| 1 | /* origin: FreeBSD /usr/src/lib/msun/src/e_exp.c */ |
| 2 | /* |
| 3 | * ==================================================== |
| 4 | * Copyright (C) 2004 by Sun Microsystems, Inc. All rights reserved. |
| 5 | * |
| 6 | * Permission to use, copy, modify, and distribute this |
| 7 | * software is freely granted, provided that this notice |
| 8 | * is preserved. |
| 9 | * ==================================================== |
| 10 | */ |
| 11 | /* exp(x) |
| 12 | * Returns the exponential of x. |
| 13 | * |
| 14 | * Method |
| 15 | * 1. Argument reduction: |
| 16 | * Reduce x to an r so that |r| <= 0.5*ln2 ~ 0.34658. |
| 17 | * Given x, find r and integer k such that |
| 18 | * |
| 19 | * x = k*ln2 + r, |r| <= 0.5*ln2. |
| 20 | * |
| 21 | * Here r will be represented as r = hi-lo for better |
| 22 | * accuracy. |
| 23 | * |
| 24 | * 2. Approximation of exp(r) by a special rational function on |
| 25 | * the interval [0,0.34658]: |
| 26 | * Write |
| 27 | * R(r**2) = r*(exp(r)+1)/(exp(r)-1) = 2 + r*r/6 - r**4/360 + ... |
| 28 | * We use a special Remez algorithm on [0,0.34658] to generate |
| 29 | * a polynomial of degree 5 to approximate R. The maximum error |
| 30 | * of this polynomial approximation is bounded by 2**-59. In |
| 31 | * other words, |
| 32 | * R(z) ~ 2.0 + P1*z + P2*z**2 + P3*z**3 + P4*z**4 + P5*z**5 |
| 33 | * (where z=r*r, and the values of P1 to P5 are listed below) |
| 34 | * and |
| 35 | * | 5 | -59 |
| 36 | * | 2.0+P1*z+...+P5*z - R(z) | <= 2 |
| 37 | * | | |
| 38 | * The computation of exp(r) thus becomes |
| 39 | * 2*r |
| 40 | * exp(r) = 1 + ---------- |
| 41 | * R(r) - r |
| 42 | * r*c(r) |
| 43 | * = 1 + r + ----------- (for better accuracy) |
| 44 | * 2 - c(r) |
| 45 | * where |
| 46 | * 2 4 10 |
| 47 | * c(r) = r - (P1*r + P2*r + ... + P5*r ). |
| 48 | * |
| 49 | * 3. Scale back to obtain exp(x): |
| 50 | * From step 1, we have |
| 51 | * exp(x) = 2^k * exp(r) |
| 52 | * |
| 53 | * Special cases: |
| 54 | * exp(INF) is INF, exp(NaN) is NaN; |
| 55 | * exp(-INF) is 0, and |
| 56 | * for finite argument, only exp(0)=1 is exact. |
| 57 | * |
| 58 | * Accuracy: |
| 59 | * according to an error analysis, the error is always less than |
| 60 | * 1 ulp (unit in the last place). |
| 61 | * |
| 62 | * Misc. info. |
| 63 | * For IEEE double |
| 64 | * if x > 709.782712893383973096 then exp(x) overflows |
| 65 | * if x < -745.133219101941108420 then exp(x) underflows |
| 66 | */ |
| 67 | |
| 68 | use super::scalbn; |
| 69 | |
| 70 | const HALF: [f64; 2] = [0.5, -0.5]; |
| 71 | const LN2HI: f64 = 6.93147180369123816490e-01; /* 0x3fe62e42, 0xfee00000 */ |
| 72 | const LN2LO: f64 = 1.90821492927058770002e-10; /* 0x3dea39ef, 0x35793c76 */ |
| 73 | const INVLN2: f64 = 1.44269504088896338700e+00; /* 0x3ff71547, 0x652b82fe */ |
| 74 | const P1: f64 = 1.66666666666666019037e-01; /* 0x3FC55555, 0x5555553E */ |
| 75 | const P2: f64 = -2.77777777770155933842e-03; /* 0xBF66C16C, 0x16BEBD93 */ |
| 76 | const P3: f64 = 6.61375632143793436117e-05; /* 0x3F11566A, 0xAF25DE2C */ |
| 77 | const P4: f64 = -1.65339022054652515390e-06; /* 0xBEBBBD41, 0xC5D26BF1 */ |
| 78 | const P5: f64 = 4.13813679705723846039e-08; /* 0x3E663769, 0x72BEA4D0 */ |
| 79 | |
| 80 | /// Exponential, base *e* (f64) |
| 81 | /// |
| 82 | /// Calculate the exponential of `x`, that is, *e* raised to the power `x` |
| 83 | /// (where *e* is the base of the natural system of logarithms, approximately 2.71828). |
| 84 | #[cfg_attr (all(test, assert_no_panic), no_panic::no_panic)] |
| 85 | pub fn exp(mut x: f64) -> f64 { |
| 86 | let x1p1023 = f64::from_bits(0x7fe0000000000000); // 0x1p1023 === 2 ^ 1023 |
| 87 | let x1p_149 = f64::from_bits(0x36a0000000000000); // 0x1p-149 === 2 ^ -149 |
| 88 | |
| 89 | let hi: f64; |
| 90 | let lo: f64; |
| 91 | let c: f64; |
| 92 | let xx: f64; |
| 93 | let y: f64; |
| 94 | let k: i32; |
| 95 | let sign: i32; |
| 96 | let mut hx: u32; |
| 97 | |
| 98 | hx = (x.to_bits() >> 32) as u32; |
| 99 | sign = (hx >> 31) as i32; |
| 100 | hx &= 0x7fffffff; /* high word of |x| */ |
| 101 | |
| 102 | /* special cases */ |
| 103 | if hx >= 0x4086232b { |
| 104 | /* if |x| >= 708.39... */ |
| 105 | if x.is_nan() { |
| 106 | return x; |
| 107 | } |
| 108 | if x > 709.782712893383973096 { |
| 109 | /* overflow if x!=inf */ |
| 110 | x *= x1p1023; |
| 111 | return x; |
| 112 | } |
| 113 | if x < -708.39641853226410622 { |
| 114 | /* underflow if x!=-inf */ |
| 115 | force_eval!((-x1p_149 / x) as f32); |
| 116 | if x < -745.13321910194110842 { |
| 117 | return 0.; |
| 118 | } |
| 119 | } |
| 120 | } |
| 121 | |
| 122 | /* argument reduction */ |
| 123 | if hx > 0x3fd62e42 { |
| 124 | /* if |x| > 0.5 ln2 */ |
| 125 | if hx >= 0x3ff0a2b2 { |
| 126 | /* if |x| >= 1.5 ln2 */ |
| 127 | k = (INVLN2 * x + i!(HALF, sign as usize)) as i32; |
| 128 | } else { |
| 129 | k = 1 - sign - sign; |
| 130 | } |
| 131 | hi = x - k as f64 * LN2HI; /* k*ln2hi is exact here */ |
| 132 | lo = k as f64 * LN2LO; |
| 133 | x = hi - lo; |
| 134 | } else if hx > 0x3e300000 { |
| 135 | /* if |x| > 2**-28 */ |
| 136 | k = 0; |
| 137 | hi = x; |
| 138 | lo = 0.; |
| 139 | } else { |
| 140 | /* inexact if x!=0 */ |
| 141 | force_eval!(x1p1023 + x); |
| 142 | return 1. + x; |
| 143 | } |
| 144 | |
| 145 | /* x is now in primary range */ |
| 146 | xx = x * x; |
| 147 | c = x - xx * (P1 + xx * (P2 + xx * (P3 + xx * (P4 + xx * P5)))); |
| 148 | y = 1. + (x * c / (2. - c) - lo + hi); |
| 149 | if k == 0 { y } else { scalbn(y, k) } |
| 150 | } |
| 151 | |