1 | /* Single-precision vector (Advanced SIMD) cos function. |
2 | |
3 | Copyright (C) 2023-2024 Free Software Foundation, Inc. |
4 | This file is part of the GNU C Library. |
5 | |
6 | The GNU C Library is free software; you can redistribute it and/or |
7 | modify it under the terms of the GNU Lesser General Public |
8 | License as published by the Free Software Foundation; either |
9 | version 2.1 of the License, or (at your option) any later version. |
10 | |
11 | The GNU C Library is distributed in the hope that it will be useful, |
12 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
13 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
14 | Lesser General Public License for more details. |
15 | |
16 | You should have received a copy of the GNU Lesser General Public |
17 | License along with the GNU C Library; if not, see |
18 | <https://www.gnu.org/licenses/>. */ |
19 | |
20 | #include "v_math.h" |
21 | |
22 | static const struct data |
23 | { |
24 | float32x4_t poly[4]; |
25 | float32x4_t range_val, inv_pi, half_pi, shift, pi_1, pi_2, pi_3; |
26 | } data = { |
27 | /* 1.886 ulp error. */ |
28 | .poly = { V4 (-0x1.555548p-3f), V4 (0x1.110df4p-7f), V4 (-0x1.9f42eap-13f), |
29 | V4 (0x1.5b2e76p-19f) }, |
30 | |
31 | .pi_1 = V4 (0x1.921fb6p+1f), |
32 | .pi_2 = V4 (-0x1.777a5cp-24f), |
33 | .pi_3 = V4 (-0x1.ee59dap-49f), |
34 | |
35 | .inv_pi = V4 (0x1.45f306p-2f), |
36 | .shift = V4 (0x1.8p+23f), |
37 | .half_pi = V4 (0x1.921fb6p0f), |
38 | .range_val = V4 (0x1p20f) |
39 | }; |
40 | |
41 | #define C(i) d->poly[i] |
42 | |
43 | static float32x4_t VPCS_ATTR NOINLINE |
44 | special_case (float32x4_t x, float32x4_t y, uint32x4_t odd, uint32x4_t cmp) |
45 | { |
46 | /* Fall back to scalar code. */ |
47 | y = vreinterpretq_f32_u32 (veorq_u32 (vreinterpretq_u32_f32 (y), odd)); |
48 | return v_call_f32 (cosf, x, y, cmp); |
49 | } |
50 | |
51 | float32x4_t VPCS_ATTR NOINLINE V_NAME_F1 (cos) (float32x4_t x) |
52 | { |
53 | const struct data *d = ptr_barrier (&data); |
54 | float32x4_t n, r, r2, r3, y; |
55 | uint32x4_t odd, cmp; |
56 | |
57 | #if WANT_SIMD_EXCEPT |
58 | r = vabsq_f32 (x); |
59 | cmp = vcgeq_u32 (vreinterpretq_u32_f32 (r), |
60 | vreinterpretq_u32_f32 (d->range_val)); |
61 | if (__glibc_unlikely (v_any_u32 (cmp))) |
62 | /* If fenv exceptions are to be triggered correctly, set any special lanes |
63 | to 1 (which is neutral w.r.t. fenv). These lanes will be fixed by |
64 | special-case handler later. */ |
65 | r = vbslq_f32 (cmp, v_f32 (1.0f), r); |
66 | #else |
67 | cmp = vcageq_f32 (x, d->range_val); |
68 | r = x; |
69 | #endif |
70 | |
71 | /* n = rint((|x|+pi/2)/pi) - 0.5. */ |
72 | n = vfmaq_f32 (d->shift, d->inv_pi, vaddq_f32 (r, d->half_pi)); |
73 | odd = vshlq_n_u32 (vreinterpretq_u32_f32 (n), 31); |
74 | n = vsubq_f32 (n, d->shift); |
75 | n = vsubq_f32 (n, v_f32 (0.5f)); |
76 | |
77 | /* r = |x| - n*pi (range reduction into -pi/2 .. pi/2). */ |
78 | r = vfmsq_f32 (r, d->pi_1, n); |
79 | r = vfmsq_f32 (r, d->pi_2, n); |
80 | r = vfmsq_f32 (r, d->pi_3, n); |
81 | |
82 | /* y = sin(r). */ |
83 | r2 = vmulq_f32 (r, r); |
84 | r3 = vmulq_f32 (r2, r); |
85 | y = vfmaq_f32 (C (2), C (3), r2); |
86 | y = vfmaq_f32 (C (1), y, r2); |
87 | y = vfmaq_f32 (C (0), y, r2); |
88 | y = vfmaq_f32 (r, y, r3); |
89 | |
90 | if (__glibc_unlikely (v_any_u32 (cmp))) |
91 | return special_case (x, y, odd, cmp); |
92 | return vreinterpretq_f32_u32 (veorq_u32 (vreinterpretq_u32_f32 (y), odd)); |
93 | } |
94 | libmvec_hidden_def (V_NAME_F1 (cos)) |
95 | HALF_WIDTH_ALIAS_F1 (cos) |
96 | |