xref: /freebsd/contrib/arm-optimized-routines/pl/math/v_hypot_1u5.c (revision 8a802df1de2d77fd0a62996bd785ca3f1326887f)
1 /*
2  * Double-precision vector hypot(x) function.
3  *
4  * Copyright (c) 2023, Arm Limited.
5  * SPDX-License-Identifier: MIT OR Apache-2.0 WITH LLVM-exception
6  */
7 
8 #include "v_math.h"
9 #include "pl_sig.h"
10 #include "pl_test.h"
11 
12 #if WANT_SIMD_EXCEPT
13 static const struct data
14 {
15   uint64x2_t tiny_bound, thres;
16 } data = {
17   .tiny_bound = V2 (0x2000000000000000), /* asuint (0x1p-511).  */
18   .thres = V2 (0x3fe0000000000000), /* asuint (0x1p511) - tiny_bound.  */
19 };
20 #else
21 static const struct data
22 {
23   uint64x2_t tiny_bound;
24   uint32x4_t thres;
25 } data = {
26   .tiny_bound = V2 (0x0360000000000000), /* asuint (0x1p-969).  */
27   .thres = V4 (0x7c900000),	 /* asuint (inf) - tiny_bound.  */
28 };
29 #endif
30 
31 static float64x2_t VPCS_ATTR NOINLINE
32 special_case (float64x2_t x, float64x2_t y, float64x2_t sqsum,
33 	      uint32x2_t special)
34 {
35   return v_call2_f64 (hypot, x, y, vsqrtq_f64 (sqsum), vmovl_u32 (special));
36 }
37 
38 /* Vector implementation of double-precision hypot.
39    Maximum error observed is 1.21 ULP:
40    _ZGVnN2vv_hypot (0x1.6a1b193ff85b5p-204, 0x1.bc50676c2a447p-222)
41     got 0x1.6a1b19400964ep-204
42    want 0x1.6a1b19400964dp-204.  */
43 #if WANT_SIMD_EXCEPT
44 
45 float64x2_t VPCS_ATTR V_NAME_D2 (hypot) (float64x2_t x, float64x2_t y)
46 {
47   const struct data *d = ptr_barrier (&data);
48 
49   float64x2_t ax = vabsq_f64 (x);
50   float64x2_t ay = vabsq_f64 (y);
51 
52   uint64x2_t ix = vreinterpretq_u64_f64 (ax);
53   uint64x2_t iy = vreinterpretq_u64_f64 (ay);
54 
55   /* Extreme values, NaNs, and infinities should be handled by the scalar
56      fallback for correct flag handling.  */
57   uint64x2_t specialx = vcgeq_u64 (vsubq_u64 (ix, d->tiny_bound), d->thres);
58   uint64x2_t specialy = vcgeq_u64 (vsubq_u64 (iy, d->tiny_bound), d->thres);
59   ax = v_zerofy_f64 (ax, specialx);
60   ay = v_zerofy_f64 (ay, specialy);
61   uint32x2_t special = vaddhn_u64 (specialx, specialy);
62 
63   float64x2_t sqsum = vfmaq_f64 (vmulq_f64 (ax, ax), ay, ay);
64 
65   if (unlikely (v_any_u32h (special)))
66     return special_case (x, y, sqsum, special);
67 
68   return vsqrtq_f64 (sqsum);
69 }
70 #else
71 
72 float64x2_t VPCS_ATTR V_NAME_D2 (hypot) (float64x2_t x, float64x2_t y)
73 {
74   const struct data *d = ptr_barrier (&data);
75 
76   float64x2_t sqsum = vfmaq_f64 (vmulq_f64 (x, x), y, y);
77 
78   uint32x2_t special = vcge_u32 (
79       vsubhn_u64 (vreinterpretq_u64_f64 (sqsum), d->tiny_bound),
80       vget_low_u32 (d->thres));
81 
82   if (unlikely (v_any_u32h (special)))
83     return special_case (x, y, sqsum, special);
84 
85   return vsqrtq_f64 (sqsum);
86 }
87 #endif
88 
89 PL_SIG (V, D, 2, hypot, -10.0, 10.0)
90 PL_TEST_ULP (V_NAME_D2 (hypot), 1.21)
91 PL_TEST_EXPECT_FENV (V_NAME_D2 (hypot), WANT_SIMD_EXCEPT)
92 PL_TEST_INTERVAL2 (V_NAME_D2 (hypot), 0, inf, 0, inf, 10000)
93 PL_TEST_INTERVAL2 (V_NAME_D2 (hypot), 0, inf, -0, -inf, 10000)
94 PL_TEST_INTERVAL2 (V_NAME_D2 (hypot), -0, -inf, 0, inf, 10000)
95 PL_TEST_INTERVAL2 (V_NAME_D2 (hypot), -0, -inf, -0, -inf, 10000)
96