Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 1 | /* |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 2 | * Copyright (c) 2020-2021 Arm Limited. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 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 ARM_COMPUTE_SVEMATH_H |
| 25 | #define ARM_COMPUTE_SVEMATH_H |
| 26 | |
Michalis Spyrou | 20fca52 | 2021-06-07 14:23:57 +0100 | [diff] [blame] | 27 | #if defined(ARM_COMPUTE_ENABLE_SVE) |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 28 | #include "src/core/NEON/wrapper/intrinsics/svcvt.h" |
| 29 | #include "src/core/NEON/wrapper/intrinsics/svdup_n.h" |
| 30 | #include "src/core/NEON/wrapper/intrinsics/svreinterpret.h" |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 31 | #include <arm_sve.h> |
| 32 | #include <array> |
| 33 | |
| 34 | namespace arm_compute |
| 35 | { |
| 36 | /** Calculate exponent. |
| 37 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 38 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 39 | * @param[in] val Input vector value in F32 format. |
| 40 | * |
| 41 | * @return The calculated exponent. |
| 42 | */ |
| 43 | svfloat32_t svexp_f32_z(svbool_t pg, svfloat32_t val); |
| 44 | |
| 45 | /** Calculate reciprocal. |
| 46 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 47 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 48 | * @param[in] x Input value. |
| 49 | * |
| 50 | * @return The calculated reciprocal. |
| 51 | */ |
| 52 | svfloat32_t svinv_f32_z(svbool_t pg, svfloat32_t x); |
| 53 | |
| 54 | /** Calculate logarithm |
| 55 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 56 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 57 | * @param[in] x Input vector value in F32 format. |
| 58 | * |
| 59 | * @return The calculated logarithm. |
| 60 | */ |
| 61 | svfloat32_t svlog_f32_z(svbool_t pg, svfloat32_t x); |
| 62 | |
| 63 | /** Calculate hyperbolic tangent. |
| 64 | * |
| 65 | * tanh(x) = (e^2x - 1)/(e^2x + 1) |
| 66 | * |
| 67 | * @note We clamp x to [-5,5] to avoid overflowing issues. |
| 68 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 69 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 70 | * @param[in] val Input vector value in F32 format. |
| 71 | * |
| 72 | * @return The calculated Hyperbolic Tangent. |
| 73 | */ |
| 74 | svfloat32_t svtanh_f32_z(svbool_t pg, svfloat32_t val); |
| 75 | |
| 76 | /** Calculate hyperbolic tangent. |
| 77 | * |
| 78 | * tanh(x) = (e^2x - 1)/(e^2x + 1) |
| 79 | * |
| 80 | * @note We clamp x to [-5,5] to avoid overflowing issues. |
| 81 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 82 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 83 | * @param[in] val Input vector value in F16 format. |
| 84 | * |
| 85 | * @return The calculated Hyperbolic Tangent. |
| 86 | */ |
| 87 | svfloat16_t svtanh_f16_z(svbool_t pg, svfloat16_t val); |
| 88 | |
| 89 | /** Calculate exponential |
| 90 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 91 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 92 | * @param[in] x Input vector value in F16 format. |
| 93 | * |
| 94 | * @return The calculated exponent. |
| 95 | */ |
| 96 | svfloat16_t svexp_f16_z(svbool_t pg, svfloat16_t x); |
| 97 | |
| 98 | /** Calculate reciprocal. |
| 99 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 100 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 101 | * @param[in] x Input value. |
| 102 | * |
| 103 | * @return The calculated reciprocal. |
| 104 | */ |
| 105 | svfloat16_t svinv_f16_z(svbool_t pg, svfloat16_t x); |
| 106 | |
| 107 | /** Calculate logarithm |
| 108 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 109 | * @param[in] pg Input predicate. |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 110 | * @param[in] x Input vector value in F32 format. |
| 111 | * |
| 112 | * @return The calculated logarithm. |
| 113 | */ |
| 114 | svfloat16_t svlog_f16_z(svbool_t pg, svfloat16_t x); |
| 115 | |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 116 | /** Calculate inverse square root. |
| 117 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 118 | * @param[in] pg Input predicate. |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 119 | * @param[in] val Input value. |
| 120 | * |
| 121 | * @return The calculated inverse square root. |
| 122 | */ |
| 123 | template <typename VectorType> |
| 124 | inline VectorType svinvsqrt(svbool_t pg, VectorType val) |
| 125 | { |
| 126 | auto sqrt_reciprocal = svrsqrte(val); |
| 127 | sqrt_reciprocal = svmul_z(pg, svrsqrts(svmul_z(pg, val, sqrt_reciprocal), sqrt_reciprocal), sqrt_reciprocal); |
| 128 | sqrt_reciprocal = svmul_z(pg, svrsqrts(svmul_z(pg, val, sqrt_reciprocal), sqrt_reciprocal), sqrt_reciprocal); |
| 129 | return sqrt_reciprocal; |
| 130 | } |
| 131 | |
| 132 | /** Calculate sine. |
| 133 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 134 | * @param[in] pg Input predicate. |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 135 | * @param[in] val Input vector value in radians, F32 format. |
| 136 | * |
| 137 | * @return The calculated sine. |
| 138 | */ |
| 139 | svfloat32_t svsin_f32_z(svbool_t pg, svfloat32_t val); |
| 140 | |
| 141 | /** Calculate sine. |
| 142 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 143 | * @param[in] pg Input predicate. |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 144 | * @param[in] val Input vector value in radians, F16 format. |
| 145 | * |
| 146 | * @return The calculated sine. |
| 147 | */ |
| 148 | svfloat16_t svsin_f16_z(svbool_t pg, svfloat16_t val); |
| 149 | |
| 150 | /** Calculate n power of a number. |
| 151 | * |
| 152 | * pow(x,n) = e^(n*log(x)) |
| 153 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 154 | * @param[in] pg Input predicate. |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 155 | * @param[in] a Input vector value in F32 format. |
| 156 | * @param[in] b Powers to raise the input to. |
| 157 | * |
| 158 | * @return The calculated power. |
| 159 | */ |
| 160 | svfloat32_t svpow_f32_z(svbool_t pg, svfloat32_t a, svfloat32_t b); |
| 161 | |
| 162 | /** Calculate n power of a number. |
| 163 | * |
| 164 | * pow(x,n) = e^(n*log(x)) |
| 165 | * |
Michalis Spyrou | a3c9a3b | 2020-12-08 21:02:16 +0000 | [diff] [blame] | 166 | * @param[in] pg Input predicate. |
Sang-Hoon Park | 0870db4 | 2020-12-08 18:42:19 +0000 | [diff] [blame] | 167 | * @param[in] a Input vector value in F16 format. |
| 168 | * @param[in] b Powers to raise the input to. |
| 169 | * |
| 170 | * @return The calculated power. |
| 171 | */ |
| 172 | svfloat16_t svpow_f16_z(svbool_t pg, svfloat16_t a, svfloat16_t b); |
| 173 | |
Sang-Hoon Park | dcf3c7e | 2021-03-04 17:03:46 +0000 | [diff] [blame] | 174 | /** Convert and pack four 32-bit float vectors into an 8-bit integer vector |
| 175 | * |
| 176 | * @param[in] in_0 The first float vector |
| 177 | * @param[in] in_1 The second float vector |
| 178 | * @param[in] in_2 The third float vector |
| 179 | * @param[in] in_3 The fourth float vector |
| 180 | * |
| 181 | * @return The converted integer vector |
| 182 | */ |
| 183 | template <typename int_vec_type> |
| 184 | int_vec_type convert_float_to_int(const svfloat32_t &in_0, const svfloat32_t &in_1, const svfloat32_t &in_2, const svfloat32_t &in_3); |
| 185 | |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 186 | } // namespace arm_compute |
| 187 | #include "src/core/NEON/SVEMath.inl" |
Michalis Spyrou | 20fca52 | 2021-06-07 14:23:57 +0100 | [diff] [blame] | 188 | #endif /* defined(ARM_COMPUTE_ENABLE_SVE) */ |
Michalis Spyrou | aa51a5b | 2020-11-22 00:49:42 +0000 | [diff] [blame] | 189 | #endif /* ARM_COMPUTE_SVEMATH_H */ |