/* * // Copyright (c) Radzivon Bartoshyk 7/2025. All rights reserved. * // * // Redistribution and use in source and binary forms, with or without modification, * // are permitted provided that the following conditions are met: * // * // 1. Redistributions of source code must retain the above copyright notice, this * // list of conditions and the following disclaimer. * // * // 2. Redistributions in binary form must reproduce the above copyright notice, * // this list of conditions and the following disclaimer in the documentation * // and/or other materials provided with the distribution. * // * // 3. Neither the name of the copyright holder nor the names of its * // contributors may be used to endorse or promote products derived from * // this software without specific prior written permission. * // * // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" * // AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE * // DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE * // FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * // DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR * // SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER * // CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, * // OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE * // OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ use crate::common::{dd_fmla, f_fmla}; use crate::double_double::DoubleDouble; use crate::hyperbolic::acosh::{ ACOSH_ASINH_B, ACOSH_ASINH_L1, ACOSH_ASINH_L2, ACOSH_ASINH_LL, ACOSH_ASINH_R1, ACOSH_ASINH_R2, ACOSH_ASINH_REFINE_T2, ACOSH_ASINH_REFINE_T4, ACOSH_SINH_REFINE_T1, ACOSH_SINH_REFINE_T3, lpoly_xd_generic, }; #[cold] fn asinh_refine(x: f64, a: f64, z: DoubleDouble) -> f64 { let mut t = z.hi.to_bits(); let ex: i32 = (t >> 52) as i32; let e = ex.wrapping_sub(0x3ff) + if z.lo == 0.0 { 1i32 } else { 0i32 }; t &= 0x000fffffffffffff; t |= 0x3ffu64 << 52; let ed = e as f64; let v = (a - ed + f64::from_bits(0x3ff0000800000000)).to_bits(); let i = (v.wrapping_sub(0x3ffu64 << 52)) >> (52 - 16); let i1 = (i >> 12) & 0x1f; let i2 = (i >> 8) & 0xf; let i3 = (i >> 4) & 0xf; let i4 = i & 0xf; const L20: f64 = f64::from_bits(0x3fd62e42fefa3800); const L21: f64 = f64::from_bits(0x3d1ef35793c76800); const L22: f64 = f64::from_bits(0xba49ff0342542fc3); let el2 = L22 * ed; let el1 = L21 * ed; let el0 = L20 * ed; let mut dl0: f64; let ll0i1 = ACOSH_ASINH_LL[0][i1 as usize]; let ll1i2 = ACOSH_ASINH_LL[1][i2 as usize]; let ll2i3 = ACOSH_ASINH_LL[2][i3 as usize]; let ll3i4 = ACOSH_ASINH_LL[3][i4 as usize]; dl0 = f64::from_bits(ll0i1.0) + f64::from_bits(ll1i2.0) + (f64::from_bits(ll2i3.0) + f64::from_bits(ll3i4.0)); let dl1 = f64::from_bits(ll0i1.1) + f64::from_bits(ll1i2.1) + (f64::from_bits(ll2i3.1) + f64::from_bits(ll3i4.1)); let dl2 = f64::from_bits(ll0i1.2) + f64::from_bits(ll1i2.2) + (f64::from_bits(ll2i3.2) + f64::from_bits(ll3i4.2)); dl0 += el0; let t12 = f64::from_bits(ACOSH_SINH_REFINE_T1[i1 as usize]) * f64::from_bits(ACOSH_ASINH_REFINE_T2[i2 as usize]); let t34 = f64::from_bits(ACOSH_SINH_REFINE_T3[i3 as usize]) * f64::from_bits(ACOSH_ASINH_REFINE_T4[i4 as usize]); let th = t12 * t34; let tl = dd_fmla(t12, t34, -th); let dh = th * f64::from_bits(t); let dl = dd_fmla(th, f64::from_bits(t), -dh); let sh = tl * f64::from_bits(t); let sl = dd_fmla(tl, f64::from_bits(t), -sh); let mut dx = DoubleDouble::from_exact_add(dh - 1., dl); if z.lo != 0.0 { t = z.lo.to_bits(); t = t.wrapping_sub((e as i64).wrapping_shl(52) as u64); dx.lo += th * f64::from_bits(t); } dx = DoubleDouble::add(dx, DoubleDouble::new(sl, sh)); const CL: [u64; 3] = [0xbfc0000000000000, 0x3fb9999999a0754f, 0xbfb55555555c3157]; let sl0 = f_fmla(dx.hi, f64::from_bits(CL[2]), f64::from_bits(CL[1])); let sl1 = f_fmla(dx.hi, sl0, f64::from_bits(CL[0])); let sl = dx.hi * sl1; const CH: [(u64, u64); 3] = [ (0x39024b67ee516e3b, 0x3fe0000000000000), (0xb91932ce43199a8d, 0xbfd0000000000000), (0x3c655540c15cf91f, 0x3fc5555555555555), ]; let mut s = lpoly_xd_generic(dx, CH, sl); s = DoubleDouble::quick_mult(dx, s); s = DoubleDouble::add(s, DoubleDouble::new(el2, el1)); s = DoubleDouble::add(s, DoubleDouble::new(dl2, dl1)); let mut v02 = DoubleDouble::from_exact_add(dl0, s.hi); let mut v12 = DoubleDouble::from_exact_add(v02.lo, s.lo); let scale = f64::copysign(2., x); v02.hi *= scale; v12.hi *= scale; v12.lo *= scale; t = v12.hi.to_bits(); if (t & 0x000fffffffffffff) == 0 { let w = v12.lo.to_bits(); if ((w ^ t) >> 63) != 0 { t = t.wrapping_sub(1); } else { t = t.wrapping_add(1); } v12.hi = f64::from_bits(t); } v02.hi + v12.hi } #[cold] fn as_asinh_zero(x: f64, x2h: f64, x2l: f64) -> f64 { static CH: [(u64, u64); 12] = [ (0xbc65555555555555, 0xbfc5555555555555), (0x3c499999999949df, 0x3fb3333333333333), (0x3c32492496091b0c, 0xbfa6db6db6db6db7), (0x3c1c71a35cfa0671, 0x3f9f1c71c71c71c7), (0x3c317f937248cf81, 0xbf96e8ba2e8ba2e9), (0xbc374e3c1dfd4c3d, 0x3f91c4ec4ec4ec4f), (0xbc238e7a467ecc55, 0xbf8c999999999977), (0x3c2a83c7bace55eb, 0x3f87a87878786c7e), (0xbc2d024df7fa0542, 0xbf83fde50d764083), (0xbc2ba9c13deb261f, 0x3f812ef3ceae4d12), (0xbc1546da9bc5b32a, 0xbf7df3bd104aa267), (0x3c140d284a1d67f9, 0x3f7a685fc5de7a04), ]; const CL: [u64; 5] = [ 0xbf77828d553ec800, 0x3f751712f7bee368, 0xbf72e6d98527bcc6, 0x3f70095da47b392c, 0xbf63b92d6368192c, ]; let yw0 = f_fmla(x2h, f64::from_bits(CL[4]), f64::from_bits(CL[3])); let yw1 = f_fmla(x2h, yw0, f64::from_bits(CL[2])); let yw2 = f_fmla(x2h, yw1, f64::from_bits(CL[1])); let y2 = x2h * f_fmla(x2h, yw2, f64::from_bits(CL[0])); let mut y1 = lpoly_xd_generic(DoubleDouble::new(x2l, x2h), CH, y2); y1 = DoubleDouble::quick_mult(y1, DoubleDouble::new(x2l, x2h)); y1 = DoubleDouble::quick_mult_f64(y1, x); let y0 = DoubleDouble::from_exact_add(x, y1.hi); let mut p = DoubleDouble::from_exact_add(y0.lo, y1.lo); let mut t = p.hi.to_bits(); if (t & 0x000fffffffffffff) == 0 { let w = p.lo.to_bits(); if ((w ^ t) >> 63) != 0 { t = t.wrapping_sub(1); } else { t = t.wrapping_add(1); } p.hi = f64::from_bits(t); } y0.hi + p.hi } /// Huperbolic sine function /// /// Max ULP 0.5 pub fn f_asinh(x: f64) -> f64 { let ax = x.abs(); let ix = ax.to_bits(); let u = ix; if u < 0x3fbb000000000000u64 { // |x| < 0x1.bp-4 // for |x| < 0x1.7137449123ef7p-26, asinh(x) rounds to x to nearest // for |x| < 0x1p-1022 we have underflow but not for 0x1p-1022 (to nearest) if u < 0x3e57137449123ef7u64 { // |x| < 0x1.7137449123ef7p-26 if u == 0 { return x; } let res = f_fmla(f64::from_bits(0xbc30000000000000), x, x); return res; } let x2h = x * x; let x2l = dd_fmla(x, x, -x2h); let x3h = x2h * x; let sl; if u < 0x3f93000000000000u64 { // |x| < 0x1.3p-6 if u < 0x3f30000000000000u64 { // |x| < 0x1p-12 if u < 0x3e5a000000000000u64 { // |x| < 0x1.ap-26 sl = x3h * f64::from_bits(0xbfc5555555555555); } else { const CL: [u64; 2] = [0xbfc5555555555555, 0x3fb3333327c57c60]; let sl0 = f_fmla(x2h, f64::from_bits(CL[1]), f64::from_bits(CL[0])); sl = x3h * sl0; } } else { const CL: [u64; 4] = [ 0xbfc5555555555555, 0x3fb333333332f2ff, 0xbfa6db6d9a665159, 0x3f9f186866d775f0, ]; let sl0 = f_fmla(x2h, f64::from_bits(CL[3]), f64::from_bits(CL[2])); let sl1 = f_fmla(x2h, sl0, f64::from_bits(CL[1])); sl = x3h * f_fmla(x2h, sl1, f64::from_bits(CL[0])); } } else { const CL: [u64; 7] = [ 0xbfc5555555555555, 0x3fb3333333333310, 0xbfa6db6db6da466c, 0x3f9f1c71c2ea7be4, 0xbf96e8b651b09d72, 0x3f91c309fc0e69c2, 0xbf8bab7833c1e000, ]; let c1 = f_fmla(x2h, f64::from_bits(CL[2]), f64::from_bits(CL[1])); let c3 = f_fmla(x2h, f64::from_bits(CL[4]), f64::from_bits(CL[3])); let c5 = f_fmla(x2h, f64::from_bits(CL[6]), f64::from_bits(CL[5])); let x4 = x2h * x2h; let sl0 = f_fmla(x4, c5, c3); let sl1 = f_fmla(x4, sl0, c1); sl = x3h * f_fmla(x2h, sl1, f64::from_bits(CL[0])); } let eps = f64::from_bits(0x3ca6000000000000) * x3h; let lb = x + (sl - eps); let ub = x + (sl + eps); if lb == ub { return lb; } return as_asinh_zero(x, x2h, x2l); } // |x| >= 0x1.bp-4 let mut x2h: f64 = 0.; let mut x2l: f64 = 0.; let mut off: i32 = 0x3ff; let va: DoubleDouble = if u < 0x4190000000000000 { // x < 0x1p+26 x2h = x * x; x2l = dd_fmla(x, x, -x2h); let mut dt = if u < 0x3ff0000000000000 { DoubleDouble::from_exact_add(1., x2h) } else { DoubleDouble::from_exact_add(x2h, 1.) }; dt.lo += x2l; let ah = dt.hi.sqrt(); let rs = 0.5 / dt.hi; let al = (dt.lo - dd_fmla(ah, ah, -dt.hi)) * (rs * ah); let mut ma = DoubleDouble::from_exact_add(ah, ax); ma.lo += al; ma } else if u < 0x4330000000000000 { DoubleDouble::new(0.5 / ax, 2. * ax) } else { if u >= 0x7ff0000000000000u64 { return x + x; } // +-inf or nan off = 0x3fe; DoubleDouble::new(0., ax) }; let mut t = va.hi.to_bits(); let ex = (t >> 52) as i32; let e = ex.wrapping_sub(off); t &= 0x000fffffffffffff; let ed = e as f64; let i = t >> (52 - 5); let d = (t & 0x00007fffffffffff) as i64; let b_i = ACOSH_ASINH_B[i as usize]; let j: u64 = t .wrapping_add((b_i[0] as u64).wrapping_shl(33)) .wrapping_add((b_i[1] as i64).wrapping_mul(d >> 16) as u64) >> (52 - 10); t |= 0x3ffu64 << 52; let i1 = (j >> 5) as i32; let i2 = j & 0x1f; let r = f64::from_bits(ACOSH_ASINH_R1[i1 as usize]) * f64::from_bits(ACOSH_ASINH_R2[i2 as usize]); let dx = dd_fmla(r, f64::from_bits(t), -1.); let dx2 = dx * dx; const C: [u64; 5] = [ 0xbfe0000000000000, 0x3fd5555555555530, 0xbfcfffffffffffa0, 0x3fc99999e33a6366, 0xbfc555559ef9525f, ]; let fw0 = f_fmla(dx, f64::from_bits(C[3]), f64::from_bits(C[2])); let fw1 = f_fmla(dx, f64::from_bits(C[1]), f64::from_bits(C[0])); let fw2 = f_fmla(dx2, f64::from_bits(C[4]), fw0); let f = dx2 * f_fmla(dx2, fw2, fw1); const L2H: f64 = f64::from_bits(0x3fe62e42fefa3800); const L2L: f64 = f64::from_bits(0x3d2ef35793c76730); let l1i1 = ACOSH_ASINH_L1[i1 as usize]; let l1i2 = ACOSH_ASINH_L2[i2 as usize]; let mut lh = f_fmla(L2H, ed, f64::from_bits(l1i1.1) + f64::from_bits(l1i2.1)); let mut ll = f_fmla( L2L, ed, f64::from_bits(l1i1.0) + f64::from_bits(l1i2.0) + va.lo / va.hi + f, ); ll += dx; lh *= f64::copysign(1., x); ll *= f64::copysign(1., x); let eps = 1.63e-19; let lb = lh + (ll - eps); let ub = lh + (ll + eps); if lb == ub { return lb; } if ax < f64::from_bits(0x3fd0000000000000) { return as_asinh_zero(x, x2h, x2l); } asinh_refine(x, f64::from_bits(0x3ff71547652b82fe) * lb.abs(), va) } #[cfg(test)] mod tests { use super::*; #[test] fn test_asinh() { assert_eq!(f_asinh(-0.05268859863273256), -0.05266425100170862); assert_eq!(f_asinh(1.05268859863273256), 0.9181436936151385); } }