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 #include "arm_compute/core/Helpers.h"
25 #include "src/cpu/kernels/elementwise_binary/generic/neon/impl.h"
26 namespace arm_compute
27 {
28 namespace cpu
29 {
30 template <ArithmeticOperation op>
neon_s32_elementwise_binary(const ITensor * in1,const ITensor * in2,ITensor * out,const Window & window)31 void neon_s32_elementwise_binary(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window)
32 {
33 return elementwise_arithm_op<op, typename wrapper::traits::neon_vector<int32_t, 4>>(in1, in2, out, window);
34 }
35
36 template void neon_s32_elementwise_binary<ArithmeticOperation::ADD>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
37 template void neon_s32_elementwise_binary<ArithmeticOperation::SUB>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
38 template void neon_s32_elementwise_binary<ArithmeticOperation::DIV>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
39 template void neon_s32_elementwise_binary<ArithmeticOperation::MIN>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
40 template void neon_s32_elementwise_binary<ArithmeticOperation::MAX>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
41 template void neon_s32_elementwise_binary<ArithmeticOperation::SQUARED_DIFF>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
42 template void neon_s32_elementwise_binary<ArithmeticOperation::POWER>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
43 template void neon_s32_elementwise_binary<ArithmeticOperation::PRELU>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
44
45 template <ArithmeticOperation op>
neon_s16_elementwise_binary(const ITensor * in1,const ITensor * in2,ITensor * out,const Window & window)46 void neon_s16_elementwise_binary(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window)
47 {
48 return elementwise_arithm_op<op, typename wrapper::traits::neon_vector<int16_t, 8>>(in1, in2, out, window);
49 }
50 template void neon_s16_elementwise_binary<ArithmeticOperation::ADD>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
51 template void neon_s16_elementwise_binary<ArithmeticOperation::SUB>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
52 template void neon_s16_elementwise_binary<ArithmeticOperation::DIV>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
53 template void neon_s16_elementwise_binary<ArithmeticOperation::MIN>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
54 template void neon_s16_elementwise_binary<ArithmeticOperation::MAX>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
55 template void neon_s16_elementwise_binary<ArithmeticOperation::SQUARED_DIFF>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
56 template void neon_s16_elementwise_binary<ArithmeticOperation::POWER>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
57 template void neon_s16_elementwise_binary<ArithmeticOperation::PRELU>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
58
59 template <ComparisonOperation op>
neon_u8_comparison_elementwise_binary(const ITensor * in1,const ITensor * in2,ITensor * out,const Window & window)60 void neon_u8_comparison_elementwise_binary(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window)
61 {
62 return elementwise_comp_op_8<op, uint8_t, uint8x16_t>(in1, in2, out, window);
63 }
64 template void neon_u8_comparison_elementwise_binary<ComparisonOperation::Equal>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
65 template void neon_u8_comparison_elementwise_binary<ComparisonOperation::NotEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
66 template void neon_u8_comparison_elementwise_binary<ComparisonOperation::Greater>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
67 template void neon_u8_comparison_elementwise_binary<ComparisonOperation::GreaterEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
68 template void neon_u8_comparison_elementwise_binary<ComparisonOperation::Less>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
69 template void neon_u8_comparison_elementwise_binary<ComparisonOperation::LessEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
70
71 template <ComparisonOperation op>
neon_s16_comparison_elementwise_binary(const ITensor * in1,const ITensor * in2,ITensor * out,const Window & window)72 void neon_s16_comparison_elementwise_binary(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window)
73 {
74 return elementwise_comp_op_16<op, int16_t, int16x8_t>(in1, in2, out, window);
75 }
76 template void neon_s16_comparison_elementwise_binary<ComparisonOperation::Equal>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
77 template void neon_s16_comparison_elementwise_binary<ComparisonOperation::NotEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
78 template void neon_s16_comparison_elementwise_binary<ComparisonOperation::Greater>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
79 template void neon_s16_comparison_elementwise_binary<ComparisonOperation::GreaterEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
80 template void neon_s16_comparison_elementwise_binary<ComparisonOperation::Less>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
81 template void neon_s16_comparison_elementwise_binary<ComparisonOperation::LessEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
82
83 template <ComparisonOperation op>
neon_s32_comparison_elementwise_binary(const ITensor * in1,const ITensor * in2,ITensor * out,const Window & window)84 void neon_s32_comparison_elementwise_binary(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window)
85 {
86 return elementwise_comp_op_32<op, int32_t, int32x4_t>(in1, in2, out, window);
87 }
88 template void neon_s32_comparison_elementwise_binary<ComparisonOperation::Equal>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
89 template void neon_s32_comparison_elementwise_binary<ComparisonOperation::NotEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
90 template void neon_s32_comparison_elementwise_binary<ComparisonOperation::Greater>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
91 template void neon_s32_comparison_elementwise_binary<ComparisonOperation::GreaterEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
92 template void neon_s32_comparison_elementwise_binary<ComparisonOperation::Less>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
93 template void neon_s32_comparison_elementwise_binary<ComparisonOperation::LessEqual>(const ITensor *in1, const ITensor *in2, ITensor *out, const Window &window);
94 }
95 } // namespace arm_compute
96