1/*
2 * Copyright (c) 2014 Advanced Micro Devices, Inc.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a copy
5 * of this software and associated documentation files (the "Software"), to deal
6 * in the Software without restriction, including without limitation the rights
7 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8 * copies of the Software, and to permit persons to whom the Software is
9 * furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice shall be included in
12 * all copies or substantial portions of the Software.
13 *
14 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
15 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
16 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
17 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
18 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
19 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
20 * THE SOFTWARE.
21 */
22
23#include <clc/clc.h>
24#include <math/clc_hypot.h>
25
26#include "config.h"
27#include "math.h"
28#include "../clcmacro.h"
29
30// Returns sqrt(x*x + y*y) with no overflow or underflow unless the result warrants it
31_CLC_DEF _CLC_OVERLOAD float __clc_hypot(float x, float y)
32{
33    uint ux = as_uint(x);
34    uint aux = ux & EXSIGNBIT_SP32;
35    uint uy = as_uint(y);
36    uint auy = uy & EXSIGNBIT_SP32;
37    float retval;
38    int c = aux > auy;
39    ux = c ? aux : auy;
40    uy = c ? auy : aux;
41
42    int xexp = clamp((int)(ux >> EXPSHIFTBITS_SP32) - EXPBIAS_SP32, -126, 126);
43    float fx_exp = as_float((xexp + EXPBIAS_SP32) << EXPSHIFTBITS_SP32);
44    float fi_exp = as_float((-xexp + EXPBIAS_SP32) << EXPSHIFTBITS_SP32);
45    float fx = as_float(ux) * fi_exp;
46    float fy = as_float(uy) * fi_exp;
47    retval = sqrt(mad(fx, fx, fy*fy)) * fx_exp;
48
49    retval = ux > PINFBITPATT_SP32 | uy == 0 ? as_float(ux) : retval;
50    retval = ux == PINFBITPATT_SP32 | uy == PINFBITPATT_SP32 ? as_float(PINFBITPATT_SP32) : retval;
51    return retval;
52}
53_CLC_BINARY_VECTORIZE(_CLC_DEF _CLC_OVERLOAD, float, __clc_hypot, float, float)
54
55#ifdef cl_khr_fp64
56_CLC_DEF _CLC_OVERLOAD double __clc_hypot(double x, double y)
57{
58    ulong ux = as_ulong(x) & ~SIGNBIT_DP64;
59    int xexp = ux >> EXPSHIFTBITS_DP64;
60    x = as_double(ux);
61
62    ulong uy = as_ulong(y) & ~SIGNBIT_DP64;
63    int yexp = uy >> EXPSHIFTBITS_DP64;
64    y = as_double(uy);
65
66    int c = xexp > EXPBIAS_DP64 + 500 | yexp > EXPBIAS_DP64 + 500;
67    double preadjust = c ? 0x1.0p-600 : 1.0;
68    double postadjust = c ? 0x1.0p+600 : 1.0;
69
70    c = xexp < EXPBIAS_DP64 - 500 | yexp < EXPBIAS_DP64 - 500;
71    preadjust = c ? 0x1.0p+600 : preadjust;
72    postadjust = c ? 0x1.0p-600 : postadjust;
73
74    double ax = x * preadjust;
75    double ay = y * preadjust;
76
77    // The post adjust may overflow, but this can't be avoided in any case
78    double r = sqrt(fma(ax, ax, ay*ay)) * postadjust;
79
80    // If the difference in exponents between x and y is large
81    double s = x + y;
82    c = abs(xexp - yexp) > MANTLENGTH_DP64 + 1;
83    r = c ? s : r;
84
85    // Check for NaN
86    //c = x != x | y != y;
87    c = isnan(x) | isnan(y);
88    r = c ? as_double(QNANBITPATT_DP64) : r;
89
90    // If either is Inf, we must return Inf
91    c = x == as_double(PINFBITPATT_DP64) | y == as_double(PINFBITPATT_DP64);
92    r = c ? as_double(PINFBITPATT_DP64) : r;
93
94    return r;
95}
96
97_CLC_BINARY_VECTORIZE(_CLC_DEF _CLC_OVERLOAD, double, __clc_hypot, double, double)
98#endif
99