1 /*
2 * Single-precision SVE tanh(x) function.
3 *
4 * Copyright (c) 2023-2024, Arm Limited.
5 * SPDX-License-Identifier: MIT OR Apache-2.0 WITH LLVM-exception
6 */
7
8 #include "sv_math.h"
9 #include "test_sig.h"
10 #include "test_defs.h"
11 #include "sv_expm1f_inline.h"
12
13 /* Largest value of x for which tanhf(x) rounds to 1 (or -1 for negative). */
14 #define BoringBound 0x1.205966p+3f
15
16 static const struct data
17 {
18 struct sv_expm1f_data expm1f_consts;
19 uint32_t onef, special_bound;
20 float boring_bound;
21 } data = {
22 .expm1f_consts = SV_EXPM1F_DATA,
23 .onef = 0x3f800000,
24 .special_bound = 0x7f800000,
25 .boring_bound = BoringBound,
26 };
27
28 static svfloat32_t NOINLINE
special_case(svfloat32_t x,svbool_t pg,svbool_t is_boring,svfloat32_t boring,svfloat32_t q,svbool_t special)29 special_case (svfloat32_t x, svbool_t pg, svbool_t is_boring,
30 svfloat32_t boring, svfloat32_t q, svbool_t special)
31 {
32 svfloat32_t y
33 = svsel_f32 (is_boring, boring, svdiv_x (pg, q, svadd_x (pg, q, 2.0)));
34 return sv_call_f32 (tanhf, x, y, special);
35 }
36
37 /* Approximation for single-precision SVE tanh(x), using a simplified
38 version of expm1f. The maximum error is 2.57 ULP:
39 _ZGVsMxv_tanhf (0x1.fc1832p-5) got 0x1.fb71a4p-5
40 want 0x1.fb71aap-5. */
SV_NAME_F1(tanh)41 svfloat32_t SV_NAME_F1 (tanh) (svfloat32_t x, const svbool_t pg)
42 {
43 const struct data *d = ptr_barrier (&data);
44
45 svfloat32_t ax = svabs_x (pg, x);
46 svuint32_t iax = svreinterpret_u32 (ax);
47 svuint32_t sign = sveor_x (pg, svreinterpret_u32 (x), iax);
48 svfloat32_t boring = svreinterpret_f32 (svorr_x (pg, sign, d->onef));
49 svbool_t special = svcmpgt (pg, iax, d->special_bound);
50 svbool_t is_boring = svacgt (pg, x, d->boring_bound);
51
52 /* tanh(x) = (e^2x - 1) / (e^2x + 1). */
53 svfloat32_t q = expm1f_inline (svmul_x (svptrue_b32 (), x, 2.0), pg,
54 &d->expm1f_consts);
55
56 if (unlikely (svptest_any (pg, special)))
57 return special_case (x, pg, is_boring, boring, q, special);
58 svfloat32_t y = svdiv_x (pg, q, svadd_x (pg, q, 2.0));
59 return svsel_f32 (is_boring, boring, y);
60 }
61
62 TEST_SIG (SV, F, 1, tanh, -10.0, 10.0)
63 TEST_ULP (SV_NAME_F1 (tanh), 2.07)
64 TEST_DISABLE_FENV (SV_NAME_F1 (tanh))
65 TEST_SYM_INTERVAL (SV_NAME_F1 (tanh), 0, 0x1p-23, 1000)
66 TEST_SYM_INTERVAL (SV_NAME_F1 (tanh), 0x1p-23, BoringBound, 100000)
67 TEST_SYM_INTERVAL (SV_NAME_F1 (tanh), BoringBound, inf, 100)
68 CLOSE_SVE_ATTR
69