diff options
| author | Sheri Zhang <sheri.zhang@arm.com> | 2020-09-23 11:22:50 +0100 |
|---|---|---|
| committer | Sheri Zhang <sheri.zhang@arm.com> | 2020-10-06 09:19:36 +0000 |
| commit | 4d91dc68adf8a4cc07285fe781469231230df3b9 (patch) | |
| tree | 4b8b53ab30f86921031fd2b6b9ff35dfdecc222b | |
| parent | 47ae441b320c0a9f79f8e6036a0b12a1bf68f9ca (diff) | |
| download | ComputeLibrary-4d91dc68adf8a4cc07285fe781469231230df3b9.tar.gz | |
COMPMID-3181: Remove padding from NEReductionOperationKernel
COMPMID-3803: Remove padding from NEComplexPixelWiseMultiplicationKernel
Signed-off-by: Sheri Zhang <sheri.zhang@arm.com>
Change-Id: I309fc4ab62bacbca9203d2680a9d6d52f76f70e6
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/4078
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Manuel Bottini <manuel.bottini@arm.com>
Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
6 files changed, 1065 insertions, 655 deletions
diff --git a/arm_compute/core/NEON/kernels/NEReductionOperationKernel.h b/arm_compute/core/NEON/kernels/NEReductionOperationKernel.h index 180697f364..b913ea4e90 100644 --- a/arm_compute/core/NEON/kernels/NEReductionOperationKernel.h +++ b/arm_compute/core/NEON/kernels/NEReductionOperationKernel.h @@ -81,14 +81,12 @@ public: // Inherited methods overridden: void run(const Window &window, const ThreadInfo &info) override; - BorderSize border_size() const override; private: const ITensor *_input; ITensor *_output; unsigned int _reduction_axis; ReductionOperation _op; - BorderSize _border_size; }; } // namespace arm_compute #endif /*ARM_COMPUTE_NEREDUCTIONOPERATIONKERNEL_H */ diff --git a/arm_compute/runtime/NEON/functions/NEReductionOperation.h b/arm_compute/runtime/NEON/functions/NEReductionOperation.h index ab6928b241..aafccb00e3 100644 --- a/arm_compute/runtime/NEON/functions/NEReductionOperation.h +++ b/arm_compute/runtime/NEON/functions/NEReductionOperation.h @@ -26,7 +26,6 @@ #include "arm_compute/runtime/IFunction.h" -#include "arm_compute/core/NEON/kernels/NEFillBorderKernel.h" #include "arm_compute/core/NEON/kernels/NEReductionOperationKernel.h" #include "arm_compute/runtime/NEON/functions/NEReshapeLayer.h" #include "arm_compute/runtime/Tensor.h" @@ -37,7 +36,7 @@ class ITensor; /** Basic function to simulate a reduction operation. This function calls the following NEON kernels: * - * -# @ref NEFillBorderKernel + * -# @ref NEReshapeLayer * -# @ref NEReductionOperationKernel * */ @@ -74,7 +73,6 @@ public: private: MemoryGroup _memory_group; NEReductionOperationKernel _reduction_kernel; - NEFillBorderKernel _fill_border_kernel; NEReshapeLayer _reshape; Tensor _output_internal; size_t _window_split; diff --git a/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp b/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp index 84683ea69f..4466c24604 100644 --- a/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp +++ b/src/core/NEON/kernels/NEPixelWiseMultiplicationKernel.cpp @@ -1018,33 +1018,115 @@ void mul_F32_F32_F32(const ITensor *in1, const ITensor *in2, ITensor *out, const } } -void c_mul_F32_F32_F32_n(const void *__restrict input1_ptr, const void *__restrict input2_ptr, void *__restrict output_ptr) +void c_mul_F32_F32_F32_n(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window) { - const auto input1 = static_cast<const float *__restrict>(input1_ptr); - const auto input2 = static_cast<const float *__restrict>(input2_ptr); - const auto output = static_cast<float *__restrict>(output_ptr); + // Create input windows + Window input1_win = window.broadcast_if_dimension_le_one(in1->info()->tensor_shape()); + Window input2_win = window.broadcast_if_dimension_le_one(in2->info()->tensor_shape()); + + // Clear X Dimension on execution window as we handle manually + Window win = window; + win.set(Window::DimX, Window::Dimension(0, 1, 1)); + + constexpr int window_step_x = 8 / sizeof(float); + const auto window_start_x = static_cast<int>(window.x().start()); + 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); + + if(is_broadcast_across_x) + { + const bool is_broadcast_input_2 = input2_win.x().step() == 0; + Window broadcast_win = is_broadcast_input_2 ? input2_win : input1_win; + Window non_broadcast_win = !is_broadcast_input_2 ? input2_win : input1_win; + const ITensor *broadcast_tensor = is_broadcast_input_2 ? in2 : in1; + const ITensor *non_broadcast_tensor = !is_broadcast_input_2 ? in2 : in1; - const float32x4_t a = wrapper::vloadq(input1); - float32x4_t b = wrapper::vloadq(input2); + // Clear X Dimension on execution window as we handle manually + non_broadcast_win.set(Window::DimX, Window::Dimension(0, 1, 1)); - using ExactTagType = typename wrapper::traits::neon_vector<float, 2>::tag_type; + Iterator broadcast_input(broadcast_tensor, broadcast_win); + Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); + Iterator output(out, win); - const float32x4_t mask = { -1.0f, 1.0f, -1.0f, 1.0f }; - const float32x2_t tmp00 = wrapper::vdup_n(wrapper::vgetlane(a, 0), ExactTagType{}); - const float32x2_t tmp01 = wrapper::vdup_n(wrapper::vgetlane(a, 1), ExactTagType{}); - const float32x2_t tmp10 = wrapper::vdup_n(wrapper::vgetlane(a, 2), ExactTagType{}); - const float32x2_t tmp11 = wrapper::vdup_n(wrapper::vgetlane(a, 3), ExactTagType{}); + execute_window_loop(win, [&](const Coordinates &) + { + const auto non_broadcast_input_ptr = reinterpret_cast<const float *>(non_broadcast_input.ptr()); + const auto output_ptr = reinterpret_cast<float *>(output.ptr()); - const float32x4_t tmp0 = wrapper::vcombine(tmp00, tmp10); - const float32x4_t tmp1 = wrapper::vcombine(tmp01, tmp11); + const float broadcast_value = *reinterpret_cast<const float *>(broadcast_input.ptr()); - float32x4_t res = wrapper::vmul(tmp0, b); + int x = window_start_x; + // Compute left-over elements + for(; x < window_end_x; ++x) + { + const auto broadcast_value0 = *(non_broadcast_input_ptr + 2 * x); + const auto broadcast_value1 = *(non_broadcast_input_ptr + 2 * x + 1); + auto res1 = broadcast_value * (broadcast_value0 - broadcast_value1); + auto res2 = broadcast_value * (broadcast_value1 + broadcast_value0); + *(output_ptr + 2 * x) = res1; + *(output_ptr + 2 * x + 1) = res2; + } + }, + broadcast_input, non_broadcast_input, output); + } + else + { + // Clear X Dimension on execution window as we handle manually + input1_win.set(Window::DimX, Window::Dimension(0, 1, 1)); + input2_win.set(Window::DimX, Window::Dimension(0, 1, 1)); - b = wrapper::vrev64(b); - b = wrapper::vmul(b, mask); + Iterator input1(in1, input1_win); + Iterator input2(in2, input2_win); + Iterator output(out, win); - res = wrapper::vmla(res, tmp1, b); - wrapper::vstore(output, res); + execute_window_loop(win, [&](const Coordinates &) + { + const auto input1_ptr = reinterpret_cast<const float *>(input1.ptr()); + const auto input2_ptr = reinterpret_cast<const float *>(input2.ptr()); + const auto output_ptr = reinterpret_cast<float *>(output.ptr()); + + using ExactTagType = typename wrapper::traits::neon_vector<float, 2>::tag_type; + + // Compute window_step_x elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + const float32x4_t a = wrapper::vloadq(input1_ptr + 2 * x); + float32x4_t b = wrapper::vloadq(input2_ptr + 2 * x); + + const float32x4_t mask = { -1.0f, 1.0f, -1.0f, 1.0f }; + const float32x2_t tmp00 = wrapper::vdup_n(wrapper::vgetlane(a, 0), ExactTagType{}); + const float32x2_t tmp01 = wrapper::vdup_n(wrapper::vgetlane(a, 1), ExactTagType{}); + const float32x2_t tmp10 = wrapper::vdup_n(wrapper::vgetlane(a, 2), ExactTagType{}); + const float32x2_t tmp11 = wrapper::vdup_n(wrapper::vgetlane(a, 3), ExactTagType{}); + + const float32x4_t tmp0 = wrapper::vcombine(tmp00, tmp10); + const float32x4_t tmp1 = wrapper::vcombine(tmp01, tmp11); + + float32x4_t res = wrapper::vmul(tmp0, b); + + b = wrapper::vrev64(b); + b = wrapper::vmul(b, mask); + + res = wrapper::vmla(res, tmp1, b); + wrapper::vstore(output_ptr + 2 * x, res); + } + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + const auto a0 = *(input1_ptr + 2 * x); + const auto a1 = *(input1_ptr + 2 * x + 1); + const auto b0 = *(input2_ptr + 2 * x); + const auto b1 = *(input2_ptr + 2 * x + 1); + auto res1 = a0 * b0 - a1 * b1; + auto res2 = a0 * b1 + a1 * b0; + *(output_ptr + 2 * x) = res1; + *(output_ptr + 2 * x + 1) = res2; + } + }, + input1, input2, output); + } } #ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC @@ -1507,8 +1589,6 @@ void NEPixelWiseMultiplicationKernel::run_op(ITensorPack &tensors, const Window } namespace { -constexpr unsigned int num_elems_processed_per_iteration_complex = 2; - Status validate_arguments_complex(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output) { ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input1, 2, DataType::F32); @@ -1527,9 +1607,13 @@ Status validate_arguments_complex(const ITensorInfo *input1, const ITensorInfo * return Status{}; } +} // namespace -std::pair<Status, Window> validate_and_configure_window_complex(ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output) +void NEComplexPixelWiseMultiplicationKernel::configure(ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output) { + ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output); + ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_complex(input1, input2, output)); + const std::pair<TensorShape, ValidRegion> broadcast_pair = ITensorInfo::broadcast_shape_and_valid_region(*input1, *input2); const TensorShape &out_shape = broadcast_pair.first; const ValidRegion &valid_region = broadcast_pair.second; @@ -1538,43 +1622,19 @@ std::pair<Status, Window> validate_and_configure_window_complex(ITensorInfo *inp const TensorInfo out_info(out_shape, input1->num_channels(), input1->data_type()); auto_init_if_empty(*output, out_info); - Window win = calculate_max_window(valid_region, Steps(num_elems_processed_per_iteration_complex)); - Window win_input1 = win.broadcast_if_dimension_le_one(*input1); - Window win_input2 = win.broadcast_if_dimension_le_one(*input2); - - AccessWindowHorizontal input1_access(input1, 0, num_elems_processed_per_iteration_complex); - AccessWindowHorizontal input2_access(input2, 0, num_elems_processed_per_iteration_complex); - AccessWindowHorizontal output_access(output, 0, num_elems_processed_per_iteration_complex); - - bool window_changed = update_window_and_padding(win_input1, input1_access) - || update_window_and_padding(win_input2, input2_access) - || update_window_and_padding(win, output_access); - - output_access.set_valid_region(win, valid_region); - - Status err = (window_changed) ? ARM_COMPUTE_CREATE_ERROR(ErrorCode::RUNTIME_ERROR, "Insufficient Padding!") : Status{}; - return std::make_pair(err, win); -} -} // namespace - -void NEComplexPixelWiseMultiplicationKernel::configure(ITensorInfo *input1, ITensorInfo *input2, ITensorInfo *output) -{ - ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output); - ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_complex(input1, input2, output)); - // Configure kernel window - auto win_config = validate_and_configure_window_complex(input1, input2, output); - ARM_COMPUTE_ERROR_THROW_ON(win_config.first); + Coordinates coord; + coord.set_num_dimensions(output->num_dimensions()); + output->set_valid_region(valid_region); + Window win = calculate_max_window(valid_region, Steps()); - // Create kernel - INEKernel::configure(win_config.second); + INEKernel::configure(win); } Status NEComplexPixelWiseMultiplicationKernel::validate(const ITensorInfo *input1, const ITensorInfo *input2, const ITensorInfo *output) { ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, output); ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_complex(input1, input2, output)); - ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window_complex(input1->clone().get(), input2->clone().get(), output->clone().get()).first); return Status{}; } @@ -1589,14 +1649,6 @@ void NEComplexPixelWiseMultiplicationKernel::run_op(ITensorPack &tensors, const auto input2 = tensors.get_const_tensor(TensorType::ACL_SRC_1); auto output = tensors.get_tensor(TensorType::ACL_DST); - Iterator input1_it(input1, window.broadcast_if_dimension_le_one(input1->info()->tensor_shape())); - Iterator input2_it(input2, window.broadcast_if_dimension_le_one(input2->info()->tensor_shape())); - Iterator output_it(output, window); - - execute_window_loop(window, [&](const Coordinates &) - { - c_mul_F32_F32_F32_n(input1_it.ptr(), input2_it.ptr(), output_it.ptr()); - }, - input1_it, input2_it, output_it); + c_mul_F32_F32_F32_n(input1, input2, output, window); } } // namespace arm_compute diff --git a/src/core/NEON/kernels/NEReductionOperationKernel.cpp b/src/core/NEON/kernels/NEReductionOperationKernel.cpp index 5a52216eac..1691f6850c 100644 --- a/src/core/NEON/kernels/NEReductionOperationKernel.cpp +++ b/src/core/NEON/kernels/NEReductionOperationKernel.cpp @@ -45,17 +45,17 @@ namespace { // Helper function that calls vqmovun/vqmvn, vcombine and vstore, allows templating of RedOpYZW_quantized template <typename T> -void combine_and_store(int16x8_t t1, int16x8_t t2, Iterator &output) +void combine_and_store(int16x8_t t1, int16x8_t t2, Iterator &output, int offset = 0) { if(std::is_same<T, uint8_t>::value) { auto res = wrapper::vcombine(wrapper::vqmovun(t1), wrapper::vqmovun(t2)); - wrapper::vstore(output.ptr(), res); + wrapper::vstore(output.ptr() + offset, res); } else { auto res = wrapper::vcombine(wrapper::vqmovn(t1), wrapper::vqmovn(t2)); - wrapper::vstore(reinterpret_cast<int8_t *>(output.ptr()), res); + wrapper::vstore(reinterpret_cast<int8_t *>(output.ptr() + offset), res); } } @@ -342,20 +342,9 @@ public: { // Set out window Window out_window(window); - out_window.set(Window::DimX, Window::Dimension(0, 0, 0)); + out_window.set(Window::DimX, Window::Dimension(0, 1, 1)); - // 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(), op); - } - while(window.slide_window_slice_1D(in_slice) && out_window.slide_window_slice_1D(out_slice)); + f(window, out_window, input, output, op); } static void reduceY(const Window &window, const ITensor *input, ITensor *output, F f, const ReductionOperation op) { @@ -366,18 +355,7 @@ public: in_window.set(Window::DimY, Window::Dimension(0, 1, 1)); out_window.set(Window::DimY, Window::Dimension(0, output->info()->dimension(1), output->info()->dimension(1))); - // Get first input and output slices - Window in_slice = in_window.first_slice_window_2D(); - Window out_slice = out_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, op); - } - while(in_window.slide_window_slice_2D(in_slice) && out_window.slide_window_slice_2D(out_slice)); + f(in_window, out_window, input, output, 1, op); } static void reduceZ(const Window &window, const ITensor *input, ITensor *output, F f, const ReductionOperation op) { @@ -388,18 +366,7 @@ public: in_window.set(Window::DimZ, Window::Dimension(0, 1, 1)); out_window.set(Window::DimZ, Window::Dimension(0, output->info()->dimension(2), output->info()->dimension(2))); - // Get first input and output slices - Window in_slice = in_window.first_slice_window_3D(); - Window out_slice = out_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, op); - } - while(in_window.slide_window_slice_3D(in_slice) && out_window.slide_window_slice_3D(out_slice)); + f(in_window, out_window, input, output, 2, op); } static void reduceW(const Window &window, const ITensor *input, ITensor *output, F f, const ReductionOperation op) { @@ -410,18 +377,7 @@ public: 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, op); - } - while(in_window.slide_window_slice_4D(in_slice) && out_window.slide_window_slice_4D(out_slice)); + f(in_window, out_window, input, output, 3, op); } }; @@ -431,309 +387,463 @@ struct RedOpX /** NEON vector tag type. */ using ExactTagType = typename wrapper::traits::neon_vector<T, S>::tag_type; - inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, const ReductionOperation op) + inline void operator()(const Window &in_window, Window &out_window, const ITensor *in, ITensor *out, const ReductionOperation op) { - ARM_COMPUTE_UNUSED(out_slice); - auto init_res_value = static_cast<T>(0.f); - switch(op) + const TensorInfo in_info = *(in->info()); + + Iterator input(in, in_window); + Iterator output(out, out_window); + const int window_step_x = 16 / sizeof(T); + const auto window_start_x = static_cast<int>(in_window.x().start()); + const auto window_end_x = static_cast<int>(in_window.x().end()); + + execute_window_loop(in_window, [&](const Coordinates &) { - case ReductionOperation::ARG_IDX_MAX: - case ReductionOperation::ARG_IDX_MIN: - case ReductionOperation::MIN: - case ReductionOperation::MAX: + const auto input_ptr = reinterpret_cast<const T *>(input.ptr()); + + auto init_res_value = static_cast<T>(0.f); + switch(op) { - init_res_value = *reinterpret_cast<T *>(input.ptr()); - break; + case ReductionOperation::ARG_IDX_MAX: + case ReductionOperation::ARG_IDX_MIN: + case ReductionOperation::MIN: + case ReductionOperation::MAX: + { + init_res_value = static_cast<T>(*input_ptr); + break; + } + case ReductionOperation::PROD: + { + init_res_value = static_cast<T>(1.f); + break; + } + default: + break; } - case ReductionOperation::PROD: + auto vec_res_value = wrapper::vdup_n(init_res_value, ExactTagType{}); + uint32x4x4_t vec_res_idx{ { 0 } }; + + // Compute window_step_x elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) { - init_res_value = static_cast<T>(1.f); - break; + const auto vec_elements = wrapper::vloadq(input_ptr + x); + switch(op) + { + case ReductionOperation::SUM_SQUARE: + vec_res_value = wrapper::vadd(wrapper::vmul(vec_elements, vec_elements), vec_res_value); + break; + case ReductionOperation::MEAN_SUM: + case ReductionOperation::SUM: + vec_res_value = wrapper::vadd(vec_elements, vec_res_value); + break; + case ReductionOperation::PROD: + vec_res_value = wrapper::vmul(vec_elements, vec_res_value); + break; + case ReductionOperation::ARG_IDX_MIN: + { + auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + vec_res_idx = calculate_index<decltype(vec_res_value)>(x, temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::ARG_IDX_MAX: + { + auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + vec_res_idx = calculate_index<decltype(vec_res_value)>(x, temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::MIN: + { + vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + break; + } + case ReductionOperation::MAX: + { + vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + break; + } + default: + ARM_COMPUTE_ERROR("Not supported"); + } } - default: - break; - } - auto vec_res_value = wrapper::vdup_n(init_res_value, ExactTagType{}); - uint32x4x4_t vec_res_idx{ { 0 } }; - - execute_window_loop(in_slice, [&](const Coordinates & id) - { - const auto in_ptr = reinterpret_cast<const T *>(input.ptr()); - const auto vec_elements = wrapper::vloadq(in_ptr); switch(op) { - case ReductionOperation::SUM_SQUARE: - vec_res_value = wrapper::vadd(wrapper::vmul(vec_elements, vec_elements), vec_res_value); - break; - case ReductionOperation::MEAN_SUM: case ReductionOperation::SUM: - vec_res_value = wrapper::vadd(vec_elements, vec_res_value); + case ReductionOperation::MEAN_SUM: + case ReductionOperation::SUM_SQUARE: + { + auto carry_res = wrapper::vpadd(wrapper::vgethigh(vec_res_value), wrapper::vgetlow(vec_res_value)); + for(int i = 0; i < S / 4; ++i) + { + carry_res = wrapper::vpadd(carry_res, carry_res); + } + auto res = wrapper::vgetlane(carry_res, 0); + + if(op == ReductionOperation::SUM_SQUARE) + { + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res += (*(input_ptr + x)) * (*(input_ptr + x)); + } + } + else + { + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res += *(input_ptr + x); + } + } + + if(op == ReductionOperation::MEAN_SUM) + { + res /= in_info.dimension(0); + } + + *(reinterpret_cast<T *>(output.ptr())) = res; break; + } case ReductionOperation::PROD: - vec_res_value = wrapper::vmul(vec_elements, vec_res_value); + { + auto carry_res = wrapper::vmul(wrapper::vgethigh(vec_res_value), wrapper::vgetlow(vec_res_value)); + T res = 1; + for(int i = 0; i < S / 2; ++i) + { + res *= wrapper::vgetlane(carry_res, i); + } + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res *= *(input_ptr + x); + } + + *(reinterpret_cast<T *>(output.ptr())) = res; break; + } case ReductionOperation::ARG_IDX_MIN: { - auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); - vec_res_idx = calculate_index<decltype(vec_res_value)>(id.x(), temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); - vec_res_value = temp_vec_res_value; + auto idx = calculate_vector_index<decltype(vec_res_value)>(vec_res_idx, vec_res_value, op); + auto res = static_cast<T>(wrapper::vgetlane(calculate_min(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + if(*(input_ptr + x) < res) + { + idx = x; + res = *(input_ptr + x); + } + } + *(reinterpret_cast<uint32_t *>(output.ptr())) = idx; break; } case ReductionOperation::ARG_IDX_MAX: { - auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); - vec_res_idx = calculate_index<decltype(vec_res_value)>(id.x(), temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); - vec_res_value = temp_vec_res_value; + auto idx = calculate_vector_index<decltype(vec_res_value)>(vec_res_idx, vec_res_value, op); + auto res = static_cast<T>(wrapper::vgetlane(calculate_max(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + if(*(input_ptr + x) > res) + { + idx = x; + res = *(input_ptr + x); + } + } + *(reinterpret_cast<uint32_t *>(output.ptr())) = idx; break; } case ReductionOperation::MIN: { - vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + auto res = static_cast<T>(wrapper::vgetlane(calculate_min(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res = *(input_ptr + x) < res ? *(input_ptr + x) : res; + } + *(reinterpret_cast<T *>(output.ptr())) = res; break; } case ReductionOperation::MAX: { - vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + auto res = static_cast<T>(wrapper::vgetlane(calculate_max(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res = *(input_ptr + x) > res ? *(input_ptr + x) : res; + } + *(reinterpret_cast<T *>(output.ptr())) = res; break; } default: ARM_COMPUTE_ERROR("Not supported"); } }, - input); - - switch(op) - { - case ReductionOperation::SUM: - case ReductionOperation::SUM_SQUARE: - case ReductionOperation::MEAN_SUM: - { - auto carry_res = wrapper::vpadd(wrapper::vgethigh(vec_res_value), wrapper::vgetlow(vec_res_value)); - for(int i = 0; i < S / 4; ++i) - { - carry_res = wrapper::vpadd(carry_res, carry_res); - } - auto res = wrapper::vgetlane(carry_res, 0); - - if(op == ReductionOperation::MEAN_SUM) - { - res /= in_info.dimension(0); - } - - *(reinterpret_cast<T *>(output.ptr())) = res; - break; - } - case ReductionOperation::PROD: - { - auto carry_res = wrapper::vmul(wrapper::vgethigh(vec_res_value), wrapper::vgetlow(vec_res_value)); - T res = 1; - for(int i = 0; i < S / 2; ++i) - { - res *= wrapper::vgetlane(carry_res, i); - } - *(reinterpret_cast<T *>(output.ptr())) = res; - break; - } - case ReductionOperation::ARG_IDX_MIN: - case ReductionOperation::ARG_IDX_MAX: - { - auto res = calculate_vector_index<decltype(vec_res_value)>(vec_res_idx, vec_res_value, op); - *(reinterpret_cast<uint32_t *>(output.ptr())) = res; - break; - } - case ReductionOperation::MIN: - { - *(reinterpret_cast<T *>(output.ptr())) = wrapper::vgetlane(calculate_min(vec_res_value), 0); - break; - } - case ReductionOperation::MAX: - { - *(reinterpret_cast<T *>(output.ptr())) = wrapper::vgetlane(calculate_max(vec_res_value), 0); - break; - } - default: - ARM_COMPUTE_ERROR("Not supported"); - } + input, output); } }; template <typename T> struct RedOpX_quantized { - inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, const ReductionOperation op) + inline void operator()(const Window &in_window, Window &out_window, const ITensor *in, ITensor *out, const ReductionOperation op) { - ARM_COMPUTE_UNUSED(out_slice); - using PromotedType = typename wrapper::traits::promote<typename wrapper::traits::promote<T>::type>::type; + const TensorInfo in_info = *(in->info()); const UniformQuantizationInfo iq_info = in_info.quantization_info().uniform(); - auto vec_res_value1 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); - auto vec_res_value2 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); - auto vec_res_value3 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); - auto vec_res_value4 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); + Iterator input(in, in_window); + Iterator output(out, out_window); + const int window_step_x = 16 / sizeof(T); + const auto window_start_x = static_cast<int>(in_window.x().start()); + const auto window_end_x = static_cast<int>(in_window.x().end()); + + execute_window_loop(in_window, [&](const Coordinates &) + { + const auto input_ptr = reinterpret_cast<T *>(input.ptr()); - auto vec_res_value1_f = vdupq_n_f32(static_cast<float>(1.f)); - auto vec_res_value2_f = vdupq_n_f32(static_cast<float>(1.f)); - auto vec_res_value3_f = vdupq_n_f32(static_cast<float>(1.f)); - auto vec_res_value4_f = vdupq_n_f32(static_cast<float>(1.f)); + auto vec_res_value1 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); + auto vec_res_value2 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); + auto vec_res_value3 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); + auto vec_res_value4 = wrapper::vdup_n(static_cast<PromotedType>(0.f), wrapper::traits::vector_128_tag{}); - typename wrapper::traits::neon_vector<T, 16>::type vec_res_value = { 0 }; + auto vec_res_value1_f = vdupq_n_f32(static_cast<float>(1.f)); + auto vec_res_value2_f = vdupq_n_f32(static_cast<float>(1.f)); + auto vec_res_value3_f = vdupq_n_f32(static_cast<float>(1.f)); + auto vec_res_value4_f = vdupq_n_f32(static_cast<float>(1.f)); - if(op == ReductionOperation::ARG_IDX_MAX || op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::MIN || op == ReductionOperation::MAX) - { - vec_res_value = wrapper::vdup_n(*reinterpret_cast<T *>(input.ptr()), wrapper::traits::vector_128_tag{}); - } + typename wrapper::traits::neon_vector<T, 16>::type vec_res_value = { 0 }; - uint32x4x4_t vec_res_idx{ { 0 } }; - execute_window_loop(in_slice, [&](const Coordinates & id) - { - const auto vec_elements = wrapper::vloadq(reinterpret_cast<T *>(input.ptr())); - switch(op) + if(op == ReductionOperation::ARG_IDX_MAX || op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::MIN || op == ReductionOperation::MAX) { - case ReductionOperation::SUM: - case ReductionOperation::MEAN_SUM: - { - 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_res_value1 = wrapper::vadd(temp32x4t_1, vec_res_value1); - vec_res_value2 = wrapper::vadd(temp32x4t_2, vec_res_value2); - vec_res_value3 = wrapper::vadd(temp32x4t_3, vec_res_value3); - vec_res_value4 = wrapper::vadd(temp32x4t_4, vec_res_value4); - break; - } - case ReductionOperation::PROD: + vec_res_value = wrapper::vdup_n(*input_ptr, wrapper::traits::vector_128_tag{}); + } + + uint32x4x4_t vec_res_idx{ { 0 } }; + // Compute window_step_x elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + const auto vec_elements = wrapper::vloadq(input_ptr + x); + switch(op) { - const auto offset32x4f_4 = vdupq_n_f32(iq_info.offset); - const auto scale32x4f_4 = vdupq_n_f32(iq_info.scale); - - 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)); - - auto temp32x4f_1 = wrapper::vcvt<float>(temp32x4t_1); - auto temp32x4f_2 = wrapper::vcvt<float>(temp32x4t_2); - auto temp32x4f_3 = wrapper::vcvt<float>(temp32x4t_3); - auto temp32x4f_4 = wrapper::vcvt<float>(temp32x4t_4); - - //de-quantize vec_elements - temp32x4f_1 = vmulq_f32(vsubq_f32(temp32x4f_1, offset32x4f_4), scale32x4f_4); - temp32x4f_2 = vmulq_f32(vsubq_f32(temp32x4f_2, offset32x4f_4), scale32x4f_4); - temp32x4f_3 = vmulq_f32(vsubq_f32(temp32x4f_3, offset32x4f_4), scale32x4f_4); - temp32x4f_4 = vmulq_f32(vsubq_f32(temp32x4f_4, offset32x4f_4), scale32x4f_4); - - vec_res_value1_f = vmulq_f32(temp32x4f_1, vec_res_value1_f); - vec_res_value2_f = vmulq_f32(temp32x4f_2, vec_res_value2_f); - vec_res_value3_f = vmulq_f32(temp32x4f_3, vec_res_value3_f); - vec_res_value4_f = vmulq_f32(temp32x4f_4, vec_res_value4_f); - break; + case ReductionOperation::SUM: + case ReductionOperation::MEAN_SUM: + { + 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_res_value1 = wrapper::vadd(temp32x4t_1, vec_res_value1); + vec_res_value2 = wrapper::vadd(temp32x4t_2, vec_res_value2); + vec_res_value3 = wrapper::vadd(temp32x4t_3, vec_res_value3); + vec_res_value4 = wrapper::vadd(temp32x4t_4, vec_res_value4); + break; + } + case ReductionOperation::PROD: + { + const auto offset32x4f_4 = vdupq_n_f32(iq_info.offset); + const auto scale32x4f_4 = vdupq_n_f32(iq_info.scale); + + 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)); + + auto temp32x4f_1 = wrapper::vcvt<float>(temp32x4t_1); + auto temp32x4f_2 = wrapper::vcvt<float>(temp32x4t_2); + auto temp32x4f_3 = wrapper::vcvt<float>(temp32x4t_3); + auto temp32x4f_4 = wrapper::vcvt<float>(temp32x4t_4); + + //de-quantize vec_elements + temp32x4f_1 = vmulq_f32(vsubq_f32(temp32x4f_1, offset32x4f_4), scale32x4f_4); + temp32x4f_2 = vmulq_f32(vsubq_f32(temp32x4f_2, offset32x4f_4), scale32x4f_4); + temp32x4f_3 = vmulq_f32(vsubq_f32(temp32x4f_3, offset32x4f_4), scale32x4f_4); + temp32x4f_4 = vmulq_f32(vsubq_f32(temp32x4f_4, offset32x4f_4), scale32x4f_4); + + vec_res_value1_f = vmulq_f32(temp32x4f_1, vec_res_value1_f); + vec_res_value2_f = vmulq_f32(temp32x4f_2, vec_res_value2_f); + vec_res_value3_f = vmulq_f32(temp32x4f_3, vec_res_value3_f); + vec_res_value4_f = vmulq_f32(temp32x4f_4, vec_res_value4_f); + break; + } + case ReductionOperation::ARG_IDX_MIN: + { + auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + vec_res_idx = calculate_index_quantized<decltype(vec_res_value)>(x, temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::ARG_IDX_MAX: + { + auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + vec_res_idx = calculate_index_quantized<decltype(vec_res_value)>(x, temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::MIN: + { + vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + break; + } + case ReductionOperation::MAX: + { + vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + break; + } + default: + ARM_COMPUTE_ERROR("Not supported"); } + } + + switch(op) + { case ReductionOperation::ARG_IDX_MIN: { - auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); - vec_res_idx = calculate_index_quantized(id.x(), temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); - vec_res_value = temp_vec_res_value; + auto idx = calculate_vector_index_quantized<decltype(vec_res_value)>(vec_res_idx, vec_res_value, op); + auto res = static_cast<T>(wrapper::vgetlane(calculate_min(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + if(*(input_ptr + x) < res) + { + idx = x; + res = *(input_ptr + x); + } + } + *(reinterpret_cast<uint32_t *>(output.ptr())) = idx; break; } case ReductionOperation::ARG_IDX_MAX: { - auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); - vec_res_idx = calculate_index_quantized(id.x(), temp_vec_res_value, vec_res_value, vec_res_idx, op, 0); - vec_res_value = temp_vec_res_value; + auto idx = calculate_vector_index_quantized<decltype(vec_res_value)>(vec_res_idx, vec_res_value, op); + auto res = static_cast<T>(wrapper::vgetlane(calculate_max(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + if(*(input_ptr + x) > res) + { + idx = x; + res = *(input_ptr + x); + } + } + *(reinterpret_cast<uint32_t *>(output.ptr())) = idx; break; } case ReductionOperation::MIN: { - vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + auto res = static_cast<T>(wrapper::vgetlane(calculate_min(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res = *(input_ptr + x) < res ? *(input_ptr + x) : res; + } + *(reinterpret_cast<T *>(output.ptr())) = res; break; } case ReductionOperation::MAX: { - vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + auto res = static_cast<T>(wrapper::vgetlane(calculate_max(vec_res_value), 0)); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res = *(input_ptr + x) > res ? *(input_ptr + x) : res; + } + *(reinterpret_cast<T *>(output.ptr())) = res; break; } - default: - ARM_COMPUTE_ERROR("Not supported"); - } - }, - input); + case ReductionOperation::PROD: + { + auto carry_res = wrapper::vmul(vec_res_value1_f, vec_res_value2_f); + carry_res = wrapper::vmul(carry_res, vec_res_value3_f); + carry_res = wrapper::vmul(carry_res, vec_res_value4_f); - switch(op) - { - case ReductionOperation::ARG_IDX_MIN: - case ReductionOperation::ARG_IDX_MAX: - { - auto res = calculate_vector_index_quantized(vec_res_idx, vec_res_value, op); - *(reinterpret_cast<PromotedType *>(output.ptr())) = res; - break; - } - case ReductionOperation::MIN: - { - *(output.ptr()) = static_cast<T>(wrapper::vgetlane(calculate_min(vec_res_value), 0)); - break; - } - case ReductionOperation::MAX: - { - *(output.ptr()) = static_cast<T>(wrapper::vgetlane(calculate_max(vec_res_value), 0)); - break; - } - case ReductionOperation::PROD: - { - auto carry_res = wrapper::vmul(vec_res_value1_f, vec_res_value2_f); - carry_res = wrapper::vmul(carry_res, vec_res_value3_f); - carry_res = wrapper::vmul(carry_res, vec_res_value4_f); + float res = wrapper::vgetlane(carry_res, 0); + res *= wrapper::vgetlane(carry_res, 1); + res *= wrapper::vgetlane(carry_res, 2); + res *= wrapper::vgetlane(carry_res, 3); + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + //de-quantize input + if(std::is_same<T, uint8_t>::value) + { + res *= dequantize_qasymm8(*(input_ptr + x), iq_info); + } + else + { + res *= dequantize_qasymm8_signed(*(input_ptr + x), iq_info); + } + } - float res = wrapper::vgetlane(carry_res, 0); - res *= wrapper::vgetlane(carry_res, 1); - res *= wrapper::vgetlane(carry_res, 2); - res *= wrapper::vgetlane(carry_res, 3); + //re-quantize result + if(std::is_same<T, uint8_t>::value) + { + res = quantize_qasymm8(res, iq_info); + } + else + { + res = quantize_qasymm8_signed(res, iq_info); + } - //re-quantize result - if(std::is_same<T, uint8_t>::value) - { - res = quantize_qasymm8(res, iq_info); + *reinterpret_cast<T *>(output.ptr()) = static_cast<T>(res); + break; } - else + case ReductionOperation::SUM: + case ReductionOperation::MEAN_SUM: { - res = quantize_qasymm8_signed(res, iq_info); - } + auto carry_res = wrapper::vadd(vec_res_value1, vec_res_value2); + carry_res = wrapper::vadd(carry_res, vec_res_value3); + carry_res = wrapper::vadd(carry_res, vec_res_value4); - *reinterpret_cast<T *>(output.ptr()) = static_cast<T>(res); - break; - } - default: - { - auto carry_res = wrapper::vadd(vec_res_value1, vec_res_value2); - carry_res = wrapper::vadd(carry_res, vec_res_value3); - carry_res = wrapper::vadd(carry_res, vec_res_value4); + auto carry_paddition = wrapper::vpadd(wrapper::vgethigh(carry_res), wrapper::vgetlow(carry_res)); + carry_paddition = wrapper::vpadd(carry_paddition, carry_paddition); + auto res = static_cast<int32_t>(wrapper::vgetlane(carry_paddition, 0)); - auto carry_paddition = wrapper::vpadd(wrapper::vgethigh(carry_res), wrapper::vgetlow(carry_res)); - carry_paddition = wrapper::vpadd(carry_paddition, carry_paddition); - auto res = static_cast<int32_t>(wrapper::vgetlane(carry_paddition, 0)); + // Compute left-over elements + for(; x < window_end_x; ++x) + { + res += *(input_ptr + x); + } - if(op == ReductionOperation::MEAN_SUM) - { - res /= static_cast<int32_t>(in_info.dimension(0)); - } - else - { - // Subtract accumulated offsets - res -= (in_info.dimension(0) - 1) * iq_info.offset; + if(op == ReductionOperation::MEAN_SUM) + { + res /= static_cast<int32_t>(in_info.dimension(0)); + } + else + { + // Subtract accumulated offsets + res -= (in_info.dimension(0) - 1) * iq_info.offset; + } + *reinterpret_cast<T *>(output.ptr()) = utils::cast::saturate_cast<T>(res); + break; } - *reinterpret_cast<T *>(output.ptr()) = utils::cast::saturate_cast<T>(res); + default: + ARM_COMPUTE_ERROR("Not supported"); } - } + }, + input, output); } }; @@ -744,100 +854,204 @@ struct RedOpYZW using ExactTagType = typename wrapper::traits::neon_vector<T, S>::tag_type; using neon_vector = typename wrapper::traits::neon_vector<T, S>::type; - inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, int axis, const ReductionOperation op) + inline void operator()(const Window &in_window, Window &out_window, const ITensor *in, ITensor *out, int axis, const ReductionOperation op) { - ARM_COMPUTE_UNUSED(out_slice); + const TensorInfo in_info = *(in->info()); + + Iterator input(in, in_window); + Iterator output(out, out_window); + const int window_step_x = 16 / sizeof(T); + const auto window_start_x = static_cast<int>(in_window.x().start()); + const auto window_end_x = static_cast<int>(in_window.x().end()); - execute_window_loop(in_slice, [&](const Coordinates &) + execute_window_loop(in_window, [&](const Coordinates &) { - neon_vector vec_res_value = { 0 }; - switch(op) + const auto input_ptr = reinterpret_cast<T *>(input.ptr()); + + // Compute window_step_x elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) { - case ReductionOperation::ARG_IDX_MAX: - case ReductionOperation::ARG_IDX_MIN: - case ReductionOperation::MIN: - case ReductionOperation::MAX: + neon_vector vec_res_value = { 0 }; + switch(op) { - vec_res_value = wrapper::vloadq(reinterpret_cast<T *>(input.ptr())); - break; + case ReductionOperation::ARG_IDX_MAX: + case ReductionOperation::ARG_IDX_MIN: + case ReductionOperation::MIN: + case ReductionOperation::MAX: + { + vec_res_value = wrapper::vloadq(input_ptr + x); + break; + } + case ReductionOperation::PROD: + { + vec_res_value = wrapper::vdup_n(static_cast<T>(1.f), ExactTagType{}); + break; + } + default: + { + vec_res_value = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); + break; + } } - case ReductionOperation::PROD: + uint32x4x4_t vec_res_idx{ { 0 } }; + + for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) { - vec_res_value = wrapper::vdup_n(static_cast<T>(1.f), ExactTagType{}); - break; + const T *in_ptr = reinterpret_cast<T *>(input.ptr() + x * sizeof(T) + in_info.strides_in_bytes()[axis] * dim); + const auto vec_elements = wrapper::vloadq(in_ptr); + switch(op) + { + case ReductionOperation::SUM: + case ReductionOperation::MEAN_SUM: + vec_res_value = wrapper::vadd(vec_elements, vec_res_value); + break; + case ReductionOperation::SUM_SQUARE: + vec_res_value = wrapper::vadd(wrapper::vmul(vec_elements, vec_elements), vec_res_value); + break; + case ReductionOperation::PROD: + vec_res_value = wrapper::vmul(vec_elements, vec_res_value); + break; + case ReductionOperation::ARG_IDX_MIN: + { + auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + vec_res_idx = calculate_index(dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::ARG_IDX_MAX: + { + auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + vec_res_idx = calculate_index(dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::MIN: + { + vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + break; + } + case ReductionOperation::MAX: + { + vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + break; + } + default: + ARM_COMPUTE_ERROR("Not supported"); + } } - default: + + if(op == ReductionOperation::MEAN_SUM) { - vec_res_value = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); - break; + auto vec_width_inv = wrapper::vinv(wrapper::vdup_n(static_cast<T>(in_info.dimension(axis)), ExactTagType{})); + vec_res_value = wrapper::vmul(vec_res_value, vec_width_inv); + } + + if(op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::ARG_IDX_MAX) + { + wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()) + x, vec_res_idx.val[0]); +#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC + if(std::is_same<T, float16_t>::value) + { + wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()) + x + 4, vec_res_idx.val[1]); + } +#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC + } + else + { + wrapper::vstore(reinterpret_cast<T *>(output.ptr() + x * sizeof(T)), vec_res_value); } } - uint32x4x4_t vec_res_idx{ { 0 } }; - for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) + // Compute left-over elements + for(; x < window_end_x; ++x) { - const T *in_ptr = reinterpret_cast<T *>(input.ptr() + in_info.strides_in_bytes()[axis] * dim); - const auto vec_elements = wrapper::vloadq(in_ptr); + auto res_value = 0.f; switch(op) { - case ReductionOperation::SUM: - case ReductionOperation::MEAN_SUM: - vec_res_value = wrapper::vadd(vec_elements, vec_res_value); - break; - case ReductionOperation::SUM_SQUARE: - vec_res_value = wrapper::vadd(wrapper::vmul(vec_elements, vec_elements), vec_res_value); - break; - case ReductionOperation::PROD: - vec_res_value = wrapper::vmul(vec_elements, vec_res_value); - break; + case ReductionOperation::ARG_IDX_MAX: case ReductionOperation::ARG_IDX_MIN: + case ReductionOperation::MIN: + case ReductionOperation::MAX: { - auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); - vec_res_idx = calculate_index(dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); - vec_res_value = temp_vec_res_value; + res_value = *(input_ptr + x); break; } - case ReductionOperation::ARG_IDX_MAX: + case ReductionOperation::PROD: { - auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); - vec_res_idx = calculate_index(dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); - vec_res_value = temp_vec_res_value; + res_value = static_cast<T>(1.f); break; } - case ReductionOperation::MIN: + default: { - vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + res_value = static_cast<T>(0.f); break; } - case ReductionOperation::MAX: + } + + uint32_t res_idx = 0; + for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) + { + const T *in_ptr = reinterpret_cast<T *>(input.ptr() + x * sizeof(T) + in_info.strides_in_bytes()[axis] * dim); + + switch(op) { - vec_res_value = wrapper::vmax(vec_elements, vec_res_value); - break; + case ReductionOperation::SUM: + case ReductionOperation::MEAN_SUM: + res_value += *in_ptr; + break; + case ReductionOperation::SUM_SQUARE: + res_value += *in_ptr * *in_ptr; + break; + case ReductionOperation::PROD: + res_value *= *in_ptr; + break; + case ReductionOperation::ARG_IDX_MIN: + { + if(*in_ptr < res_value) + { + res_value = *in_ptr; + res_idx = dim; + } + break; + } + case ReductionOperation::ARG_IDX_MAX: + { + if(*in_ptr > res_value) + { + res_value = *in_ptr; + res_idx = dim; + } + break; + } + case ReductionOperation::MIN: + { + res_value = *in_ptr < res_value ? *in_ptr : res_value; + break; + } + case ReductionOperation::MAX: + { + res_value = *in_ptr > res_value ? *in_ptr : res_value; + break; + } + default: + ARM_COMPUTE_ERROR("Not supported"); } - default: - ARM_COMPUTE_ERROR("Not supported"); } - } - if(op == ReductionOperation::MEAN_SUM) - { - auto vec_width_inv = wrapper::vinv(wrapper::vdup_n(static_cast<T>(in_info.dimension(axis)), ExactTagType{})); - vec_res_value = wrapper::vmul(vec_res_value, vec_width_inv); - } + if(op == ReductionOperation::MEAN_SUM) + { + res_value /= in_info.dimension(axis); + } - if(op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::ARG_IDX_MAX) - { - wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()), vec_res_idx.val[0]); -#ifdef __ARM_FEATURE_FP16_VECTOR_ARITHMETIC - if(std::is_same<T, float16_t>::value) + if(op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::ARG_IDX_MAX) { - wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()) + 4, vec_res_idx.val[1]); + *(reinterpret_cast<uint32_t *>(output.ptr()) + x) = res_idx; + } + else + { + *(reinterpret_cast<T *>(output.ptr() + x * sizeof(T))) = res_value; } -#endif // __ARM_FEATURE_FP16_VECTOR_ARITHMETIC - } - else - { - wrapper::vstore(reinterpret_cast<T *>(output.ptr()), vec_res_value); } }, input, output); @@ -851,51 +1065,95 @@ struct RedOpYZW_complex using ExactTagType = typename wrapper::traits::neon_vector<T, S>::tag_type; using neon_vector = typename wrapper::traits::neon_vector<T, S>::type; - inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, int, const ReductionOperation) + inline void operator()(const Window &in_window, Window &out_window, const ITensor *in, ITensor *out, int, const ReductionOperation) { - ARM_COMPUTE_UNUSED(out_slice); ARM_COMPUTE_ERROR_ON(axis != 2); + const TensorInfo in_info = *(in->info()); + + Iterator input(in, in_window); + Iterator output(out, out_window); + const int window_step_x = 16 / sizeof(T); + const auto window_start_x = static_cast<int>(in_window.x().start()); + const auto window_end_x = static_cast<int>(in_window.x().end()); + const size_t stride_z = in_info.strides_in_bytes()[axis]; - execute_window_loop(in_slice, [&](const Coordinates &) + execute_window_loop(in_window, [&](const Coordinates &) { - neon_vector vec_res_value_0 = { 0 }; - neon_vector vec_res_value_1 = { 0 }; + // Compute window_step_x elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + neon_vector vec_res_value_0 = { 0 }; + neon_vector vec_res_value_1 = { 0 }; - vec_res_value_0 = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); - vec_res_value_1 = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); + vec_res_value_0 = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); + vec_res_value_1 = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); - for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) - { - T *in_ptr_0; - T *in_ptr_1; - switch(axis) + T *out_ptr = reinterpret_cast<T *>(output.ptr() + 2 * x * sizeof(T)); + for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) { - case 2: - in_ptr_0 = reinterpret_cast<T *>(input.ptr() + stride_z * dim); - in_ptr_1 = reinterpret_cast<T *>(input.ptr() + 16 + stride_z * dim); - break; - default: - ARM_COMPUTE_ERROR("Not supported"); - } - const auto vec_elements_0 = wrapper::vloadq(in_ptr_0); - const auto vec_elements_1 = wrapper::vloadq(in_ptr_1); + T *in_ptr_0; + T *in_ptr_1; + switch(axis) + { + case 2: + in_ptr_0 = reinterpret_cast<T *>(input.ptr() + 2 * x * sizeof(T) + stride_z * dim); + in_ptr_1 = reinterpret_cast<T *>(input.ptr() + 2 * x * sizeof(T) + 16 + stride_z * dim); + break; + default: + ARM_COMPUTE_ERROR("Not supported"); + } + const auto vec_elements_0 = wrapper::vloadq(in_ptr_0); + const auto vec_elements_1 = wrapper::vloadq(in_ptr_1); - switch(op) - { - case ReductionOperation::SUM: - vec_res_value_0 = wrapper::vadd(vec_elements_0, vec_res_value_0); - vec_res_value_1 = wrapper::vadd(vec_elements_1, vec_res_value_1); - break; - default: - ARM_COMPUTE_ERROR("Not supported"); + switch(op) + { + case ReductionOperation::SUM: + vec_res_value_0 = wrapper::vadd(vec_elements_0, vec_res_value_0); + vec_res_value_1 = wrapper::vadd(vec_elements_1, vec_res_value_1); + break; + default: + ARM_COMPUTE_ERROR("Not supported"); + } } + + wrapper::vstore(out_ptr, vec_res_value_0); + wrapper::vstore(out_ptr + 4, vec_res_value_1); } - wrapper::vstore(reinterpret_cast<T *>(output.ptr()), vec_res_value_0); - wrapper::vstore(reinterpret_cast<T *>(output.ptr() + 16), vec_res_value_1); + // Compute left-over elements + for(; x < window_end_x; ++x) + { + auto res_value_0 = 0.f; + auto res_value_1 = 0.f; + T *out_ptr = reinterpret_cast<T *>(output.ptr() + 2 * x * sizeof(T)); + for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) + { + T *in_ptr; + switch(axis) + { + case 2: + in_ptr = reinterpret_cast<T *>(input.ptr() + 2 * x * sizeof(T) + stride_z * dim); + break; + default: + ARM_COMPUTE_ERROR("Not supported"); + } + switch(op) + { + case ReductionOperation::SUM: + res_value_0 += *in_ptr; + res_value_1 += *(in_ptr + 1); + break; + default: + ARM_COMPUTE_ERROR("Not supported"); + } + } + *out_ptr = res_value_0; + *(out_ptr + 1) = res_value_1; + } }, input, output); } @@ -904,184 +1162,337 @@ struct RedOpYZW_complex template <typename T> struct RedOpYZW_quantized { - inline void operator()(Iterator &input, Iterator &output, Window &in_slice, Window &out_slice, const TensorInfo &in_info, int axis, const ReductionOperation op) + inline void operator()(const Window &in_window, Window &out_window, const ITensor *in, ITensor *out, int axis, const ReductionOperation op) { - ARM_COMPUTE_UNUSED(out_slice); + const TensorInfo in_info = *(in->info()); + + Iterator input(in, in_window); + Iterator output(out, out_window); + const int window_step_x = 16 / sizeof(T); + const auto window_start_x = static_cast<int>(in_window.x().start()); + const auto window_end_x = static_cast<int>(in_window.x().end()); using PromotedType = typename wrapper::traits::promote<typename wrapper::traits::promote<T>::type>::type; const UniformQuantizationInfo iq_info = in_info.quantization_info().uniform(); - execute_window_loop(in_slice, [&](const Coordinates &) + execute_window_loop(in_window, [&](const Coordinates &) { - uint32x4x4_t vec_res_idx{ { 0 } }; - auto vec_res_value1 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); - auto vec_res_value2 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); - auto vec_res_value3 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); - auto vec_res_value4 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); + const auto input_ptr = reinterpret_cast<T *>(input.ptr()); - auto vec_res_value1_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); - auto vec_res_value2_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); - auto vec_res_value3_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); - auto vec_res_value4_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); + // Compute window_step_x elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + uint32x4x4_t vec_res_idx{ { 0 } }; + auto vec_res_value1 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); + auto vec_res_value2 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); + auto vec_res_value3 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); + auto vec_res_value4 = wrapper::vdup_n(static_cast<PromotedType>(0), wrapper::traits::vector_128_tag{}); - auto vec_res_value = wrapper::vloadq(reinterpret_cast<T *>(input.ptr())); + auto vec_res_value1_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); + auto vec_res_value2_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); + auto vec_res_value3_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); + auto vec_res_value4_f = wrapper::vdup_n(static_cast<float>(1), wrapper::traits::vector_128_tag{}); + + auto vec_res_value = wrapper::vloadq(input_ptr + x); + + for(unsigned int index_dim = 0; index_dim < in_info.dimension(axis); ++index_dim) + { + const T *in_ptr = input_ptr + x + in_info.strides_in_bytes()[axis] * index_dim; + const auto vec_elements = wrapper::vloadq(in_ptr); + switch(op) + { + case ReductionOperation::SUM: + case ReductionOperation::MEAN_SUM: + { + 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_res_value1 = wrapper::vadd(temp32x4t_1, vec_res_value1); + vec_res_value2 = wrapper::vadd(temp32x4t_2, vec_res_value2); + vec_res_value3 = wrapper::vadd(temp32x4t_3, vec_res_value3); + vec_res_value4 = wrapper::vadd(temp32x4t_4, vec_res_value4); + break; + } + case ReductionOperation::PROD: + { + const auto offset32x4f_4 = wrapper::vdup_n(static_cast<float>(iq_info.offset), wrapper::traits::vector_128_tag{}); + const auto scale32x4f_4 = wrapper::vdup_n(iq_info.scale, wrapper::traits::vector_128_tag{}); + + 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)); + + auto temp32x4f_1 = wrapper::vcvt<float>(temp32x4t_1); + auto temp32x4f_2 = wrapper::vcvt<float>(temp32x4t_2); + auto temp32x4f_3 = wrapper::vcvt<float>(temp32x4t_3); + auto temp32x4f_4 = wrapper::vcvt<float>(temp32x4t_4); + + //de-quantize vec_elements + temp32x4f_1 = wrapper::vmul(wrapper::vsub(temp32x4f_1, offset32x4f_4), scale32x4f_4); + temp32x4f_2 = wrapper::vmul(wrapper::vsub(temp32x4f_2, offset32x4f_4), scale32x4f_4); + temp32x4f_3 = wrapper::vmul(wrapper::vsub(temp32x4f_3, offset32x4f_4), scale32x4f_4); + temp32x4f_4 = wrapper::vmul(wrapper::vsub(temp32x4f_4, offset32x4f_4), scale32x4f_4); + + vec_res_value1_f = wrapper::vmul(temp32x4f_1, vec_res_value1_f); + vec_res_value2_f = wrapper::vmul(temp32x4f_2, vec_res_value2_f); + vec_res_value3_f = wrapper::vmul(temp32x4f_3, vec_res_value3_f); + vec_res_value4_f = wrapper::vmul(temp32x4f_4, vec_res_value4_f); + break; + } + case ReductionOperation::ARG_IDX_MIN: + { + auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + vec_res_idx = calculate_index_quantized(index_dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::ARG_IDX_MAX: + { + auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + vec_res_idx = calculate_index_quantized(index_dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); + vec_res_value = temp_vec_res_value; + break; + } + case ReductionOperation::MIN: + { + vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + break; + } + case ReductionOperation::MAX: + { + vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + break; + } + default: + ARM_COMPUTE_ERROR("Not supported"); + } + } - for(unsigned int index_dim = 0; index_dim < in_info.dimension(axis); ++index_dim) - { - const T *in_ptr = reinterpret_cast<T *>(input.ptr()) + in_info.strides_in_bytes()[axis] * index_dim; - const auto vec_elements = wrapper::vloadq(in_ptr); switch(op) { + case ReductionOperation::ARG_IDX_MIN: + case ReductionOperation::ARG_IDX_MAX: + { + wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr() + 4 * x), vec_res_idx.val[0]); + wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr() + 4 * x) + 4, vec_res_idx.val[1]); + wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr() + 4 * x) + 8, vec_res_idx.val[2]); + wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr() + 4 * x) + 12, vec_res_idx.val[3]); + break; + } + case ReductionOperation::MIN: + case ReductionOperation::MAX: + { + wrapper::vstore(reinterpret_cast<T *>(output.ptr() + x), vec_res_value); + break; + } case ReductionOperation::SUM: - case ReductionOperation::MEAN_SUM: { - const auto temp16x8t_1 = wrapper::vmovl(wrapper::vgetlow(vec_elements)); - const auto temp16x8t_2 = wrapper::vmovl(wrapper::vgethigh(vec_elements)); + // Subtract offsets + auto offsets = vdupq_n_s32((in_info.dimension(axis) - 1) * iq_info.offset); - 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)); + auto vec_res_s_value1 = wrapper::vreinterpret(vec_res_value1); + auto vec_res_s_value2 = wrapper::vreinterpret(vec_res_value2); + auto vec_res_s_value3 = wrapper::vreinterpret(vec_res_value3); + auto vec_res_s_value4 = wrapper::vreinterpret(vec_res_value4); - vec_res_value1 = wrapper::vadd(temp32x4t_1, vec_res_value1); - vec_res_value2 = wrapper::vadd(temp32x4t_2, vec_res_value2); - vec_res_value3 = wrapper::vadd(temp32x4t_3, vec_res_value3); - vec_res_value4 = wrapper::vadd(temp32x4t_4, vec_res_value4); + vec_res_s_value1 = wrapper::vsub(vec_res_s_value1, offsets); + vec_res_s_value2 = wrapper::vsub(vec_res_s_value2, offsets); + vec_res_s_value3 = wrapper::vsub(vec_res_s_value3, offsets); + vec_res_s_value4 = wrapper::vsub(vec_res_s_value4, offsets); + + const auto temp16x8t_1 = wrapper::vcombine(wrapper::vqmovn(vec_res_s_value1), wrapper::vqmovn(vec_res_s_value2)); + const auto temp16x8t_2 = wrapper::vcombine(wrapper::vqmovn(vec_res_s_value3), wrapper::vqmovn(vec_res_s_value4)); + + combine_and_store<T>(temp16x8t_1, temp16x8t_2, output, x); + break; + } + case ReductionOperation::MEAN_SUM: + { + const auto vec_width_inv = wrapper::vinv(wrapper::vdup_n(static_cast<float>(in_info.dimension(axis)), wrapper::traits::vector_128_tag{})); + vec_res_value1_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value1), vec_width_inv); + vec_res_value2_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value2), vec_width_inv); + vec_res_value3_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value3), vec_width_inv); + vec_res_value4_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value4), vec_width_inv); + + vec_res_value1 = wrapper::vcvt<T>(vec_res_value1_f); + vec_res_value2 = wrapper::vcvt<T>(vec_res_value2_f); + vec_res_value3 = wrapper::vcvt<T>(vec_res_value3_f); + vec_res_value4 = wrapper::vcvt<T>(vec_res_value4_f); + + const auto temp16x8t_1 = wrapper::vcombine(wrapper::vqmovn(vec_res_value1), wrapper::vqmovn(vec_res_value2)); + const auto temp16x8t_2 = wrapper::vcombine(wrapper::vqmovn(vec_res_value3), wrapper::vqmovn(vec_res_value4)); + auto res = wrapper::vcombine(wrapper::vqmovn(temp16x8t_1), wrapper::vqmovn(temp16x8t_2)); + + wrapper::vstore(reinterpret_cast<T *>(output.ptr() + x), res); break; } case ReductionOperation::PROD: { const auto offset32x4f_4 = wrapper::vdup_n(static_cast<float>(iq_info.offset), wrapper::traits::vector_128_tag{}); - const auto scale32x4f_4 = wrapper::vdup_n(iq_info.scale, wrapper::traits::vector_128_tag{}); + const auto iscale32x4f_4 = vinvq_f32(vdupq_n_f32(iq_info.scale)); - const auto temp16x8t_1 = wrapper::vmovl(wrapper::vgetlow(vec_elements)); - const auto temp16x8t_2 = wrapper::vmovl(wrapper::vgethigh(vec_elements)); + //re-quantize + vec_res_value1_f = wrapper::vadd(wrapper::vmul(vec_res_value1_f, iscale32x4f_4), offset32x4f_4); + vec_res_value2_f = wrapper::vadd(wrapper::vmul(vec_res_value2_f, iscale32x4f_4), offset32x4f_4); + vec_res_value3_f = wrapper::vadd(wrapper::vmul(vec_res_value3_f, iscale32x4f_4), offset32x4f_4); + vec_res_value4_f = wrapper::vadd(wrapper::vmul(vec_res_value4_f, iscale32x4f_4), offset32x4f_4); - 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_res_value1 = wrapper::vcvt<T>(vec_res_value1_f); + vec_res_value2 = wrapper::vcvt<T>(vec_res_value2_f); + vec_res_value3 = wrapper::vcvt<T>(vec_res_value3_f); + vec_res_value4 = wrapper::vcvt<T>(vec_res_value4_f); - auto temp32x4f_1 = wrapper::vcvt<float>(temp32x4t_1); - auto temp32x4f_2 = wrapper::vcvt<float>(temp32x4t_2); - auto temp32x4f_3 = wrapper::vcvt<float>(temp32x4t_3); - auto temp32x4f_4 = wrapper::vcvt<float>(temp32x4t_4); + const auto temp16x8t_1 = wrapper::vcombine(wrapper::vqmovn(vec_res_value1), wrapper::vqmovn(vec_res_value2)); + const auto temp16x8t_2 = wrapper::vcombine(wrapper::vqmovn(vec_res_value3), wrapper::vqmovn(vec_res_value4)); + auto res = wrapper::vcombine(wrapper::vqmovn(temp16x8t_1), wrapper::vqmovn(temp16x8t_2)); - //de-quantize vec_elements - temp32x4f_1 = wrapper::vmul(wrapper::vsub(temp32x4f_1, offset32x4f_4), scale32x4f_4); - temp32x4f_2 = wrapper::vmul(wrapper::vsub(temp32x4f_2, offset32x4f_4), scale32x4f_4); - temp32x4f_3 = wrapper::vmul(wrapper::vsub(temp32x4f_3, offset32x4f_4), scale32x4f_4); - temp32x4f_4 = wrapper::vmul(wrapper::vsub(temp32x4f_4, offset32x4f_4), scale32x4f_4); - - vec_res_value1_f = wrapper::vmul(temp32x4f_1, vec_res_value1_f); - vec_res_value2_f = wrapper::vmul(temp32x4f_2, vec_res_value2_f); - vec_res_value3_f = wrapper::vmul(temp32x4f_3, vec_res_value3_f); - vec_res_value4_f = wrapper::vmul(temp32x4f_4, vec_res_value4_f); - break; - } - case ReductionOperation::ARG_IDX_MIN: - { - auto temp_vec_res_value = wrapper::vmin(vec_elements, vec_res_value); - vec_res_idx = calculate_index_quantized(index_dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); - vec_res_value = temp_vec_res_value; + wrapper::vstore(reinterpret_cast<T *>(output.ptr() + x), res); break; } + default: + ARM_COMPUTE_ERROR("Not supported"); + } + } + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + auto res_value = 0; + switch(op) + { case ReductionOperation::ARG_IDX_MAX: + case ReductionOperation::ARG_IDX_MIN: + case ReductionOperation::MIN: + case ReductionOperation::MAX: { - auto temp_vec_res_value = wrapper::vmax(vec_elements, vec_res_value); - vec_res_idx = calculate_index_quantized(index_dim, temp_vec_res_value, vec_res_value, vec_res_idx, op, axis); - vec_res_value = temp_vec_res_value; + res_value = *(input_ptr + x); break; } - case ReductionOperation::MIN: + case ReductionOperation::PROD: { - vec_res_value = wrapper::vmin(vec_elements, vec_res_value); + res_value = static_cast<T>(1.0f); break; } - case ReductionOperation::MAX: + default: { - vec_res_value = wrapper::vmax(vec_elements, vec_res_value); + res_value = static_cast<T>(0.0f); break; } - default: - ARM_COMPUTE_ERROR("Not supported"); } - } - - if(op == ReductionOperation::MEAN_SUM) - { - const auto vec_width_inv = wrapper::vinv(wrapper::vdup_n(static_cast<float>(in_info.dimension(axis)), wrapper::traits::vector_128_tag{})); - vec_res_value1_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value1), vec_width_inv); - vec_res_value2_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value2), vec_width_inv); - vec_res_value3_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value3), vec_width_inv); - vec_res_value4_f = wrapper::vmul(wrapper::vcvt<float>(vec_res_value4), vec_width_inv); - - vec_res_value1 = wrapper::vcvt<T>(vec_res_value1_f); - vec_res_value2 = wrapper::vcvt<T>(vec_res_value2_f); - vec_res_value3 = wrapper::vcvt<T>(vec_res_value3_f); - vec_res_value4 = wrapper::vcvt<T>(vec_res_value4_f); - } - else if(op == ReductionOperation::PROD) - { - const auto offset32x4f_4 = wrapper::vdup_n(static_cast<float>(iq_info.offset), wrapper::traits::vector_128_tag{}); - const auto iscale32x4f_4 = vinvq_f32(vdupq_n_f32(iq_info.scale)); - - //re-quantize - vec_res_value1_f = wrapper::vadd(wrapper::vmul(vec_res_value1_f, iscale32x4f_4), offset32x4f_4); - vec_res_value2_f = wrapper::vadd(wrapper::vmul(vec_res_value2_f, iscale32x4f_4), offset32x4f_4); - vec_res_value3_f = wrapper::vadd(wrapper::vmul(vec_res_value3_f, iscale32x4f_4), offset32x4f_4); - vec_res_value4_f = wrapper::vadd(wrapper::vmul(vec_res_value4_f, iscale32x4f_4), offset32x4f_4); - - vec_res_value1 = wrapper::vcvt<T>(vec_res_value1_f); - vec_res_value2 = wrapper::vcvt<T>(vec_res_value2_f); - vec_res_value3 = wrapper::vcvt<T>(vec_res_value3_f); - vec_res_value4 = wrapper::vcvt<T>(vec_res_value4_f); - } + uint32_t res_idx = 0; - if(op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::ARG_IDX_MAX) - { - wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()), vec_res_idx.val[0]); - wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()) + 4, vec_res_idx.val[1]); - wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()) + 8, vec_res_idx.val[2]); - wrapper::vstore(reinterpret_cast<uint32_t *>(output.ptr()) + 12, vec_res_idx.val[3]); - } - else if(op == ReductionOperation::MIN || op == ReductionOperation::MAX) - { - wrapper::vstore(reinterpret_cast<T *>(output.ptr()), vec_res_value); - } - else - { - if(op == ReductionOperation::SUM) + for(unsigned int dim = 0; dim < in_info.dimension(axis); ++dim) { - // Subtract offsets - auto offsets = vdupq_n_s32((in_info.dimension(axis) - 1) * iq_info.offset); - - auto vec_res_s_value1 = wrapper::vreinterpret(vec_res_value1); - auto vec_res_s_value2 = wrapper::vreinterpret(vec_res_value2); - auto vec_res_s_value3 = wrapper::vreinterpret(vec_res_value3); - auto vec_res_s_value4 = wrapper::vreinterpret(vec_res_value4); - - vec_res_s_value1 = wrapper::vsub(vec_res_s_value1, offsets); - vec_res_s_value2 = wrapper::vsub(vec_res_s_value2, offsets); - vec_res_s_value3 = wrapper::vsub(vec_res_s_value3, offsets); - vec_res_s_value4 = wrapper::vsub(vec_res_s_value4, offsets); - - const auto temp16x8t_1 = wrapper::vcombine(wrapper::vqmovn(vec_res_s_value1), wrapper::vqmovn(vec_res_s_value2)); - const auto temp16x8t_2 = wrapper::vcombine(wrapper::vqmovn(vec_res_s_value3), wrapper::vqmovn(vec_res_s_value4)); - - combine_and_store<T>(temp16x8t_1, temp16x8t_2, output); + const T *in_ptr = reinterpret_cast<T *>(input.ptr() + x + in_info.strides_in_bytes()[axis] * dim); + switch(op) + { + case ReductionOperation::SUM: + case ReductionOperation::MEAN_SUM: + { + res_value += *in_ptr; + break; + } + case ReductionOperation::SUM_SQUARE: + { + res_value += *in_ptr * *in_ptr; + break; + } + case ReductionOperation::PROD: + { + //de-quantize input + if(std::is_same<T, uint8_t>::value) + { + res_value *= dequantize_qasymm8(*input_ptr, iq_info); + } + else + { + res_value *= dequantize_qasymm8_signed(*input_ptr, iq_info); + } + break; + } + case ReductionOperation::ARG_IDX_MIN: + { + if(*in_ptr < res_value) + { + res_value = *in_ptr; + res_idx = dim; + } + break; + } + case ReductionOperation::ARG_IDX_MAX: + { + if(*in_ptr > res_value) + { + res_value = *in_ptr; + res_idx = dim; + } + break; + } + case ReductionOperation::MIN: + { + res_value = *in_ptr < res_value ? *in_ptr : res_value; + break; + } + case ReductionOperation::MAX: + { + res_value = *in_ptr > res_value ? *in_ptr : res_value; + break; + } + default: + ARM_COMPUTE_ERROR("Not supported"); + } } - else - { - const auto temp16x8t_1 = wrapper::vcombine(wrapper::vqmovn(vec_res_value1), wrapper::vqmovn(vec_res_value2)); - const auto temp16x8t_2 = wrapper::vcombine(wrapper::vqmovn(vec_res_value3), wrapper::vqmovn(vec_res_value4)); - auto res = wrapper::vcombine(wrapper::vqmovn(temp16x8t_1), wrapper::vqmovn(temp16x8t_2)); - wrapper::vstore(reinterpret_cast<T *>(output.ptr()), res); + switch(op) + { + case ReductionOperation::MEAN_SUM: + { + res_value /= in_info.dimension(axis); + *reinterpret_cast<T *>(output.ptr() + x) = utils::cast::saturate_cast<T>(res_value); + break; + } + case ReductionOperation::SUM: + { + // Subtract accumulated offsets + res_value -= (in_info.dimension(axis) - 1) * iq_info.offset; + *reinterpret_cast<T *>(output.ptr() + x) = utils::cast::saturate_cast<T>(res_value); + break; + } + case ReductionOperation::PROD: + { + //re-quantize result + if(std::is_same<T, uint8_t>::value) + { + res_value = quantize_qasymm8(res_value, iq_info); + } + else + { + res_value = quantize_qasymm8_signed(res_value, iq_info); + } + break; + *(reinterpret_cast<T *>(output.ptr() + x)) = res_value; + } + case ReductionOperation::ARG_IDX_MIN: + case ReductionOperation::ARG_IDX_MAX: + { + *(reinterpret_cast<uint32_t *>(output.ptr() + x * 4)) = res_idx; + break; + } + default: + *(reinterpret_cast<T *>(output.ptr() + x)) = res_value; } } - }, input, output); } @@ -1235,69 +1646,43 @@ Status validate_arguments(const ITensorInfo *input, const ITensorInfo *output, u return Status{}; } - -std::tuple<Status, Window> validate_and_configure_window(ITensorInfo *input, ITensorInfo *output, unsigned int axis, ReductionOperation op) -{ - // Calculate output shape and set if empty - const TensorShape output_shape = arm_compute::misc::shape_calculator::compute_reduced_shape(input->tensor_shape(), axis); - - // Output auto initialization if not yet initialized - const bool is_arg_min_max = (op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::ARG_IDX_MAX); - DataType output_data_type = is_arg_min_max ? DataType::S32 : input->data_type(); - auto_init_if_empty(*output, input->clone()->set_tensor_shape(output_shape).set_data_type(output_data_type).reset_padding().set_is_resizable(true)); - - 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() + : _input(nullptr), _output(nullptr), _reduction_axis(0), _op(ReductionOperation::SUM_SQUARE) { } -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, op); + Coordinates coord; + coord.set_num_dimensions(input->info()->num_dimensions()); + input->info()->set_valid_region(ValidRegion(coord, input->info()->tensor_shape())); + Window win = calculate_max_window(*input->info(), Steps(input->info()->dimension(0))); + INEKernel::configure(win); - ARM_COMPUTE_ERROR_THROW_ON(std::get<0>(win_config)); - - INEKernel::configure(std::get<1>(win_config)); + // Calculate output shape and set if empty + const TensorShape output_shape = arm_compute::misc::shape_calculator::compute_reduced_shape(input->info()->tensor_shape(), axis); + // Output auto initialization if not yet initialized + const bool is_arg_min_max = (op == ReductionOperation::ARG_IDX_MIN || op == ReductionOperation::ARG_IDX_MAX); + DataType output_data_type = is_arg_min_max ? DataType::S32 : input->info()->data_type(); + auto_init_if_empty(*output->info(), input->info()->clone()->set_tensor_shape(output_shape).set_data_type(output_data_type).reset_padding().set_is_resizable(true)); + output->info()->set_valid_region(ValidRegion(coord, output_shape)); } 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, op))); return Status{}; } diff --git a/src/runtime/NEON/functions/NEReductionOperation.cpp b/src/runtime/NEON/functions/NEReductionOperation.cpp index fc2e6d2912..91176bfa45 100644 --- a/src/runtime/NEON/functions/NEReductionOperation.cpp +++ b/src/runtime/NEON/functions/NEReductionOperation.cpp @@ -54,7 +54,7 @@ size_t reduction_window_split_dimension(unsigned int axis) } // namespace NEReductionOperation::NEReductionOperation(std::shared_ptr<IMemoryManager> memory_manager) - : _memory_group(memory_manager), _reduction_kernel(), _fill_border_kernel(), _reshape(), _output_internal(), _window_split(0), _reduction_axis(), _is_reshape_required(false) + : _memory_group(memory_manager), _reduction_kernel(), _reshape(), _output_internal(), _window_split(0), _reduction_axis(), _is_reshape_required(false) { } @@ -128,47 +128,6 @@ void NEReductionOperation::configure(ITensor *input, ITensor *output, unsigned i _window_split = reduction_window_split_dimension(axis); _reduction_axis = axis; - if(axis == 0) - { - // Configure fill border kernel - const BorderSize fill_border_size = _reduction_kernel.border_size(); - PixelValue pixelValue; - switch(op) - { - case ReductionOperation::PROD: - { - pixelValue = PixelValue(1, input->info()->data_type(), input->info()->quantization_info()); - break; - } - case ReductionOperation::MIN: - { - pixelValue = std::get<1>(get_min_max(input->info()->data_type())); - break; - } - case ReductionOperation::MAX: - { - pixelValue = std::get<0>(get_min_max(input->info()->data_type())); - break; - } - case ReductionOperation::ARG_IDX_MAX: - case ReductionOperation::ARG_IDX_MIN: - { - pixelValue = PixelValue(0, input->info()->data_type(), input->info()->quantization_info()); - break; - } - case ReductionOperation::MEAN_SUM: - case ReductionOperation::SUM_SQUARE: - case ReductionOperation::SUM: - { - pixelValue = PixelValue(static_cast<uint32_t>(0)); - break; - } - default: - ARM_COMPUTE_ERROR("Reduction Operation unsupported"); - } - _fill_border_kernel.configure(input, fill_border_size, (is_arg_min_max ? BorderMode::REPLICATE : BorderMode::CONSTANT), pixelValue); - } - if(_is_reshape_required) { _reshape.configure(output_internal, output); @@ -179,10 +138,6 @@ void NEReductionOperation::configure(ITensor *input, ITensor *output, unsigned i void NEReductionOperation::run() { MemoryGroupResourceScope scope_mg(_memory_group); - if(_reduction_axis == 0) - { - NEScheduler::get().schedule(&_fill_border_kernel, Window::DimY); - } NEScheduler::get().schedule(&_reduction_kernel, _window_split); if(_is_reshape_required) { diff --git a/tests/validation/NEON/ReductionOperation.cpp b/tests/validation/NEON/ReductionOperation.cpp index 47b36c630c..ed17e6968e 100644 --- a/tests/validation/NEON/ReductionOperation.cpp +++ b/tests/validation/NEON/ReductionOperation.cpp @@ -106,6 +106,28 @@ DATA_TEST_CASE(Validate, framework::DatasetMode::ALL, zip(zip(zip(zip( keep_dims)); ARM_COMPUTE_EXPECT(is_valid == expected, framework::LogLevel::ERRORS); } + +DATA_TEST_CASE(ValidateNoPadding, framework::DatasetMode::ALL, combine(combine(combine(combine(datasets::Small4DShapes(), framework::dataset::make("DataType", DataType::F32)), framework::dataset::make("Axis", +{ 0, 1 })), framework::dataset::make("ReductionOperation", {ReductionOperation::SUM,})), KeepDims), + shape, data_type, axis, op, keep_dims) +{ + TensorShape input_shape = TensorShape(shape); + TensorInfo input_info = TensorInfo(input_shape, 1, data_type); + const bool is_arg_min_max = (op == ReductionOperation::ARG_IDX_MAX) || (op == ReductionOperation::ARG_IDX_MIN); + const bool _keep_dims = keep_dims && !is_arg_min_max; + const TensorShape output_shape = arm_compute::misc::shape_calculator::compute_reduced_shape(shape, axis, keep_dims); + + // Create tensors + Tensor src = create_tensor<Tensor>(input_shape, data_type, 1, QuantizationInfo()); + Tensor dst = create_tensor<Tensor>(output_shape, data_type, 1, QuantizationInfo()); + + // Create and configure function + NEReductionOperation reduction; + reduction.configure(&src, &dst, axis, op, _keep_dims); + + validate(src.info()->padding(), PaddingSize(0, 0, 0, 0)); + validate(dst.info()->padding(), PaddingSize(0, 0, 0, 0)); +} // clang-format on // *INDENT-ON* |