xref: /aosp_15_r20/external/ComputeLibrary/src/cpu/kernels/instancenorm/generic/neon/impl.h (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2022 Arm Limited.
3  *
4  * SPDX-License-Identifier: MIT
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to
8  * deal in the Software without restriction, including without limitation the
9  * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
10  * sell copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in all
14  * copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22  * SOFTWARE.
23  */
24 #ifndef SRC_CORE_SVE_KERNELS_INSTANCENORM_IMPL_H
25 #define SRC_CORE_SVE_KERNELS_INSTANCENORM_IMPL_H
26 #include "arm_compute/core/Helpers.h"
27 namespace arm_compute
28 {
29 namespace cpu
30 {
31 template <typename T, typename AccType = T>
32 void instance_normalization_nchw(ITensor *input, ITensor *output, float gamma, float beta, float epsilon, const Window &window);
33 
34 template <typename InputType, typename AccType = InputType>
35 void vector_float_sum(AccType &result, AccType &result_square, const InputType &inputs);
36 
37 template <typename InputType, typename AccType = InputType>
38 InputType vector_float_norm(const InputType &inputs, const AccType &vec_mean, const AccType &vec_multip, const AccType &vec_beta);
39 
40 #if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS)
41 template <>
42 inline void vector_float_sum(float32x4_t &result, float32x4_t &result_square, const float16x8_t &inputs);
43 
44 template <>
45 inline float16x8_t vector_float_norm(const float16x8_t &inputs, const float32x4_t &vec_mean, const float32x4_t &vec_multip, const float32x4_t &vec_beta);
46 #endif //defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS)
47 
48 } // namespace cpu
49 } // namespace arm_compute
50 #endif //define SRC_CORE_SVE_KERNELS_INSTANCENORM_IMPL_H
51