1 /* Double-precision vector (SVE) cos function.
2
3 Copyright (C) 2023 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 "sv_math.h"
21
22 static const struct data
23 {
24 double inv_pio2, pio2_1, pio2_2, pio2_3, shift;
25 } data = {
26 /* Polynomial coefficients are hardwired in FTMAD instructions. */
27 .inv_pio2 = 0x1.45f306dc9c882p-1,
28 .pio2_1 = 0x1.921fb50000000p+0,
29 .pio2_2 = 0x1.110b460000000p-26,
30 .pio2_3 = 0x1.1a62633145c07p-54,
31 /* Original shift used in AdvSIMD cos,
32 plus a contribution to set the bit #0 of q
33 as expected by trigonometric instructions. */
34 .shift = 0x1.8000000000001p52
35 };
36
37 #define RangeVal 0x4160000000000000 /* asuint64 (0x1p23). */
38
39 static svfloat64_t NOINLINE
40 special_case (svfloat64_t x, svfloat64_t y, svbool_t out_of_bounds)
41 {
42 return sv_call_f64 (cos, x, y, out_of_bounds);
43 }
44
45 /* A fast SVE implementation of cos based on trigonometric
46 instructions (FTMAD, FTSSEL, FTSMUL).
47 Maximum measured error: 2.108 ULPs.
48 SV_NAME_D1 (cos)(0x1.9b0ba158c98f3p+7) got -0x1.fddd4c65c7f07p-3
49 want -0x1.fddd4c65c7f05p-3. */
50 svfloat64_t SV_NAME_D1 (cos) (svfloat64_t x, const svbool_t pg)
51 {
52 const struct data *d = ptr_barrier (&data);
53
54 svfloat64_t r = svabs_f64_x (pg, x);
55 svbool_t out_of_bounds
56 = svcmpge_n_u64 (pg, svreinterpret_u64_f64 (r), RangeVal);
57
58 /* Load some constants in quad-word chunks to minimise memory access. */
59 svbool_t ptrue = svptrue_b64 ();
60 svfloat64_t invpio2_and_pio2_1 = svld1rq_f64 (ptrue, &d->inv_pio2);
61 svfloat64_t pio2_23 = svld1rq_f64 (ptrue, &d->pio2_2);
62
63 /* n = rint(|x|/(pi/2)). */
64 svfloat64_t q = svmla_lane_f64 (sv_f64 (d->shift), r, invpio2_and_pio2_1, 0);
65 svfloat64_t n = svsub_n_f64_x (pg, q, d->shift);
66
67 /* r = |x| - n*(pi/2) (range reduction into -pi/4 .. pi/4). */
68 r = svmls_lane_f64 (r, n, invpio2_and_pio2_1, 1);
69 r = svmls_lane_f64 (r, n, pio2_23, 0);
70 r = svmls_lane_f64 (r, n, pio2_23, 1);
71
72 /* cos(r) poly approx. */
73 svfloat64_t r2 = svtsmul_f64 (r, svreinterpret_u64_f64 (q));
74 svfloat64_t y = sv_f64 (0.0);
75 y = svtmad_f64 (y, r2, 7);
76 y = svtmad_f64 (y, r2, 6);
77 y = svtmad_f64 (y, r2, 5);
78 y = svtmad_f64 (y, r2, 4);
79 y = svtmad_f64 (y, r2, 3);
80 y = svtmad_f64 (y, r2, 2);
81 y = svtmad_f64 (y, r2, 1);
82 y = svtmad_f64 (y, r2, 0);
83
84 /* Final multiplicative factor: 1.0 or x depending on bit #0 of q. */
85 svfloat64_t f = svtssel_f64 (r, svreinterpret_u64_f64 (q));
86 /* Apply factor. */
87 y = svmul_f64_x (pg, f, y);
88
89 if (__glibc_unlikely (svptest_any (pg, out_of_bounds)))
90 return special_case (x, y, out_of_bounds);
91 return y;
92 }