[352] | 1 | /* origin: FreeBSD /usr/src/lib/msun/src/s_sinf.c */
|
---|
| 2 | /*
|
---|
| 3 | * Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
|
---|
| 4 | * Optimized by Bruce D. Evans.
|
---|
| 5 | */
|
---|
| 6 | /*
|
---|
| 7 | * ====================================================
|
---|
| 8 | * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
|
---|
| 9 | *
|
---|
| 10 | * Developed at SunPro, a Sun Microsystems, Inc. business.
|
---|
| 11 | * Permission to use, copy, modify, and distribute this
|
---|
| 12 | * software is freely granted, provided that this notice
|
---|
| 13 | * is preserved.
|
---|
| 14 | * ====================================================
|
---|
| 15 | */
|
---|
| 16 |
|
---|
| 17 | #define _GNU_SOURCE
|
---|
| 18 | #include "libm.h"
|
---|
| 19 |
|
---|
| 20 | /* Small multiples of pi/2 rounded to double precision. */
|
---|
| 21 | static const double
|
---|
| 22 | s1pio2 = 1*M_PI_2, /* 0x3FF921FB, 0x54442D18 */
|
---|
| 23 | s2pio2 = 2*M_PI_2, /* 0x400921FB, 0x54442D18 */
|
---|
| 24 | s3pio2 = 3*M_PI_2, /* 0x4012D97C, 0x7F3321D2 */
|
---|
| 25 | s4pio2 = 4*M_PI_2; /* 0x401921FB, 0x54442D18 */
|
---|
| 26 |
|
---|
| 27 | void sincosf(float x, float *sin, float *cos)
|
---|
| 28 | {
|
---|
| 29 | double y;
|
---|
| 30 | float_t s, c;
|
---|
| 31 | uint32_t ix;
|
---|
| 32 | unsigned n, sign;
|
---|
| 33 |
|
---|
| 34 | GET_FLOAT_WORD(ix, x);
|
---|
| 35 | sign = ix >> 31;
|
---|
| 36 | ix &= 0x7fffffff;
|
---|
| 37 |
|
---|
| 38 | /* |x| ~<= pi/4 */
|
---|
| 39 | if (ix <= 0x3f490fda) {
|
---|
| 40 | /* |x| < 2**-12 */
|
---|
| 41 | if (ix < 0x39800000) {
|
---|
| 42 | /* raise inexact if x!=0 and underflow if subnormal */
|
---|
| 43 | FORCE_EVAL(ix < 0x00100000 ? x/0x1p120f : x+0x1p120f);
|
---|
| 44 | *sin = x;
|
---|
| 45 | *cos = 1.0f;
|
---|
| 46 | return;
|
---|
| 47 | }
|
---|
| 48 | *sin = __sindf(x);
|
---|
| 49 | *cos = __cosdf(x);
|
---|
| 50 | return;
|
---|
| 51 | }
|
---|
| 52 |
|
---|
| 53 | /* |x| ~<= 5*pi/4 */
|
---|
| 54 | if (ix <= 0x407b53d1) {
|
---|
| 55 | if (ix <= 0x4016cbe3) { /* |x| ~<= 3pi/4 */
|
---|
| 56 | if (sign) {
|
---|
| 57 | *sin = -__cosdf(x + s1pio2);
|
---|
| 58 | *cos = __sindf(x + s1pio2);
|
---|
| 59 | } else {
|
---|
| 60 | *sin = __cosdf(s1pio2 - x);
|
---|
| 61 | *cos = __sindf(s1pio2 - x);
|
---|
| 62 | }
|
---|
| 63 | return;
|
---|
| 64 | }
|
---|
| 65 | /* -sin(x+c) is not correct if x+c could be 0: -0 vs +0 */
|
---|
| 66 | *sin = -__sindf(sign ? x + s2pio2 : x - s2pio2);
|
---|
| 67 | *cos = -__cosdf(sign ? x + s2pio2 : x - s2pio2);
|
---|
| 68 | return;
|
---|
| 69 | }
|
---|
| 70 |
|
---|
| 71 | /* |x| ~<= 9*pi/4 */
|
---|
| 72 | if (ix <= 0x40e231d5) {
|
---|
| 73 | if (ix <= 0x40afeddf) { /* |x| ~<= 7*pi/4 */
|
---|
| 74 | if (sign) {
|
---|
| 75 | *sin = __cosdf(x + s3pio2);
|
---|
| 76 | *cos = -__sindf(x + s3pio2);
|
---|
| 77 | } else {
|
---|
| 78 | *sin = -__cosdf(x - s3pio2);
|
---|
| 79 | *cos = __sindf(x - s3pio2);
|
---|
| 80 | }
|
---|
| 81 | return;
|
---|
| 82 | }
|
---|
| 83 | *sin = __sindf(sign ? x + s4pio2 : x - s4pio2);
|
---|
| 84 | *cos = __cosdf(sign ? x + s4pio2 : x - s4pio2);
|
---|
| 85 | return;
|
---|
| 86 | }
|
---|
| 87 |
|
---|
| 88 | /* sin(Inf or NaN) is NaN */
|
---|
| 89 | if (ix >= 0x7f800000) {
|
---|
| 90 | *sin = *cos = x - x;
|
---|
| 91 | return;
|
---|
| 92 | }
|
---|
| 93 |
|
---|
| 94 | /* general argument reduction needed */
|
---|
| 95 | n = __rem_pio2f(x, &y);
|
---|
| 96 | s = __sindf(y);
|
---|
| 97 | c = __cosdf(y);
|
---|
| 98 | switch (n&3) {
|
---|
| 99 | case 0:
|
---|
| 100 | *sin = s;
|
---|
| 101 | *cos = c;
|
---|
| 102 | break;
|
---|
| 103 | case 1:
|
---|
| 104 | *sin = c;
|
---|
| 105 | *cos = -s;
|
---|
| 106 | break;
|
---|
| 107 | case 2:
|
---|
| 108 | *sin = -s;
|
---|
| 109 | *cos = -c;
|
---|
| 110 | break;
|
---|
| 111 | case 3:
|
---|
| 112 | default:
|
---|
| 113 | *sin = -c;
|
---|
| 114 | *cos = s;
|
---|
| 115 | break;
|
---|
| 116 | }
|
---|
| 117 | }
|
---|