| 1 |
/* |
| 2 |
* // Copyright (c) Radzivon Bartoshyk 9/2025. All rights reserved. |
| 3 |
* // |
| 4 |
* // Redistribution and use in source and binary forms, with or without modification, |
| 5 |
* // are permitted provided that the following conditions are met: |
| 6 |
* // |
| 7 |
* // 1. Redistributions of source code must retain the above copyright notice, this |
| 8 |
* // list of conditions and the following disclaimer. |
| 9 |
* // |
| 10 |
* // 2. Redistributions in binary form must reproduce the above copyright notice, |
| 11 |
* // this list of conditions and the following disclaimer in the documentation |
| 12 |
* // and/or other materials provided with the distribution. |
| 13 |
* // |
| 14 |
* // 3. Neither the name of the copyright holder nor the names of its |
| 15 |
* // contributors may be used to endorse or promote products derived from |
| 16 |
* // this software without specific prior written permission. |
| 17 |
* // |
| 18 |
* // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
| 19 |
* // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 20 |
* // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE |
| 21 |
* // DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE |
| 22 |
* // FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 23 |
* // DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR |
| 24 |
* // SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER |
| 25 |
* // CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, |
| 26 |
* // OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 27 |
* // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 28 |
*/ |
| 29 |
use crate::common::{dd_fmla, f_fmla}; |
| 30 |
use crate::double_double::DoubleDouble; |
| 31 |
use crate::exponents::{EXP_REDUCE_T0, EXP_REDUCE_T1}; |
| 32 |
use crate::rounding::CpuRound; |
| 33 |
|
| 34 |
#[inline(always)] |
| 35 |
fn exp_poly(z: f64) -> DoubleDouble { |
| 36 |
/* The following is a degree-4 polynomial generated by Sollya for exp(x) |
| 37 |
over [-2^-12.905,2^-12.905] |
| 38 |
with absolute error < 2^-74.34 (see sollya/Q_1.sollya). */ |
| 39 |
const Q_1: [u64; 5] = [ |
| 40 |
0x3ff0000000000000, |
| 41 |
0x3ff0000000000000, |
| 42 |
0x3fe0000000000000, |
| 43 |
0x3fc5555555997996, |
| 44 |
0x3fa5555555849d8d, |
| 45 |
]; |
| 46 |
let mut q = dd_fmla(f64::from_bits(Q_1[4]), z, f64::from_bits(Q_1[3])); |
| 47 |
q = dd_fmla(q, z, f64::from_bits(Q_1[2])); |
| 48 |
let h0 = dd_fmla(q, z, f64::from_bits(Q_1[1])); |
| 49 |
|
| 50 |
let v1 = DoubleDouble::from_exact_mult(z, h0); |
| 51 |
DoubleDouble::f64_add(f64::from_bits(Q_1[0]), v1) |
| 52 |
} |
| 53 |
|
| 54 |
#[inline] |
| 55 |
pub(crate) fn i0_exp(r: f64) -> DoubleDouble { |
| 56 |
const INVLOG2: f64 = f64::from_bits(0x40b71547652b82fe); |
| 57 |
|
| 58 |
let k = (r * INVLOG2).cpu_round(); |
| 59 |
|
| 60 |
const LOG_2E: DoubleDouble = DoubleDouble::new( |
| 61 |
f64::from_bits(0x3d0718432a1b0e26), |
| 62 |
f64::from_bits(0x3f262e42ff000000), |
| 63 |
); |
| 64 |
|
| 65 |
let zh = f_fmla(LOG_2E.lo, k, f_fmla(-LOG_2E.hi, k, r)); |
| 66 |
|
| 67 |
let bk = unsafe { |
| 68 |
k.to_int_unchecked::<i64>() // k is already integer, this is just a conversion |
| 69 |
}; |
| 70 |
let mk = (bk >> 12) + 0x3ff; |
| 71 |
let i2 = (bk >> 6) & 0x3f; |
| 72 |
let i1 = bk & 0x3f; |
| 73 |
|
| 74 |
let t0 = DoubleDouble::from_bit_pair(EXP_REDUCE_T0[i2 as usize]); |
| 75 |
let t1 = DoubleDouble::from_bit_pair(EXP_REDUCE_T1[i1 as usize]); |
| 76 |
let mut de = DoubleDouble::quick_mult(t1, t0); |
| 77 |
let q = exp_poly(zh); |
| 78 |
de = DoubleDouble::quick_mult(de, q); |
| 79 |
|
| 80 |
let mut du = (mk as u64).wrapping_shl(52); |
| 81 |
du = f64::from_bits(du).to_bits(); |
| 82 |
DoubleDouble::quick_mult_f64(de, f64::from_bits(du)) |
| 83 |
} |
| 84 |
|
| 85 |
#[inline(always)] |
| 86 |
fn exp_poly_dd(z: DoubleDouble) -> DoubleDouble { |
| 87 |
// Generated by Sollya: |
| 88 |
// d = [-2^-12.905,2^-12.905]; |
| 89 |
// f = exp(x); |
| 90 |
// w = 1; |
| 91 |
// p = remez(f, 6, d, w); |
| 92 |
// pf = fpminimax(f, [|0,1,2,3,4,5,6|], [|1, 107...|], d, absolute, floating, 0, p); |
| 93 |
// err_p = -log2(dirtyinfnorm(pf*w-f, d)); |
| 94 |
// display = decimal; |
| 95 |
const Q_1: [(u64, u64); 7] = [ |
| 96 |
(0x0000000000000000, 0x3ff0000000000000), |
| 97 |
(0x3a20e40000000000, 0x3ff0000000000000), |
| 98 |
(0x3a04820000000000, 0x3fe0000000000000), |
| 99 |
(0xbc756423c5338a66, 0x3fc5555555555556), |
| 100 |
(0xbc5560f74db5556c, 0x3fa5555555555556), |
| 101 |
(0x3c3648eca89bc6ac, 0x3f8111111144fbee), |
| 102 |
(0xbbd53d924ae90c8c, 0x3f56c16c16ffeecc), |
| 103 |
]; |
| 104 |
let mut p = DoubleDouble::quick_mul_add( |
| 105 |
z, |
| 106 |
DoubleDouble::from_bit_pair(Q_1[6]), |
| 107 |
DoubleDouble::from_bit_pair(Q_1[5]), |
| 108 |
); |
| 109 |
p = DoubleDouble::quick_mul_add(z, p, DoubleDouble::from_bit_pair(Q_1[4])); |
| 110 |
p = DoubleDouble::quick_mul_add(z, p, DoubleDouble::from_bit_pair(Q_1[3])); |
| 111 |
p = DoubleDouble::quick_mul_add(z, p, DoubleDouble::from_bit_pair(Q_1[2])); |
| 112 |
p = DoubleDouble::quick_mul_add(z, p, DoubleDouble::from_bit_pair(Q_1[1])); |
| 113 |
DoubleDouble::quick_mul_add_f64(z, p, f64::from_bits(0x3ff0000000000000)) |
| 114 |
} |
| 115 |
|
| 116 |
#[cold] |
| 117 |
pub(crate) fn i0_exp_accurate(r: f64) -> DoubleDouble { |
| 118 |
const INVLOG2: f64 = f64::from_bits(0x40b71547652b82fe); |
| 119 |
|
| 120 |
let k = (r * INVLOG2).cpu_round(); |
| 121 |
|
| 122 |
const L2: DoubleDouble = DoubleDouble::new( |
| 123 |
f64::from_bits(0x3d0718432a1b0e26), |
| 124 |
f64::from_bits(0x3f262e42ff000000), |
| 125 |
); |
| 126 |
const L2LL: f64 = f64::from_bits(0x3999ff0342542fc3); |
| 127 |
let dx = f_fmla(-L2.hi, k, r); |
| 128 |
let dx_dd = DoubleDouble::quick_mult_f64(DoubleDouble::new(L2LL, L2.lo), k); |
| 129 |
let dz = DoubleDouble::full_add_f64(dx_dd, dx); |
| 130 |
|
| 131 |
let bk = unsafe { |
| 132 |
k.to_int_unchecked::<i64>() // k is already integer, this is just a conversion |
| 133 |
}; |
| 134 |
let mk = (bk >> 12) + 0x3ff; |
| 135 |
let i2 = (bk >> 6) & 0x3f; |
| 136 |
let i1 = bk & 0x3f; |
| 137 |
|
| 138 |
let t0 = DoubleDouble::from_bit_pair(EXP_REDUCE_T0[i2 as usize]); |
| 139 |
let t1 = DoubleDouble::from_bit_pair(EXP_REDUCE_T1[i1 as usize]); |
| 140 |
let mut de = DoubleDouble::quick_mult(t1, t0); |
| 141 |
let q = exp_poly_dd(dz); |
| 142 |
de = DoubleDouble::quick_mult(de, q); |
| 143 |
|
| 144 |
let mut du = (mk as u64).wrapping_shl(52); |
| 145 |
du = f64::from_bits(du).to_bits(); |
| 146 |
DoubleDouble::quick_mult_f64(de, f64::from_bits(du)) |
| 147 |
} |
| 148 |
|
| 149 |
#[cfg(test)] |
| 150 |
mod tests { |
| 151 |
use super::*; |
| 152 |
#[test] |
| 153 |
fn test_i0() { |
| 154 |
assert_eq!(i0_exp(0.5).to_f64(), 1.6487212707001282); |
| 155 |
assert_eq!(i0_exp_accurate(0.5).to_f64(), 1.6487212707001282); |
| 156 |
} |
| 157 |
} |