| 1 | //===-- Common utils for exp10f16 -------------------------------*- C++ -*-===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #ifndef LLVM_LIBC_SRC___SUPPORT_MATH_EXP10F16_UTILS_H |
| 10 | #define LLVM_LIBC_SRC___SUPPORT_MATH_EXP10F16_UTILS_H |
| 11 | |
| 12 | #include "include/llvm-libc-macros/float16-macros.h" |
| 13 | |
| 14 | #ifdef LIBC_TYPES_HAS_FLOAT16 |
| 15 | |
| 16 | #include "exp10_float16_constants.h" |
| 17 | #include "expf16_utils.h" |
| 18 | #include "src/__support/FPUtil/FPBits.h" |
| 19 | |
| 20 | namespace LIBC_NAMESPACE_DECL { |
| 21 | |
| 22 | LIBC_INLINE ExpRangeReduction exp10_range_reduction(float16 x) { |
| 23 | // For -8 < x < 5, to compute 10^x, we perform the following range reduction: |
| 24 | // find hi, mid, lo, such that: |
| 25 | // x = (hi + mid) * log2(10) + lo, in which |
| 26 | // hi is an integer, |
| 27 | // mid * 2^3 is an integer, |
| 28 | // -2^(-4) <= lo < 2^(-4). |
| 29 | // In particular, |
| 30 | // hi + mid = round(x * 2^3) * 2^(-3). |
| 31 | // Then, |
| 32 | // 10^x = 10^(hi + mid + lo) = 2^((hi + mid) * log2(10)) + 10^lo |
| 33 | // We store 2^mid in the lookup table EXP2_MID_BITS, and compute 2^hi * 2^mid |
| 34 | // by adding hi to the exponent field of 2^mid. 10^lo is computed using a |
| 35 | // degree-4 minimax polynomial generated by Sollya. |
| 36 | |
| 37 | float xf = x; |
| 38 | float kf = fputil::nearest_integer(x: xf * (LOG2F_10 * 0x1.0p+3f)); |
| 39 | int x_hi_mid = static_cast<int>(kf); |
| 40 | unsigned x_hi = static_cast<unsigned>(x_hi_mid) >> 3; |
| 41 | unsigned x_mid = static_cast<unsigned>(x_hi_mid) & 0x7; |
| 42 | // lo = x - (hi + mid) = round(x * 2^3 * log2(10)) * log10(2) * (-2^(-3)) + x |
| 43 | float lo = fputil::multiply_add(x: kf, y: LOG10F_2 * -0x1.0p-3f, z: xf); |
| 44 | |
| 45 | uint32_t exp2_hi_mid_bits = |
| 46 | EXP2_MID_BITS[x_mid] + |
| 47 | static_cast<uint32_t>(x_hi << fputil::FPBits<float>::FRACTION_LEN); |
| 48 | float exp2_hi_mid = fputil::FPBits<float>(exp2_hi_mid_bits).get_val(); |
| 49 | // Degree-4 minimax polynomial generated by Sollya with the following |
| 50 | // commands: |
| 51 | // > display = hexadecimal; |
| 52 | // > P = fpminimax((10^x - 1)/x, 3, [|SG...|], [-2^-4, 2^-4]); |
| 53 | // > 1 + x * P; |
| 54 | float exp10_lo = fputil::polyeval(x: lo, a0: 0x1p+0f, a: 0x1.26bb14p+1f, a: 0x1.53526p+1f, |
| 55 | a: 0x1.04b434p+1f, a: 0x1.2bcf9ep+0f); |
| 56 | return {.exp_hi_mid: exp2_hi_mid, .exp_lo: exp10_lo}; |
| 57 | } |
| 58 | |
| 59 | } // namespace LIBC_NAMESPACE_DECL |
| 60 | |
| 61 | #endif // LIBC_TYPES_HAS_FLOAT16 |
| 62 | |
| 63 | #endif // LLVM_LIBC_SRC___SUPPORT_MATH_EXP10F16_UTILS_H |
| 64 | |