xref: /freebsd/contrib/arm-optimized-routines/pl/math/sv_hypot_1u5.c (revision 7fdf597e96a02165cfe22ff357b857d5fa15ed8a)
1 /*
2  * Double-precision SVE 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 "sv_math.h"
9 #include "pl_sig.h"
10 #include "pl_test.h"
11 
12 static const struct data
13 {
14   uint64_t tiny_bound, thres;
15 } data = {
16   .tiny_bound = 0x0c80000000000000, /* asuint (0x1p-102).  */
17   .thres = 0x7300000000000000,	    /* asuint (inf) - tiny_bound.  */
18 };
19 
20 static svfloat64_t NOINLINE
21 special_case (svfloat64_t sqsum, svfloat64_t x, svfloat64_t y, svbool_t pg,
22 	      svbool_t special)
23 {
24   return sv_call2_f64 (hypot, x, y, svsqrt_x (pg, sqsum), special);
25 }
26 
27 /* SVE implementation of double-precision hypot.
28    Maximum error observed is 1.21 ULP:
29    _ZGVsMxvv_hypot (-0x1.6a22d0412cdd3p+352, 0x1.d3d89bd66fb1ap+330)
30     got 0x1.6a22d0412cfp+352
31    want 0x1.6a22d0412cf01p+352.  */
32 svfloat64_t SV_NAME_D2 (hypot) (svfloat64_t x, svfloat64_t y, svbool_t pg)
33 {
34   const struct data *d = ptr_barrier (&data);
35 
36   svfloat64_t sqsum = svmla_x (pg, svmul_x (pg, x, x), y, y);
37 
38   svbool_t special = svcmpge (
39       pg, svsub_x (pg, svreinterpret_u64 (sqsum), d->tiny_bound), d->thres);
40 
41   if (unlikely (svptest_any (pg, special)))
42     return special_case (sqsum, x, y, pg, special);
43   return svsqrt_x (pg, sqsum);
44 }
45 
46 PL_SIG (SV, D, 2, hypot, -10.0, 10.0)
47 PL_TEST_ULP (SV_NAME_D2 (hypot), 0.71)
48 PL_TEST_INTERVAL2 (SV_NAME_D2 (hypot), 0, inf, 0, inf, 10000)
49 PL_TEST_INTERVAL2 (SV_NAME_D2 (hypot), 0, inf, -0, -inf, 10000)
50 PL_TEST_INTERVAL2 (SV_NAME_D2 (hypot), -0, -inf, 0, inf, 10000)
51 PL_TEST_INTERVAL2 (SV_NAME_D2 (hypot), -0, -inf, -0, -inf, 10000)
52