/* * Copyright (c) 2017-2018 ARM Limited. * * SPDX-License-Identifier: MIT * * Permission is hereby granted, free of charge, to any person obtaining a copy * of this software and associated documentation files (the "Software"), to * deal in the Software without restriction, including without limitation the * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or * sell copies of the Software, and to permit persons to whom the Software is * furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice shall be included in all * copies or substantial portions of the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ #include "arm_compute/core/NEON/kernels/NEReductionOperationKernel.h" #include "arm_compute/core/Coordinates.h" #include "arm_compute/core/Helpers.h" #include "arm_compute/core/IAccessWindow.h" #include "arm_compute/core/ITensor.h" #include "arm_compute/core/NEON/INEKernel.h" #include "arm_compute/core/NEON/NEMath.h" #include "arm_compute/core/TensorInfo.h" #include "arm_compute/core/Validate.h" #include "arm_compute/core/NEON/wrapper/wrapper.h" #include namespace arm_compute { namespace { template class Reducer { public: static void reduceX(const Window &window, const ITensor *input, ITensor *output, F f) { // Set out window Window out_window(window); out_window.set(Window::DimX, Window::Dimension(0, 0, 0)); // Get first input and output slices Window in_slice = window.first_slice_window_1D(); Window out_slice = out_window.first_slice_window_1D(); do { Iterator in(input, in_slice); Iterator out(output, out_slice); f(in, out, in_slice, out_slice, *input->info()); } while(window.slide_window_slice_1D(in_slice) && out_window.slide_window_slice_1D(out_slice)); } static void reduceY(const Window &window, const ITensor *input, ITensor *output, F f) { // Set in window Window in_window(window); in_window.set(Window::DimY, Window::Dimension(0, 1, 1)); // Get first input and output slices Window in_slice = in_window.first_slice_window_2D(); Window out_slice = window.first_slice_window_2D(); do { Iterator in(input, in_slice); Iterator out(output, out_slice); f(in, out, in_slice, out_slice, *input->info(), 1); } while(in_window.slide_window_slice_2D(in_slice) && window.slide_window_slice_2D(out_slice)); } static void reduceZ(const Window &window, const ITensor *input, ITensor *output, F f) { // Set in window Window in_window(window); in_window.set(Window::DimZ, Window::Dimension(0, 1, 1)); // Get first input and output slices Window in_slice = in_window.first_slice_window_3D(); Window out_slice = window.first_slice_window_3D(); do { Iterator in(input, in_slice); Iterator out(output, out_slice); f(in, out, in_slice, out_slice, *input->info(), 2); } while(in_window.slide_window_slice_3D(in_slice) && window.slide_window_slice_3D(out_slice)); } static void reduceW(const Window &window, const ITensor *input, ITensor *output, F f) { // Set in/out window Window in_window(window); Window out_window(window); in_window.set(3, Window::Dimension(0, 1, 1)); out_window.set(3, Window::Dimension(0, 1, 1)); // Get first input and output slices Window in_slice = in_window.first_slice_window_4D(); Window out_slice = out_window.first_slice_window_4D(); do { Iterator in(input, in_slice); Iterator out(output, out_slice); f(in, out, in_slice, out_slice, *input->info(), 3); } while(in_window.slide_window_slice_4D(in_slice) && out_window.slide_window_slice_4D(out_slice)); } }; template struct RedOpX { /** NEON vector tag type. */ using ExactTagType = typename wrapper::traits::neon_vector::tag_type; inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info) { ARM_COMPUTE_UNUSED(out_slice); auto vec_sum_value = wrapper::vdup_n(static_cast(0.f), ExactTagType{}); execute_window_loop(in_slice, [&](const Coordinates & id) { const auto in_ptr = reinterpret_cast(input.ptr()); const auto vec_elements = wrapper::vloadq(in_ptr); if(op == ReductionOperation::SUM_SQUARE) { vec_sum_value = wrapper::vadd(wrapper::vmul(vec_elements, vec_elements), vec_sum_value); } else { vec_sum_value = wrapper::vadd(vec_elements, vec_sum_value); } }, input); auto carry_addition = wrapper::vpadd(wrapper::vgethigh(vec_sum_value), wrapper::vgetlow(vec_sum_value)); carry_addition = wrapper::vpadd(carry_addition, carry_addition); auto res = wrapper::vgetlane(carry_addition, 0); if(op == ReductionOperation::MEAN_SUM) { res /= in_info.dimension(0); } *(reinterpret_cast(output.ptr())) = res; } }; template struct RedOpX_qasymm8 { inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info) { ARM_COMPUTE_UNUSED(out_slice); auto vec_sum_value1 = vdupq_n_u32(static_cast(0.f)); auto vec_sum_value2 = vdupq_n_u32(static_cast(0.f)); auto vec_sum_value3 = vdupq_n_u32(static_cast(0.f)); auto vec_sum_value4 = vdupq_n_u32(static_cast(0.f)); execute_window_loop(in_slice, [&](const Coordinates & id) { const auto vec_elements = wrapper::vloadq(input.ptr()); const auto temp16x8t_1 = wrapper::vmovl(wrapper::vgetlow(vec_elements)); const auto temp16x8t_2 = wrapper::vmovl(wrapper::vgethigh(vec_elements)); const auto temp32x4t_1 = wrapper::vmovl(wrapper::vgetlow(temp16x8t_1)); const auto temp32x4t_2 = wrapper::vmovl(wrapper::vgethigh(temp16x8t_1)); const auto temp32x4t_3 = wrapper::vmovl(wrapper::vgetlow(temp16x8t_2)); const auto temp32x4t_4 = wrapper::vmovl(wrapper::vgethigh(temp16x8t_2)); vec_sum_value1 = wrapper::vadd(temp32x4t_1, vec_sum_value1); vec_sum_value2 = wrapper::vadd(temp32x4t_2, vec_sum_value2); vec_sum_value3 = wrapper::vadd(temp32x4t_3, vec_sum_value3); vec_sum_value4 = wrapper::vadd(temp32x4t_4, vec_sum_value4); }, input); auto carry_addition = wrapper::vadd(vec_sum_value1, vec_sum_value2); carry_addition = wrapper::vadd(carry_addition, vec_sum_value3); carry_addition = wrapper::vadd(carry_addition, vec_sum_value4); auto carry_paddition = wrapper::vpadd(wrapper::vgethigh(carry_addition), wrapper::vgetlow(carry_addition)); carry_paddition = wrapper::vpadd(carry_paddition, carry_paddition); auto res = wrapper::vgetlane(carry_paddition, 0); if(op == ReductionOperation::MEAN_SUM) { res /= in_info.dimension(0); } *(output.ptr()) = static_cast(res); } }; template struct RedOpYZW { /** NEON vector tag type. */ using ExactTagType = typename wrapper::traits::neon_vector::tag_type; inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, int axis) { ARM_COMPUTE_UNUSED(out_slice); execute_window_loop(in_slice, [&](const Coordinates & id) { auto vec_sum_value = wrapper::vdup_n(static_cast(0.f), ExactTagType{}); for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) { T *in_ptr; switch(axis) { case 1: in_ptr = reinterpret_cast(input.ptr() + in_info.offset_element_in_bytes(Coordinates(0, dim))); break; case 2: in_ptr = reinterpret_cast(input.ptr() + in_info.offset_element_in_bytes(Coordinates(0, 0, dim))); break; case 3: in_ptr = reinterpret_cast(input.ptr() + in_info.offset_element_in_bytes(Coordinates(0, 0, 0, dim))); break; default: ARM_COMPUTE_ERROR("Not supported"); } const auto vec_elements = wrapper::vloadq(in_ptr); if(op == ReductionOperation::SUM_SQUARE) { vec_sum_value = wrapper::vadd(wrapper::vmul(vec_elements, vec_elements), vec_sum_value); } else { vec_sum_value = wrapper::vadd(vec_elements, vec_sum_value); } } if(op == ReductionOperation::MEAN_SUM) { auto vec_width_inv = wrapper::vinv(wrapper::vdup_n(static_cast(in_info.dimension(axis)), ExactTagType{})); vec_sum_value = wrapper::vmul(vec_sum_value, vec_width_inv); } wrapper::vstore(reinterpret_cast(output.ptr()), vec_sum_value); }, input, output); } }; template struct RedOpYZW_qasymm8 { inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, int axis) { ARM_COMPUTE_UNUSED(out_slice); execute_window_loop(in_slice, [&](const Coordinates & id) { auto vec_sum_value1 = vdupq_n_u32(static_cast(0.f)); auto vec_sum_value2 = vdupq_n_u32(static_cast(0.f)); auto vec_sum_value3 = vdupq_n_u32(static_cast(0.f)); auto vec_sum_value4 = vdupq_n_u32(static_cast(0.f)); for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) { uint8_t *in_ptr; switch(axis) { case 1: in_ptr = input.ptr() + in_info.offset_element_in_bytes(Coordinates(0, dim)); break; case 2: in_ptr = input.ptr() + in_info.offset_element_in_bytes(Coordinates(0, 0, dim)); break; case 3: in_ptr = input.ptr() + in_info.offset_element_in_bytes(Coordinates(0, 0, 0, dim)); break; default: ARM_COMPUTE_ERROR("Not supported"); } const auto vec_elements = wrapper::vloadq(in_ptr); const auto temp16x8t_1 = wrapper::vmovl(wrapper::vgetlow(vec_elements)); const auto temp16x8t_2 = wrapper::vmovl(wrapper::vgethigh(vec_elements)); const auto temp32x4t_1 = wrapper::vmovl(wrapper::vgetlow(temp16x8t_1)); const auto temp32x4t_2 = wrapper::vmovl(wrapper::vgethigh(temp16x8t_1)); const auto temp32x4t_3 = wrapper::vmovl(wrapper::vgetlow(temp16x8t_2)); const auto temp32x4t_4 = wrapper::vmovl(wrapper::vgethigh(temp16x8t_2)); vec_sum_value1 = wrapper::vadd(temp32x4t_1, vec_sum_value1); vec_sum_value2 = wrapper::vadd(temp32x4t_2, vec_sum_value2); vec_sum_value3 = wrapper::vadd(temp32x4t_3, vec_sum_value3); vec_sum_value4 = wrapper::vadd(temp32x4t_4, vec_sum_value4); } if(op == ReductionOperation::MEAN_SUM) { const auto vec_width_inv = wrapper::vinv(vdupq_n_f32(in_info.dimension(axis))); const auto vec_sum_value1_f = wrapper::vmul(vcvtq_f32_u32(vec_sum_value1), vec_width_inv); const auto vec_sum_value2_f = wrapper::vmul(vcvtq_f32_u32(vec_sum_value2), vec_width_inv); const auto vec_sum_value3_f = wrapper::vmul(vcvtq_f32_u32(vec_sum_value3), vec_width_inv); const auto vec_sum_value4_f = wrapper::vmul(vcvtq_f32_u32(vec_sum_value4), vec_width_inv); vec_sum_value1 = vcvtq_u32_f32(vec_sum_value1_f); vec_sum_value2 = vcvtq_u32_f32(vec_sum_value2_f); vec_sum_value3 = vcvtq_u32_f32(vec_sum_value3_f); vec_sum_value4 = vcvtq_u32_f32(vec_sum_value4_f); } const auto temp16x8t_1 = vcombine_u16(wrapper::vqmovn(vec_sum_value1), wrapper::vqmovn(vec_sum_value2)); const auto temp16x8t_2 = vcombine_u16(wrapper::vqmovn(vec_sum_value3), wrapper::vqmovn(vec_sum_value4)); auto res = vcombine_u8(wrapper::vqmovn(temp16x8t_1), wrapper::vqmovn(temp16x8t_2)); wrapper::vstore(output.ptr(), res); }, input, output); } }; void reduce_sumsq(const Window &window, const ITensor *input, ITensor *output, unsigned int axis) { switch(axis) { case 0: switch(input->info()->data_type()) { #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceX(window, input, output, RedOpX()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceX(window, input, output, RedOpX()); case DataType::QASYMM8: default: ARM_COMPUTE_ERROR("Not supported"); } case 1: switch(input->info()->data_type()) { #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceY(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceY(window, input, output, RedOpYZW()); case DataType::QASYMM8: default: ARM_COMPUTE_ERROR("Not supported"); } case 2: switch(input->info()->data_type()) { #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceZ(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceZ(window, input, output, RedOpYZW()); case DataType::QASYMM8: default: ARM_COMPUTE_ERROR("Not supported"); } case 3: switch(input->info()->data_type()) { #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceW(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceW(window, input, output, RedOpYZW()); case DataType::QASYMM8: default: ARM_COMPUTE_ERROR("Not supported"); } default: ARM_COMPUTE_ERROR("Unsupported reduction axis"); } } void reduce_sum(const Window &window, const ITensor *input, ITensor *output, unsigned int axis) { switch(axis) { case 0: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceX(window, input, output, RedOpX_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceX(window, input, output, RedOpX()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceX(window, input, output, RedOpX()); default: ARM_COMPUTE_ERROR("Not supported"); } case 1: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceY(window, input, output, RedOpYZW_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceY(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceY(window, input, output, RedOpYZW()); default: ARM_COMPUTE_ERROR("Not supported"); } case 2: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceZ(window, input, output, RedOpYZW_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceZ(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceZ(window, input, output, RedOpYZW()); default: ARM_COMPUTE_ERROR("Not supported"); } case 3: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceW(window, input, output, RedOpYZW_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceW(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceW(window, input, output, RedOpYZW()); default: ARM_COMPUTE_ERROR("Not supported"); } default: ARM_COMPUTE_ERROR("Unsupported reduction axis"); } } void reduce_mean_sum(const Window &window, const ITensor *input, ITensor *output, unsigned int axis) { switch(axis) { case 0: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceX(window, input, output, RedOpX_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceX(window, input, output, RedOpX()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceX(window, input, output, RedOpX()); default: ARM_COMPUTE_ERROR("Not supported"); } case 1: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceY(window, input, output, RedOpYZW_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceY(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceY(window, input, output, RedOpYZW()); default: ARM_COMPUTE_ERROR("Not supported"); } case 2: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceZ(window, input, output, RedOpYZW_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceZ(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceZ(window, input, output, RedOpYZW()); default: ARM_COMPUTE_ERROR("Not supported"); } case 3: switch(input->info()->data_type()) { case DataType::QASYMM8: return Reducer>::reduceW(window, input, output, RedOpYZW_qasymm8()); #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F16: return Reducer>::reduceW(window, input, output, RedOpYZW()); #endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC case DataType::F32: return Reducer>::reduceW(window, input, output, RedOpYZW()); default: ARM_COMPUTE_ERROR("Not supported"); } default: ARM_COMPUTE_ERROR("Unsupported reduction axis"); } } TensorShape calculate_output_shape(const TensorShape &input_shape, unsigned int axis) { TensorShape output_shape{ input_shape }; output_shape.set(axis, 1); return output_shape; } Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, unsigned int axis, ReductionOperation op) { ARM_COMPUTE_UNUSED(op); ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input, output); ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QASYMM8, DataType::F16, DataType::F32); ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis >= TensorShape::num_max_dimensions, "Reduction axis greater than max number of dimensions"); ARM_COMPUTE_RETURN_ERROR_ON_MSG(axis > 3, "Unsupported reduction axis"); if(output->total_size() != 0) { ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input, output); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_LAYOUT(input, output); const TensorShape output_shape = calculate_output_shape(input->tensor_shape(), axis); const TensorInfo tensor_info_reshaped = input->clone()->set_tensor_shape(output_shape); ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(output, &tensor_info_reshaped); } return Status{}; } std::tuple validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, unsigned int axis) { // Calculate output shape and set if empty const TensorShape output_shape = calculate_output_shape(input->tensor_shape(), axis); // Output auto initialization if not yet initialized auto_init_if_empty(*output, output_shape, 1, input->data_type()); unsigned int num_elems_processed_per_iteration = 16 / data_size_from_type(input->data_type()); // Configure kernel window Window win = calculate_max_window(*input, Steps(num_elems_processed_per_iteration)); AccessWindowHorizontal input_access(input, 0, num_elems_processed_per_iteration); AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration); bool window_changed = update_window_and_padding(win, input_access, output_access); output_access.set_valid_region(win, ValidRegion(Coordinates(), output->tensor_shape())); Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{}; return std::make_tuple(err, win); } } // namespace NEReductionOperationKernel::NEReductionOperationKernel() : _input(nullptr), _output(nullptr), _reduction_axis(0), _op(ReductionOperation::SUM_SQUARE), _border_size() { } BorderSize NEReductionOperationKernel::border_size() const { return _border_size; } void NEReductionOperationKernel::configure(const ITensor *input, ITensor *output, unsigned int axis, ReductionOperation op) { ARM_COMPUTE_ERROR_ON_NULLPTR(input, output); ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), output->info(), axis, op)); unsigned int num_elems_processed_per_iteration = 16 / data_size_from_type(input->info()->data_type()); _input = input; _output = output; _border_size = (axis == 0) ? BorderSize(0, num_elems_processed_per_iteration - (input->info()->dimension(0) % num_elems_processed_per_iteration), 0, 0) : BorderSize(); _op = op; _reduction_axis = axis; // Configure kernel window auto win_config = validate_and_configure_window(_input->info(), _output->info(), axis); ARM_COMPUTE_ERROR_THROW_ON(std::get<0>(win_config)); INEKernel::configure(std::get<1>(win_config)); } Status NEReductionOperationKernel::validate(const ITensorInfo *input, const ITensorInfo *output, unsigned int axis, ReductionOperation op) { ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, output, axis, op)); ARM_COMPUTE_RETURN_ON_ERROR(std::get<0>(validate_and_configure_window(input->clone().get(), output->clone().get(), axis))); return Status{}; } void NEReductionOperationKernel::run(const Window &window, const ThreadInfo &info) { ARM_COMPUTE_UNUSED(info); ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window); switch(_op) { case ReductionOperation::SUM_SQUARE: reduce_sumsq(window, _input, _output, _reduction_axis); break; case ReductionOperation::MEAN_SUM: reduce_mean_sum(window, _input, _output, _reduction_axis); break; case ReductionOperation::SUM: reduce_sum(window, _input, _output, _reduction_axis); break; default: ARM_COMPUTE_ERROR("Unsupported reduction operation."); } } } // namespace arm_compute