diff options
Diffstat (limited to 'src/cpu/kernels/activation')
| -rw-r--r-- | src/cpu/kernels/activation/generic/sve/fp16.cpp | 29 | ||||
| -rw-r--r-- | src/cpu/kernels/activation/list.h | 9 |
2 files changed, 33 insertions, 5 deletions
diff --git a/src/cpu/kernels/activation/generic/sve/fp16.cpp b/src/cpu/kernels/activation/generic/sve/fp16.cpp index 97399e01e0..19d9126556 100644 --- a/src/cpu/kernels/activation/generic/sve/fp16.cpp +++ b/src/cpu/kernels/activation/generic/sve/fp16.cpp @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023 Arm Limited. + * Copyright (c) 2020-2024 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -30,6 +30,7 @@ #include "arm_compute/function_info/ActivationLayerInfo.h" #include "src/core/NEON/SVEMath.h" +#include "src/cpu/kernels/lut/list.h" #include <arm_sve.h> #include <cmath> @@ -141,6 +142,32 @@ void sve_fp16_activation(const ITensor *src, ITensor *dst, const ActivationLayer }, input, output); } + +void sve_fp16_activation_lut(const ITensor *src, + ITensor *dst, + const ActivationLayerInfo &act_info, + const Window &window) +{ + ARM_COMPUTE_ERROR_ON(src->info()->data_type() != DataType::F16); + const auto window_start_x = window.x().start(); + const auto window_end_x = window.x().end(); + const auto size = window_end_x - window_start_x; + Window win_collapsed = window.collapse_if_possible(window, Window::DimZ); + win_collapsed.set(Window::DimX, Window::Dimension(0, 1, 1)); + + Iterator input(src, win_collapsed); + Iterator output(dst, win_collapsed); + execute_window_loop( + win_collapsed, + [&](const Coordinates &) + { + const auto input_ptr = reinterpret_cast<const uint16_t *>(input.ptr()); + auto output_ptr = reinterpret_cast<uint16_t *>(output.ptr()); + lut_u16_sve(reinterpret_cast<const uint16_t *>(act_info.lut_fp16().data()), 1U /* num_strings (UNUSED) */, + size, input_ptr + window_start_x, output_ptr + window_start_x); + }, + input, output); +} } // namespace cpu } // namespace arm_compute #endif /* defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */ diff --git a/src/cpu/kernels/activation/list.h b/src/cpu/kernels/activation/list.h index 6550ddfeca..8c24adc3fe 100644 --- a/src/cpu/kernels/activation/list.h +++ b/src/cpu/kernels/activation/list.h @@ -1,5 +1,5 @@ /* - * Copyright (c) 2020-2023 Arm Limited. + * Copyright (c) 2020-2024 Arm Limited. * * SPDX-License-Identifier: MIT * @@ -21,8 +21,8 @@ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -#ifndef SRC_CORE_NEON_KERNELS_ACTIVATION_LIST_H -#define SRC_CORE_NEON_KERNELS_ACTIVATION_LIST_H +#ifndef ACL_SRC_CPU_KERNELS_ACTIVATION_LIST_H +#define ACL_SRC_CPU_KERNELS_ACTIVATION_LIST_H namespace arm_compute { @@ -42,6 +42,7 @@ DECLARE_ACTIVATION_KERNEL(sve2_qasymm8_signed_activation); DECLARE_ACTIVATION_KERNEL(neon_qsymm16_activation); DECLARE_ACTIVATION_KERNEL(sve2_qsymm16_activation); DECLARE_ACTIVATION_KERNEL(sve_fp16_activation); +DECLARE_ACTIVATION_KERNEL(sve_fp16_activation_lut); DECLARE_ACTIVATION_KERNEL(sve_fp32_activation); DECLARE_ACTIVATION_KERNEL(neon_fp16_activation); DECLARE_ACTIVATION_KERNEL(neon_fp32_activation); @@ -50,4 +51,4 @@ DECLARE_ACTIVATION_KERNEL(neon_fp32_activation); } // namespace cpu } // namespace arm_compute -#endif /* SRC_CORE_NEON_KERNELS_ACTIVATION_LIST_H */ +#endif // ACL_SRC_CPU_KERNELS_ACTIVATION_LIST_H |