diff options
author | Pablo Marquez Tello <pablo.tello@arm.com> | 2023-09-04 16:08:33 +0100 |
---|---|---|
committer | Pablo Marquez Tello <pablo.tello@arm.com> | 2023-09-13 09:16:23 +0000 |
commit | 145e82e74916a801bd12720564c837c8286042d0 (patch) | |
tree | a66edf213bde752b572376dfa636191cb1e45880 /src/cpu/kernels/instancenorm/generic/neon/fp16.cpp | |
parent | cf219a4be6e9e9637193b5c9aa4f1eedd0a23900 (diff) | |
download | ComputeLibrary-145e82e74916a801bd12720564c837c8286042d0.tar.gz |
Changes to InstanceNrom to enable fp16 in armv8a multi_isa builds
* Code guarded with __ARM_FEATURE_FP16_VECTOR_ARITHMETIC needs
to be moved to an fp16.cpp file to allow compilation with
-march=armv8.2-a+fp16
* Partially resolves MLCE-1102
Change-Id: If53ff1927948b3ad7c9e3c9347bc2af38764e342
Signed-off-by: Pablo Marquez Tello <pablo.tello@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/10243
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Benchmark: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/cpu/kernels/instancenorm/generic/neon/fp16.cpp')
-rw-r--r-- | src/cpu/kernels/instancenorm/generic/neon/fp16.cpp | 139 |
1 files changed, 136 insertions, 3 deletions
diff --git a/src/cpu/kernels/instancenorm/generic/neon/fp16.cpp b/src/cpu/kernels/instancenorm/generic/neon/fp16.cpp index e9fcc84b35..2b7d91b144 100644 --- a/src/cpu/kernels/instancenorm/generic/neon/fp16.cpp +++ b/src/cpu/kernels/instancenorm/generic/neon/fp16.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2022 Arm Limited. + * Copyright (c) 2022-2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -22,20 +22,153 @@ * SOFTWARE. */ #if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) +#include "arm_compute/core/Helpers.h" +#include "src/core/NEON/wrapper/wrapper.h" #include "src/cpu/kernels/instancenorm/generic/neon/impl.h" + namespace arm_compute { namespace cpu { +namespace +{ +template <typename InputType, typename AccType> +void vector_float_sum_fp16(AccType &result, AccType &result_square, const InputType &inputs) +{ + result = wrapper::vadd(result, inputs); + result_square = wrapper::vadd(result_square, wrapper::vmul(inputs, inputs)); +} + +template <typename InputType, typename AccType> +InputType vector_float_norm_fp16(const InputType &inputs, const AccType &vec_mean, const AccType &vec_multip, const AccType &vec_beta) +{ + return wrapper::vadd(wrapper::vmul(wrapper::vsub(inputs, vec_mean), vec_multip), vec_beta); +} + +template <> +inline void vector_float_sum_fp16(float32x4_t &result, float32x4_t &result_square, const float16x8_t &inputs) +{ + vector_float_sum_fp16(result, result_square, wrapper::vcvt<float>(wrapper::vgetlow(inputs))); + vector_float_sum_fp16(result, result_square, wrapper::vcvt<float>(wrapper::vgethigh(inputs))); +} +template <> +inline float16x8_t vector_float_norm_fp16(const float16x8_t &inputs, const float32x4_t &vec_mean, const float32x4_t &vec_multip, const float32x4_t &vec_beta) +{ + const auto input_low = wrapper::vcvt<float>(wrapper::vgetlow(inputs)); + const auto input_high = wrapper::vcvt<float>(wrapper::vgethigh(inputs)); + const auto result_low = wrapper::vcvt<float16_t>(vector_float_norm_fp16(input_low, vec_mean, vec_multip, vec_beta)); + const auto result_high = wrapper::vcvt<float16_t>(vector_float_norm_fp16(input_high, vec_mean, vec_multip, vec_beta)); + float16x8_t result = wrapper::vcombine(result_low, result_high); + + return result; +} + +template <typename AccType> +void instance_normalization_nchw_fp16(const ITensor *input, ITensor *output, float gamma, float beta, float epsilon, const Window &window) +{ + /** SIMD vector tag type. */ + using ExactTagType = typename wrapper::traits::neon_bitvector_tag_t<float16_t, wrapper::traits::BitWidth::W128>; + + // Clear X/Y dimensions on execution window as we handle the planes manually + Window win = window; + win.set(Window::DimX, Window::Dimension(0, 1, 1)); + win.set(Window::DimY, Window::Dimension(0, 1, 1)); + + constexpr int window_step_x = 16 / sizeof(float16_t); + const unsigned int elements_plane = input->info()->dimension(0) * output->info()->dimension(1); + + Iterator input_it(input, win); + execute_window_loop(win, [&](const Coordinates & id) + { + Window win_plane = window; + win_plane.set(Window::DimX, Window::Dimension(0, 1, 1)); + win_plane.set(Window::DimZ, Window::Dimension(id[2], id[2] + 1, 1)); + win_plane.set(3, Window::Dimension(id[3], id[3] + 1, 1)); + + Iterator input_plane_it(input, win_plane); + Iterator output_plane_it(output, win_plane); + + auto sum_h_w = static_cast<AccType>(0.f); + auto sum_squares_h_w = static_cast<AccType>(0.f); + + execute_window_loop(win_plane, [&](const Coordinates &) + { + const auto input_ptr = reinterpret_cast<const float16_t *>(input_plane_it.ptr()); + + auto vec_sum_h_w = wrapper::vdup_n(static_cast<AccType>(0.f), ExactTagType{}); + auto vec_sum_squares_h_w = wrapper::vdup_n(static_cast<AccType>(0.f), ExactTagType{}); + + // Compute S elements per iteration + int x = window.x().start(); + for(; x <= (window.x().end() - window_step_x); x += window_step_x) + { + auto vec_input_val = wrapper::vloadq(input_ptr + x); + vector_float_sum_fp16(vec_sum_h_w, vec_sum_squares_h_w, vec_input_val); + } + + auto vec2_sum_h_w = wrapper::vpadd(wrapper::vgethigh(vec_sum_h_w), wrapper::vgetlow(vec_sum_h_w)); + auto vec2_sum_squares_h_w = wrapper::vpadd(wrapper::vgethigh(vec_sum_squares_h_w), wrapper::vgetlow(vec_sum_squares_h_w)); + + vec2_sum_h_w = wrapper::vpadd(vec2_sum_h_w, vec2_sum_h_w); + vec2_sum_squares_h_w = wrapper::vpadd(vec2_sum_squares_h_w, vec2_sum_squares_h_w); + + sum_h_w += wrapper::vgetlane(vec2_sum_h_w, 0); + sum_squares_h_w += wrapper::vgetlane(vec2_sum_squares_h_w, 0); + + // Compute left-over elements + for(; x < window.x().end(); ++x) + { + const auto value = static_cast<AccType>(*(input_ptr + x)); + sum_h_w += value; + sum_squares_h_w += value * value; + } + }, + input_plane_it, output_plane_it); + + const auto mean_h_w = sum_h_w / elements_plane; + const auto var_h_w = sum_squares_h_w / elements_plane - mean_h_w * mean_h_w; + + const auto multip_h_w = gamma / std::sqrt(var_h_w + epsilon); + const auto vec_mean_h_w = wrapper::vdup_n(static_cast<AccType>(mean_h_w), ExactTagType{}); + const auto vec_multip_h_w = wrapper::vdup_n(static_cast<AccType>(multip_h_w), ExactTagType{}); + const auto vec_beta = wrapper::vdup_n(static_cast<AccType>(beta), ExactTagType{}); + + execute_window_loop(win_plane, [&](const Coordinates &) + { + auto input_ptr = reinterpret_cast<const float16_t *>(input_plane_it.ptr()); + auto output_ptr = reinterpret_cast<float16_t *>(output_plane_it.ptr()); + + // Compute S elements per iteration + int x = window.x().start(); + for(; x <= (window.x().end() - window_step_x); x += window_step_x) + { + const auto vec_val = wrapper::vloadq(input_ptr + x); + const auto normalized_vec = vector_float_norm_fp16(vec_val, vec_mean_h_w, vec_multip_h_w, vec_beta); + wrapper::vstore(output_ptr + x, normalized_vec); + } + + // Compute left-over elements + for(; x < window.x().end(); ++x) + { + const auto val = static_cast<AccType>(*(input_ptr + x)); + *(output_ptr + x) = static_cast<float16_t>((val - mean_h_w) * multip_h_w + beta); + } + }, + input_plane_it, output_plane_it); + }, + input_it); +} +} + void neon_fp16_instancenorm(ITensor *input, ITensor *output, float gamma, float beta, float epsilon, bool use_mixed_precision, const Window &window) { if(use_mixed_precision) { - return instance_normalization_nchw<float16_t, float>(input, output, gamma, beta, epsilon, window); + return instance_normalization_nchw_fp16<float>(input, output, gamma, beta, epsilon, window); } else { - return instance_normalization_nchw<float16_t>(input, output, gamma, beta, epsilon, window); + return instance_normalization_nchw_fp16<float16_t>(input, output, gamma, beta, epsilon, window); } } } // namespace cpu |