xref: /aosp_15_r20/external/ComputeLibrary/src/cpu/kernels/floor/neon/fp16.cpp (revision c217d954acce2dbc11938adb493fc0abd69584f3)
1 /*
2  * Copyright (c) 2020-2021 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 #if defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS)
25 
26 #include "src/common/utils/Validate.h"
27 #include "src/core/NEON/NEMath.h"
28 
29 #include <arm_neon.h>
30 #include <cmath>
31 #include <cstddef>
32 
33 namespace arm_compute
34 {
35 namespace cpu
36 {
37 constexpr int step = 8;
38 
fp16_neon_floor(const void * src,void * dst,int len)39 void fp16_neon_floor(const void *src, void *dst, int len)
40 {
41     ARM_COMPUTE_ASSERT_NOT_NULLPTR(src);
42     ARM_COMPUTE_ASSERT_NOT_NULLPTR(dst);
43     ARM_COMPUTE_ASSERT(len >= 0);
44 
45     auto psrc = static_cast<const __fp16 *>(src);
46     auto pdst = static_cast<__fp16 *>(dst);
47 
48     for(; len >= step; len -= step)
49     {
50         vst1q_f16(pdst, vfloorq_f16(vld1q_f16(psrc)));
51         psrc += step;
52         pdst += step;
53     }
54 
55     for(; len > 0; --len)
56     {
57         *pdst = std::floor(*psrc);
58         ++psrc;
59         ++pdst;
60     }
61 }
62 } // namespace cpu
63 } // namespace arm_compute
64 #endif /* defined(__ARM_FEATURE_FP16_VECTOR_ARITHMETIC) && defined(ENABLE_FP16_KERNELS) */
65