You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

128 lines
3.9 KiB

/*
* Copyright (c) 2014 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include <clc/clc.h>
#include "ep_log.h"
#include "math.h"
#include "../clcmacro.h"
_CLC_OVERLOAD _CLC_DEF float acosh(float x) {
uint ux = as_uint(x);
// Arguments greater than 1/sqrt(epsilon) in magnitude are
// approximated by acosh(x) = ln(2) + ln(x)
// For 2.0 <= x <= 1/sqrt(epsilon) the approximation is
// acosh(x) = ln(x + sqrt(x*x-1)) */
int high = ux > 0x46000000U;
int med = ux > 0x40000000U;
float w = x - 1.0f;
float s = w*w + 2.0f*w;
float t = x*x - 1.0f;
float r = sqrt(med ? t : s) + (med ? x : w);
float v = (high ? x : r) - (med ? 1.0f : 0.0f);
float z = log1p(v) + (high ? 0x1.62e430p-1f : 0.0f);
z = ux >= PINFBITPATT_SP32 ? x : z;
z = x < 1.0f ? as_float(QNANBITPATT_SP32) : z;
return z;
}
_CLC_UNARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, float, acosh, float)
#ifdef cl_khr_fp64
#pragma OPENCL EXTENSION cl_khr_fp64 : enable
_CLC_OVERLOAD _CLC_DEF double acosh(double x) {
const double recrteps = 0x1.6a09e667f3bcdp+26; // 1/sqrt(eps) = 9.49062656242515593767e+07
//log2_lead and log2_tail sum to an extra-precise version of log(2)
const double log2_lead = 0x1.62e42ep-1;
const double log2_tail = 0x1.efa39ef35793cp-25;
// Handle x >= 128 here
int xlarge = x > recrteps;
double r = x + sqrt(fma(x, x, -1.0));
r = xlarge ? x : r;
int xexp;
double r1, r2;
__clc_ep_log(r, &xexp, &r1, &r2);
double dxexp = xexp + xlarge;
r1 = fma(dxexp, log2_lead, r1);
r2 = fma(dxexp, log2_tail, r2);
double ret1 = r1 + r2;
// Handle 1 < x < 128 here
// We compute the value
// t = x - 1.0 + sqrt(2.0*(x - 1.0) + (x - 1.0)*(x - 1.0))
// using simulated quad precision.
double t = x - 1.0;
double u1 = t * 2.0;
// (t,0) * (t,0) -> (v1, v2)
double v1 = t * t;
double v2 = fma(t, t, -v1);
// (u1,0) + (v1,v2) -> (w1,w2)
r = u1 + v1;
double s = (((u1 - r) + v1) + v2);
double w1 = r + s;
double w2 = (r - w1) + s;
// sqrt(w1,w2) -> (u1,u2)
double p1 = sqrt(w1);
double a1 = p1*p1;
double a2 = fma(p1, p1, -a1);
double temp = (((w1 - a1) - a2) + w2);
double p2 = MATH_DIVIDE(temp * 0.5, p1);
u1 = p1 + p2;
double u2 = (p1 - u1) + p2;
// (u1,u2) + (t,0) -> (r1,r2)
r = u1 + t;
s = ((u1 - r) + t) + u2;
// r1 = r + s;
// r2 = (r - r1) + s;
// t = r1 + r2;
t = r + s;
// For arguments 1.13 <= x <= 1.5 the log1p function is good enough
double ret2 = log1p(t);
ulong ux = as_ulong(x);
double ret = x >= 128.0 ? ret1 : ret2;
ret = ux >= 0x7FF0000000000000 ? x : ret;
ret = x == 1.0 ? 0.0 : ret;
ret = (ux & SIGNBIT_DP64) != 0UL | x < 1.0 ? as_double(QNANBITPATT_DP64) : ret;
return ret;
}
_CLC_UNARY_VECTORIZE(_CLC_OVERLOAD _CLC_DEF, double, acosh, double)
#endif