diff options
Diffstat (limited to 'src/cpu/kernels/range/generic')
-rw-r--r-- | src/cpu/kernels/range/generic/neon/impl.cpp | 93 | ||||
-rw-r--r-- | src/cpu/kernels/range/generic/neon/impl.h | 60 |
2 files changed, 52 insertions, 101 deletions
diff --git a/src/cpu/kernels/range/generic/neon/impl.cpp b/src/cpu/kernels/range/generic/neon/impl.cpp deleted file mode 100644 index f91251c914..0000000000 --- a/src/cpu/kernels/range/generic/neon/impl.cpp +++ /dev/null @@ -1,93 +0,0 @@ -/* - * Copyright (c) 2021 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 "src/cpu/kernels/range/generic/neon/impl.h" - -#include "arm_compute/core/Helpers.h" -#include "arm_compute/core/TensorInfo.h" -#include "src/core/NEON/wrapper/wrapper.h" -#include "src/core/common/Registrars.h" - -namespace arm_compute -{ -namespace cpu -{ -template <typename T> -void neon_range_function(ITensor *output, float start, float step, const Window &window) -{ - /** SIMD vector tag type. */ - using ExactTagType = typename wrapper::traits::neon_bitvector<T, wrapper::traits::BitWidth::W128>::tag_type; - - const auto step_vec = wrapper::vdup_n(static_cast<T>(step), ExactTagType{}); - const auto start_vec = wrapper::vdup_n(static_cast<T>(start), ExactTagType{}); - auto id_vec = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); - - const auto window_start_x = static_cast<int>(window.x().start()); - const auto window_end_x = static_cast<int>(window.x().end()); - const int window_step_x = 16 / sizeof(T); - - Window win{ window }; - win.set(Window::DimX, Window::Dimension(0, 1, 1)); - Iterator output_it(output, win); - - execute_window_loop(win, [&](const Coordinates &) - { - int x = window_start_x; - const auto out_ptr = reinterpret_cast<T *>(output_it.ptr()); - for(; x <= (window_end_x - window_step_x); x += window_step_x) - { - for(int count = 0; count < window_step_x; ++count) - { - id_vec = wrapper::vsetlane(static_cast<T>(x + count), id_vec, count); - } - - // start + step * id - const auto res_vec = wrapper::vmla(start_vec, id_vec, step_vec); - wrapper::vstore(out_ptr + x, res_vec); - } - - // Compute left-over elements - for(; x < window_end_x; ++x) - { - const auto res = start + x * step; - *(out_ptr + x) = res; - } - - }, - output_it); -} - -template void neon_range_function<uint8_t>(ITensor *output, float start, float step, const Window &window); -template void neon_range_function<uint16_t>(ITensor *output, float start, float step, const Window &window); -template void neon_range_function<uint32_t>(ITensor *output, float start, float step, const Window &window); -template void neon_range_function<int8_t>(ITensor *output, float start, float step, const Window &window); -template void neon_range_function<int16_t>(ITensor *output, float start, float step, const Window &window); -template void neon_range_function<int32_t>(ITensor *output, float start, float step, const Window &window); -template void neon_range_function<float32_t>(ITensor *output, float start, float step, const Window &window); - -#if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) -template void neon_range_function<float16_t>(ITensor *output, float start, float step, const Window &window); -#endif /* defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */ - -} // namespace cpu -} // namespace arm_compute diff --git a/src/cpu/kernels/range/generic/neon/impl.h b/src/cpu/kernels/range/generic/neon/impl.h index 7ac2fc9c11..62144e6776 100644 --- a/src/cpu/kernels/range/generic/neon/impl.h +++ b/src/cpu/kernels/range/generic/neon/impl.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2021 Arm Limited. + * Copyright (c) 2021, 2023 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -21,18 +21,62 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -#ifndef SRC_CORE_NEON_KERNELS_RANGE_IMPL_H -#define SRC_CORE_NEON_KERNELS_RANGE_IMPL_H +#ifndef ACL_SRC_CPU_KERNELS_RANGE_GENERIC_NEON_IMPL_H +#define ACL_SRC_CPU_KERNELS_RANGE_GENERIC_NEON_IMPL_H + +#include "arm_compute/core/Helpers.h" +#include "arm_compute/core/TensorInfo.h" +#include "src/core/NEON/wrapper/wrapper.h" +#include "src/core/common/Registrars.h" namespace arm_compute { -class ITensor; -class Window; - namespace cpu { template <typename T> -void neon_range_function(ITensor *output, float start, float step, const Window &window); +void neon_range_function(ITensor *output, float start, float step, const Window &window) +{ + /** SIMD vector tag type. */ + using ExactTagType = typename wrapper::traits::neon_bitvector<T, wrapper::traits::BitWidth::W128>::tag_type; + + const auto step_vec = wrapper::vdup_n(static_cast<T>(step), ExactTagType{}); + const auto start_vec = wrapper::vdup_n(static_cast<T>(start), ExactTagType{}); + auto id_vec = wrapper::vdup_n(static_cast<T>(0.f), ExactTagType{}); + + const auto window_start_x = static_cast<int>(window.x().start()); + const auto window_end_x = static_cast<int>(window.x().end()); + const int window_step_x = 16 / sizeof(T); + + Window win{ window }; + win.set(Window::DimX, Window::Dimension(0, 1, 1)); + Iterator output_it(output, win); + + execute_window_loop(win, [&](const Coordinates &) + { + int x = window_start_x; + const auto out_ptr = reinterpret_cast<T *>(output_it.ptr()); + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + for(int count = 0; count < window_step_x; ++count) + { + id_vec = wrapper::vsetlane(static_cast<T>(x + count), id_vec, count); + } + + // start + step * id + const auto res_vec = wrapper::vmla(start_vec, id_vec, step_vec); + wrapper::vstore(out_ptr + x, res_vec); + } + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + const auto res = start + x * step; + *(out_ptr + x) = res; + } + + }, + output_it); +} } // namespace cpu } // namespace arm_compute -#endif //SRC_CORE_NEON_KERNELS_RANGE_IMPL_H +#endif // ACL_SRC_CPU_KERNELS_RANGE_GENERIC_NEON_IMPL_H |