +++ b/src/core/NEON/kernels/NEScharr3x3Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -161,7 +161,7 @@ void NEScharr3x3Kernel::run(const Window &window, const ThreadInfo &info) if(_run_scharr_x && _run_scharr_y) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); @@ -197,7 +197,7 @@ void NEScharr3x3Kernel::run(const Window &window, const ThreadInfo &info) } else if(_run_scharr_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); @@ -232,7 +232,7 @@ void NEScharr3x3Kernel::run(const Window &window, const ThreadInfo &info) } else if(_run_scharr_y) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); diff --git a/src/core/NEON/kernels/NESelectKernel.cpp b/src/core/NEON/kernels/NESelectKernel.cpp index f2697bcc6d..c03e5f0bca 100644 --- a/src/core/NEON/kernels/NESelectKernel.cpp +++ b/src/core/NEON/kernels/NESelectKernel.cpp @@ -54,7 +54,7 @@ void select_op(const ITensor *cond, const ITensor *in1, const ITensor *in2, ITen Iterator input2(in2, win); Iterator output(out, win); - execute_window_loop(win, [&](const Coordinates & id) + execute_window_loop(win, [&](const Coordinates &) { auto output_ptr = reinterpret_cast<ScalarType *>(output.ptr()); const auto condition_ptr = reinterpret_cast<const uint8_t *>(condition.ptr()); diff --git a/src/core/NEON/kernels/NESobel3x3Kernel.cpp b/src/core/NEON/kernels/NESobel3x3Kernel.cpp index 5a80630a76..7a27203bef 100644 --- a/src/core/NEON/kernels/NESobel3x3Kernel.cpp +++ b/src/core/NEON/kernels/NESobel3x3Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -42,7 +42,7 @@ NESobel3x3Kernel::NESobel3x3Kernel() BorderSize NESobel3x3Kernel::border_size() const { - return BorderSize(1); + return BorderSize{ 1 }; } void NESobel3x3Kernel::configure(const ITensor *input, ITensor *output_x, ITensor *output_y, bool border_undefined) @@ -117,7 +117,7 @@ void NESobel3x3Kernel::run(const Window &window, const ThreadInfo &info) if(_run_sobel_y && _run_sobel_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); @@ -181,7 +181,7 @@ void NESobel3x3Kernel::run(const Window &window, const ThreadInfo &info) } else if(_run_sobel_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t mid_data = vld1q_u8(input_mid_ptr + input.offset()); @@ -229,7 +229,7 @@ void NESobel3x3Kernel::run(const Window &window, const ThreadInfo &info) } else if(_run_sobel_y) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t top_data = vld1q_u8(input_top_ptr + input.offset()); const uint8x16_t bot_data = vld1q_u8(input_bot_ptr + input.offset()); diff --git a/src/core/NEON/kernels/NESobel5x5Kernel.cpp b/src/core/NEON/kernels/NESobel5x5Kernel.cpp index 30e7817aa4..a92cfc2308 100644 --- a/src/core/NEON/kernels/NESobel5x5Kernel.cpp +++ b/src/core/NEON/kernels/NESobel5x5Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -120,7 +120,7 @@ void NESobel5x5HorKernel::run(const Window &window, const ThreadInfo &info) static const int16x8_t two = vdupq_n_s16(2); static const int16x8_t minustwo = vdupq_n_s16(-2); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -154,7 +154,7 @@ void NESobel5x5HorKernel::run(const Window &window, const ThreadInfo &info) static const int16x8_t two = vdupq_n_s16(2); static const int16x8_t minustwo = vdupq_n_s16(-2); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -180,7 +180,7 @@ void NESobel5x5HorKernel::run(const Window &window, const ThreadInfo &info) static const int16x8_t six = vdupq_n_s16(6); static const int16x8_t four = vdupq_n_s16(4); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); @@ -211,7 +211,7 @@ NESobel5x5VertKernel::NESobel5x5VertKernel() BorderSize NESobel5x5VertKernel::border_size() const { - return BorderSize(2, 0); + return BorderSize{ 2, 0 }; } void NESobel5x5VertKernel::configure(ITensor *input_x, ITensor *input_y, ITensor *output_x, ITensor *output_y, bool border_undefined) @@ -312,7 +312,7 @@ void NESobel5x5VertKernel::run(const Window &window, const ThreadInfo &info) if(_run_sobel_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Convert offset from uint8_t* to uint16_t* const size_t input_offset_high_s16 = input_x.offset() / 2; @@ -361,7 +361,7 @@ void NESobel5x5VertKernel::run(const Window &window, const ThreadInfo &info) if(_run_sobel_y) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { // Convert offset from uint8_t* to uint16_t* const size_t input_offset_high_s16 = input_y.offset() / 2; diff --git a/src/core/NEON/kernels/NESobel7x7Kernel.cpp b/src/core/NEON/kernels/NESobel7x7Kernel.cpp index 40a3e31a39..f2b42cc5dd 100644 --- a/src/core/NEON/kernels/NESobel7x7Kernel.cpp +++ b/src/core/NEON/kernels/NESobel7x7Kernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -215,7 +215,7 @@ void NESobel7x7HorKernel::run(const Window &window, const ThreadInfo &info) if(_run_sobel_y && _run_sobel_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr() - 3); @@ -244,7 +244,7 @@ void NESobel7x7HorKernel::run(const Window &window, const ThreadInfo &info) } else if(_run_sobel_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr() - 3); @@ -269,7 +269,7 @@ void NESobel7x7HorKernel::run(const Window &window, const ThreadInfo &info) } else if(_run_sobel_y) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr() - 3); @@ -301,7 +301,7 @@ NESobel7x7VertKernel::NESobel7x7VertKernel() BorderSize NESobel7x7VertKernel::border_size() const { - return BorderSize(3, 0); + return BorderSize{ 3, 0 }; } void NESobel7x7VertKernel::configure(const ITensor *input_x, const ITensor *input_y, ITensor *output_x, ITensor *output_y, bool border_undefined) @@ -382,7 +382,7 @@ void NESobel7x7VertKernel::run(const Window &window, const ThreadInfo &info) if(_run_sobel_x) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { auto in_ptr = reinterpret_cast<int32_t *>(input_x.ptr()) - 3 * in_x_stride; @@ -453,7 +453,7 @@ void NESobel7x7VertKernel::run(const Window &window, const ThreadInfo &info) if(_run_sobel_y) { - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { auto in_ptr = reinterpret_cast<int32_t *>(input_y.ptr()) - 3 * in_y_stride; diff --git a/src/core/NEON/kernels/NETableLookupKernel.cpp b/src/core/NEON/kernels/NETableLookupKernel.cpp index 958f4a9cfb..536c2201c2 100644 --- a/src/core/NEON/kernels/NETableLookupKernel.cpp +++ b/src/core/NEON/kernels/NETableLookupKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -61,7 +61,7 @@ void NETableLookupKernel::tableLookup(const Window &window) Iterator input = Iterator(_input, window); Iterator output = Iterator(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { auto input_ptr = reinterpret_cast<const T *>(input.ptr()); auto output_ptr = reinterpret_cast<T *>(output.ptr()); @@ -92,7 +92,7 @@ void NETableLookupKernel::tableLookup<uint8_t>(const Window &window) Iterator input = Iterator(_input, window); Iterator output = Iterator(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8_t *input_ptr = input.ptr(); uint8_t *output_ptr = output.ptr(); diff --git a/src/core/NEON/kernels/NEThresholdKernel.cpp b/src/core/NEON/kernels/NEThresholdKernel.cpp index 5ef06931cc..ae9c62bc92 100644 --- a/src/core/NEON/kernels/NEThresholdKernel.cpp +++ b/src/core/NEON/kernels/NEThresholdKernel.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2016, 2017 ARM Limited. + * Copyright (c) 2016-2019 ARM Limited. * * SPDX-License-Identifier: MIT * @@ -86,7 +86,7 @@ inline void NEThresholdKernel::run_binary(const Window &window) Iterator input(_input, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); const uint8x16_t mask = vcgtq_u8(data, threshold); @@ -106,7 +106,7 @@ inline void NEThresholdKernel::run_range(const Window &window) Iterator input(_input, window); Iterator output(_output, window); - execute_window_loop(window, [&](const Coordinates & id) + execute_window_loop(window, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(input.ptr()); diff --git a/src/core/NEON/kernels/NEUpsampleLayerKernel.cpp b/src/core/NEON/kernels/NEUpsampleLayerKernel.cpp index aae85c6741..d3d88b3bf0 100644 --- a/src/core/NEON/kernels/NEUpsampleLayerKernel.cpp +++ b/src/core/NEON/kernels/NEUpsampleLayerKernel.cpp @@ -130,7 +130,7 @@ void NEUpsampleLayerKernel::upsample_f32_nchw(const arm_compute::Window &window) Iterator output(_output, window_out); const int offset_y_out = _output->info()->strides_in_bytes().y() / sizeof(float); - execute_window_loop(window_out, [&](const Coordinates & id) + execute_window_loop(window_out, [&](const Coordinates &) { const float32x4_t data = vld1q_f32(reinterpret_cast<const float *>(input.ptr())); const float32x4_t data_out1 = { vgetq_lane_f32(data, 0), vgetq_lane_f32(data, 0), vgetq_lane_f32(data, 1), vgetq_lane_f32(data, 1) }; @@ -157,7 +157,7 @@ void NEUpsampleLayerKernel::upsample_f32_nhwc(const arm_compute::Window &window) const int offset_y_out = _output->info()->strides_in_bytes().y() / sizeof(float); const int offset_z_out = _output->info()->strides_in_bytes().z() / sizeof(float); - execute_window_loop(window_out, [&](const Coordinates & id) + execute_window_loop(window_out, [&](const Coordinates &) { const float32x4_t data = vld1q_f32(reinterpret_cast<const float *>(input.ptr())); auto out = reinterpret_cast<float *>(output.ptr()); @@ -182,7 +182,7 @@ void NEUpsampleLayerKernel::upsample_qasymm8_nchw(const arm_compute::Window &win Iterator output(_output, window_out); const int offset_y_out = _output->info()->strides_in_bytes().y() / sizeof(uint8_t); - execute_window_loop(window_out, [&](const Coordinates & id) + execute_window_loop(window_out, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(reinterpret_cast<const uint8_t *>(input.ptr())); const uint8x16_t data_out1 = { vgetq_lane_u8(data, 0), vgetq_lane_u8(data, 0), vgetq_lane_u8(data, 1), vgetq_lane_u8(data, 1), @@ -218,7 +218,7 @@ void NEUpsampleLayerKernel::upsample_qasymm8_nhwc(const arm_compute::Window &win const int offset_y_out = _output->info()->strides_in_bytes().y() / sizeof(uint8_t); const int offset_z_out = _output->info()->strides_in_bytes().z() / sizeof(uint8_t); - execute_window_loop(window_out, [&](const Coordinates & id) + execute_window_loop(window_out, [&](const Coordinates &) { const uint8x16_t data = vld1q_u8(reinterpret_cast<const uint8_t *>(input.ptr())); auto out = reinterpret_cast<uint8_t *>(output.ptr()); @@ -245,7 +245,7 @@ void NEUpsampleLayerKernel::upsample_f16_nchw(const arm_compute::Window &window) Iterator output(_output, window_out); const int offset_y_out = _output->info()->strides_in_bytes().y() / sizeof(float16_t); - execute_window_loop(window_out, [&](const Coordinates & id) + execute_window_loop(window_out, [&](const Coordinates &) { const float16x8_t data = vld1q_f16(reinterpret_cast<const float16_t *>(input.ptr())); const float16x8_t data_out1 = { vgetq_lane_f16(data, 0), vgetq_lane_f16(data, 0), vgetq_lane_f16(data, 1), vgetq_lane_f16(data, 1), @@ -278,7 +278,7 @@ void NEUpsampleLayerKernel::upsample_f16_nhwc(const arm_compute::Window &window) const int offset_y_out = _output->info()->strides_in_bytes().y() / sizeof(float16_t); const int offset_z_out = _output->info()->strides_in_bytes().z() / sizeof(float16_t); - execute_window_loop(window_out, [&](const Coordinates & id) + execute_window_loop(window_out, [&](const Coordinates &) { const float16x8_t data = vld1q_f16(reinterpret_cast<const float16_t *>(input.ptr())); auto out = reinterpret_cast<float16_t *>(output.ptr()); diff --git a/src/runtime/CL/functions/CLCropResize.cpp b/src/runtime/CL/functions/CLCropResize.cpp index 2cacef1bb1..b22809eb09 100644 --- a/src/runtime/CL/functions/CLCropResize.cpp +++ b/src/runtime/CL/functions/CLCropResize.cpp @@ -57,9 +57,9 @@ inline void run_crop(const ICLTensor *input, ICLTensor *output, uint32_t batch_i bool is_width_flipped = end[0] < start[0]; bool is_height_flipped = end[1] < start[1]; /** The number of rows out of bounds at the start and end of output. */ - int32_t rows_out_of_bounds[2]; + std::array<int32_t, 2> rows_out_of_bounds{ 0 }; /** The number of columns out of bounds at the start and end of output. */ - int32_t cols_out_of_bounds[2]; + std::array<int32_t, 2> cols_out_of_bounds{ 0 }; if(is_height_flipped) { rows_out_of_bounds[0] = start[1] >= static_cast<int32_t>(input->info()->dimension(2)) ? std::min(start[1] - input->info()->dimension(2) + 1, output->info()->dimension(2)) : 0; @@ -164,7 +164,7 @@ inline void run_crop(const ICLTensor *input, ICLTensor *output, uint32_t batch_i } // namespace CLCropResize::CLCropResize() - : _input(nullptr), _boxes(nullptr), _box_ind(nullptr), _output(nullptr), _num_boxes(0), _method(), _extrapolation_value(0), _scale(), _copy() + : _input(nullptr), _boxes(nullptr), _box_ind(nullptr), _output(nullptr), _num_boxes(0), _method(), _extrapolation_value(0), _scale(), _copy(), _crop_results(), _scaled_results() { } @@ -210,20 +210,19 @@ void CLCropResize::configure(const ICLTensor *input, ICLTensor *boxes, ICLTensor // - A scale function is used to resize the cropped image to the size specified by crop_size. // - A tensor is required to hold the final scaled image before it is copied into the 4D output // that will hold all final cropped and scaled 3D images using CLCopyKernel. - _crop_results = arm_compute::support::cpp14::make_unique<CLTensor[]>(_num_boxes); - _scale = arm_compute::support::cpp14::make_unique<CLScale[]>(_num_boxes); - _scaled_results = arm_compute::support::cpp14::make_unique<CLTensor[]>(_num_boxes); - _copy = arm_compute::support::cpp14::make_unique<CLCopyKernel[]>(_num_boxes); - for(unsigned int i = 0; i < _num_boxes; ++i) { + auto crop_tensor = support::cpp14::make_unique<CLTensor>(); TensorInfo crop_result_info(1, DataType::F32); crop_result_info.set_data_layout(DataLayout::NHWC); - _crop_results[i].allocator()->init(crop_result_info); + crop_tensor->allocator()->init(crop_result_info); + _crop_results.emplace_back(std::move(crop_tensor)); + auto scale_tensor = support::cpp14::make_unique<CLTensor>(); TensorInfo scaled_result_info(out_shape, 1, DataType::F32); scaled_result_info.set_data_layout(DataLayout::NHWC); - _scaled_results[i].allocator()->init(scaled_result_info); + scale_tensor->allocator()->init(scaled_result_info); + _scaled_results.emplace_back(std::move(scale_tensor)); } } @@ -240,32 +239,37 @@ void CLCropResize::run() // Size of the crop box in _boxes and thus the shape of _crop_results[i] // may not be known until run-time and so the kernels cannot be configured until then. uint32_t batch_index; - Coordinates start, end; - configure_crop(_input, _boxes, _box_ind, &_crop_results[i], i, start, end, batch_index); - _scale[i].configure(&_crop_results[i], &_scaled_results[i], _method, BorderMode::CONSTANT, PixelValue(_extrapolation_value), SamplingPolicy::TOP_LEFT); + Coordinates start{}; + Coordinates end{}; + configure_crop(_input, _boxes, _box_ind, _crop_results[i].get(), i, start, end, batch_index); + + auto scale_kernel = support::cpp14::make_unique<CLScale>(); + scale_kernel->configure(_crop_results[i].get(), _scaled_results[i].get(), _method, BorderMode::CONSTANT, PixelValue(_extrapolation_value), SamplingPolicy::TOP_LEFT); + _scale.emplace_back(std::move(scale_kernel)); Window win = calculate_max_window(*_output->info()); win.set(3, Window::Dimension(i, i + 1, 1)); - _copy[i].configure(&_scaled_results[i], _output, PaddingList(), &win); - _crop_results[i].allocator()->allocate(); - _scaled_results[i].allocator()->allocate(); + auto copy_kernel = support::cpp14::make_unique<CLCopyKernel>(); + copy_kernel->configure(_scaled_results[i].get(), _output, PaddingList(), &win); + _copy.emplace_back(std::move(copy_kernel)); + + _crop_results[i]->allocator()->allocate(); + _scaled_results[i]->allocator()->allocate(); - run_crop(_input, &_crop_results[i], batch_index, start, end, _extrapolation_value); + run_crop(_input, _crop_results[i].get(), batch_index, start, end, _extrapolation_value); } _boxes->unmap(CLScheduler::get().queue()); _box_ind->unmap(CLScheduler::get().queue()); CLScheduler::get().sync(); - for(unsigned int i = 0; i < _num_boxes; ++i) + for(auto &kernel : _scale) { - // Scale the cropped image - _scale[i].run(); + kernel->run(); } CLScheduler::get().sync(); - for(unsigned int i = 0; i < _num_boxes; ++i) + for(auto &kernel : _copy) { - // Copy scaled image into output. - CLScheduler::get().enqueue(_copy[i]); + CLScheduler::get().enqueue(*kernel, true); } CLScheduler::get().sync(); } diff --git a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp index add5e16498..7b9229c4ae 100644 --- a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp +++ b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp @@ -231,7 +231,8 @@ void CLFullyConnectedLayer::configure(const ICLTensor *input, const ICLTensor *w if(_is_quantized) { float multiplier = input->info()->quantization_info().scale * weights->info()->quantization_info().scale / output->info()->quantization_info().scale; - int output_multiplier, output_shift; + int output_multiplier; + int output_shift; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); _gemmlowp_output_stage.configure(&_gemmlowp_output, biases, output, output_multiplier, output_shift, output->info()->quantization_info().offset); _gemmlowp_output.allocator()->allocate(); diff --git a/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp b/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp index cd537087d3..049db1d461 100644 --- a/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp +++ b/src/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.cpp @@ -249,7 +249,8 @@ Status CLGEMMLowpMatrixMultiplyCore::validate(const ITensorInfo *a, const ITenso ARM_COMPUTE_RETURN_ON_ERROR(CLGEMMReshapeRHSMatrixKernel::validate(b, &tmp_b_info, rhs_info)); } - TensorInfo info_vector_sum_col, info_vector_sum_row; + TensorInfo info_vector_sum_col{}; + TensorInfo info_vector_sum_row{}; // Validate matrix B reduction kernel only if _a_offset is not equal to 0 if(a_offset != 0) diff --git a/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp b/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp index 5133756993..3bb69b1ffc 100644 --- a/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp +++ b/src/runtime/NEON/functions/NEDepthwiseConvolutionLayer.cpp @@ -112,7 +112,8 @@ void NEDepthwiseConvolutionLayer3x3::configure_generic(ITensor const QuantizationInfo output_quant_info = (output->info()->total_size() == 0) ? input->info()->quantization_info() : output->info()->quantization_info(); float multiplier = input->info()->quantization_info().scale * weights->info()->quantization_info().scale / output_quant_info.scale; - int output_multiplier, output_shift; + int output_multiplier; + int output_shift; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); _output_stage_kernel.configure(&_accumulator, biases, _is_nchw ? output : &_permuted_output, output_multiplier, output_shift, output_quant_info.offset); _accumulator.allocator()->allocate(); @@ -461,7 +462,8 @@ void NEDepthwiseConvolutionLayer::configure(ITensor *input, const ITensor *weigh const QuantizationInfo output_quant_info = output->info()->quantization_info(); float multiplier = input->info()->quantization_info().scale * weights->info()->quantization_info().scale / output_quant_info.scale; - int output_multiplier, output_shift; + int output_multiplier; + int output_shift; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); _output_stage_kernel.configure(&_output_reshaped, biases, output_to_use, output_multiplier, output_shift, output_quant_info.offset); _output_reshaped.allocator()->allocate(); diff --git a/src/runtime/NEON/functions/NEFullyConnectedLayer.cpp b/src/runtime/NEON/functions/NEFullyConnectedLayer.cpp index 273da278e8..e1a17db6d4 100644 --- a/src/runtime/NEON/functions/NEFullyConnectedLayer.cpp +++ b/src/runtime/NEON/functions/NEFullyConnectedLayer.cpp @@ -228,7 +228,8 @@ void NEFullyConnectedLayer::configure(const ITensor *input, const ITensor *weigh if(_is_quantized) { float multiplier = input->info()->quantization_info().scale * weights->info()->quantization_info().scale / output->info()->quantization_info().scale; - int output_multiplier, output_shift; + int output_multiplier; + int output_shift; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); _gemmlowp_output_stage.configure(&_gemmlowp_output, biases, output, output_multiplier, output_shift, output->info()->quantization_info().offset); _gemmlowp_output.allocator()->allocate(); diff --git a/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp b/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp index beac4ffe5f..a2c4e8a8b1 100644 --- a/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp +++ b/src/runtime/NEON/functions/NEGEMMConvolutionLayer.cpp @@ -118,7 +118,8 @@ void NEGEMMConvolutionLayer::configure_mm(const ITensor *input, const ITensor *w const QuantizationInfo output_quant_info = (output->info()->total_size() == 0) ? input_quantization_info : output->info()->quantization_info(); float multiplier = input_quantization_info.scale * weights->info()->quantization_info().scale / output_quant_info.scale; - int output_multiplier, output_shift; + int output_multiplier; + int output_shift; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); // Merge activation with output stage @@ -184,7 +185,8 @@ Status NEGEMMConvolutionLayer::validate_mm(const ITensorInfo *input, const ITens const QuantizationInfo output_quant_info = (output->total_size() == 0) ? input_quantization_info : output->quantization_info(); float multiplier = input_quantization_info.scale * weights->quantization_info().scale / output_quant_info.scale; - int output_multiplier, output_shift; + int output_multiplier; + int output_shift; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); // Merge activation with output stage @@ -412,7 +414,10 @@ Status NEGEMMConvolutionLayer::validate(const ITensorInfo *input, const ITensorI const unsigned int kernel_width = weights->dimension(idx_width); const unsigned int kernel_height = weights->dimension(idx_height); - TensorInfo im2col_reshaped_info, info_gemm, tmp_info, weights_reshaped_info; + TensorInfo im2col_reshaped_info{}; + TensorInfo info_gemm{}; + TensorInfo tmp_info{}; + TensorInfo weights_reshaped_info{}; const ITensorInfo *gemm_input_to_use = input; const ITensorInfo *gemm_output_to_use = output; const ITensorInfo *weights_to_use = weights; diff --git a/src/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.cpp b/src/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.cpp index 5ce7bc5f2c..54f49a6707 100644 --- a/src/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.cpp +++ b/src/runtime/NEON/functions/NEGEMMLowpMatrixMultiplyCore.cpp @@ -287,7 +287,8 @@ Status NEGEMMLowpMatrixMultiplyCore::validate(const ITensorInfo *a, const ITenso } } - TensorInfo info_vector_sum_col, info_vector_sum_row; + TensorInfo info_vector_sum_col{}; + TensorInfo info_vector_sum_row{}; // Validate matrix B reduction kernel only if _a_offset is not equal to 0 if(a_offset != 0) diff --git a/src/runtime/NEON/functions/NEWinogradConvolutionLayer.cpp b/src/runtime/NEON/functions/NEWinogradConvolutionLayer.cpp index 2332450c04..1513786ae5 100644 --- a/src/runtime/NEON/functions/NEWinogradConvolutionLayer.cpp +++ b/src/runtime/NEON/functions/NEWinogradConvolutionLayer.cpp @@ -162,7 +162,7 @@ inline Tensor4DShape internal_get_input_shape(const arm_compute::ITensor *input) const int in_channels = input->info()->dimension(get_data_layout_dimension_index(data_layout, DataLayoutDimension::CHANNEL)); const int in_batches = input->info()->dimension(3); - return Tensor4DShape({ in_batches, in_height, in_width, in_channels }); + return Tensor4DShape{ in_batches, in_height, in_width, in_channels }; } Status validate_arguments(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const PadStrideInfo &conv_info) @@ -234,7 +234,7 @@ bool check_support_fast_math(const Size2D &output_tile, const Size2D &kernel_siz } //namespace -NEWinogradConvolutionLayer::NEWinogradConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager) +NEWinogradConvolutionLayer::NEWinogradConvolutionLayer(const std::shared_ptr<IMemoryManager> &memory_manager) : _memory_group(memory_manager), _gemm_function(memory_manager), _transform_input_kernel(nullptr), _transform_output_kernel(nullptr), _transform_weights_kernel(nullptr), _activationlayer_function(), _permute_input(), _permute_weights(), _permute_output(), _input_transformed(), _output_transformed(), _input_workspace(), _output_workspace(), _kernel_storage(), _input_nhwc(), _output_nhwc(), _weights_hwio(), _input(), _weights(), _output(), _is_prepared(false), _is_activationlayer_enabled(false) @@ -428,7 +428,9 @@ void NEWinogradConvolutionLayer::configure(const ITensor *input, const ITensor * d_strides.set(2, 0); d_strides.set(3, data_type_size * output_matrix_stride); - TensorInfo a_info, b_info, d_info; + TensorInfo a_info{}; + TensorInfo b_info{}; + TensorInfo d_info{}; a_info.init(a_shape, 1, data_type, a_strides, 0, input_storage_size); b_info.init(b_shape, 1, data_type, b_strides, 0, kernel_storage_size); d_info.init(d_shape, 1, data_type, d_strides, 0, output_storage_size); |