inari 1.0.0

A Rust implementation of interval arithmetic
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204

use std::{
    arch::{aarch64::*, asm},
    mem::transmute,
};

pub(crate) const HAS_MAXIMUM: bool = true;

pub(crate) type F64X2 = float64x2_t;

pub(crate) fn abs(x: F64X2) -> F64X2 {
    unsafe { vabsq_f64(x) }
}

pub(crate) fn add_rn(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { vaddq_f64(x, y) }
}

pub(crate) fn all(x: F64X2) -> bool {
    unsafe { transmute::<_, u64>(vmovn_u64(transmute(x))) == !0u64 }
}

pub(crate) fn and(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vandq_u64(transmute(x), transmute(y))) }
}

pub(crate) fn any(x: F64X2) -> bool {
    unsafe { transmute::<_, u64>(vmovn_u64(transmute(x))) != 0u64 }
}

pub(crate) fn bitmask(x: F64X2) -> u32 {
    let [a, b] = unsafe { transmute::<_, [u64; 2]>(x) };
    (b as u32) & 0x2 | (a as u32) & 0x1
}

pub(crate) fn ceil(x: F64X2) -> F64X2 {
    unsafe { vrndpq_f64(x) }
}

pub(crate) fn constant(x: f64, y: f64) -> F64X2 {
    unsafe { transmute([x, y]) }
}

pub(crate) fn eq(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vceqq_f64(x, y)) }
}

pub(crate) fn extract(x: F64X2) -> [f64; 2] {
    unsafe { transmute::<_, [f64; 2]>(x) }
}

pub(crate) fn extract0(x: F64X2) -> f64 {
    unsafe { transmute::<_, [f64; 2]>(x)[0] }
}

pub(crate) fn extract1(x: F64X2) -> f64 {
    unsafe { transmute::<_, [f64; 2]>(x)[1] }
}

pub(crate) fn floor(x: F64X2) -> F64X2 {
    unsafe { vrndmq_f64(x) }
}

pub(crate) fn ge(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vcgeq_f64(x, y)) }
}

pub(crate) fn gt(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vcgtq_f64(x, y)) }
}

pub(crate) fn hadd_rn(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { constant(vaddvq_f64(x), vaddvq_f64(y)) }
}

pub(crate) fn le(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vcleq_f64(x, y)) }
}

pub(crate) fn lt(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vcltq_f64(x, y)) }
}

pub(crate) fn max(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { vmaxnmq_f64(x, y) }
}

pub(crate) fn maximum(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { vmaxq_f64(x, y) }
}

pub(crate) fn min(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { vminnmq_f64(x, y) }
}

pub(crate) fn minimum(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { vminq_f64(x, y) }
}

pub(crate) fn neg(x: F64X2) -> F64X2 {
    unsafe { vnegq_f64(x) }
}

pub(crate) fn neg0(x: F64X2) -> F64X2 {
    constant(-extract0(x), extract1(x))
}

pub(crate) fn or(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { transmute(vorrq_u64(transmute(x), transmute(y))) }
}

pub(crate) fn round(x: F64X2) -> F64X2 {
    unsafe { vrndaq_f64(x) }
}

pub(crate) fn round_ties_to_even(x: F64X2) -> F64X2 {
    unsafe { vrndnq_f64(x) }
}

pub(crate) fn shuffle02(x: F64X2, y: F64X2) -> F64X2 {
    constant(extract0(x), extract0(y))
}

pub(crate) fn shuffle03(x: F64X2, y: F64X2) -> F64X2 {
    constant(extract0(x), extract1(y))
}

pub(crate) fn shuffle13(x: F64X2, y: F64X2) -> F64X2 {
    constant(extract1(x), extract1(y))
}

pub(crate) fn splat(x: f64) -> F64X2 {
    unsafe { transmute([x, x]) }
}

pub(crate) fn sub_rn(x: F64X2, y: F64X2) -> F64X2 {
    unsafe { vsubq_f64(x, y) }
}

pub(crate) fn swap(x: F64X2) -> F64X2 {
    shuffle12(x, x)
}

pub(crate) fn trunc(x: F64X2) -> F64X2 {
    unsafe { vrndq_f64(x) }
}

fn shuffle12(x: F64X2, y: F64X2) -> F64X2 {
    constant(extract1(x), extract0(y))
}

macro_rules! impl_op_round {
    ($t:ty, $f:ident ($x:ident $(,$y:ident)*), $inst:literal, rd) => {
        impl_op_round!($t, $f ($x $(,$y)*), $inst, 0x800000);
    };

    ($t:ty, $f:ident ($x:ident $(,$y:ident)*), $inst:literal, ru) => {
        impl_op_round!($t, $f ($x $(,$y)*), $inst, 0x400000);
    };

    ($t:ty, $f:ident ($x:ident $(,$y:ident)*), $inst:literal, $fpcr:literal) => {
        pub(crate) fn $f(mut $x: $t, $($y: $t,)*) -> $t {
            unsafe {
                asm!(
                    "mrs {fpcr_bak}, fpcr",
                    "msr fpcr, {fpcr:x}",
                    $inst,
                    "msr fpcr, {fpcr_bak}",
                    $x = inout(vreg) $x,
                    $($y = in(vreg) $y,)*
                    fpcr = in(reg) $fpcr,
                    fpcr_bak = out(reg) _,
                    options(pure, nomem, nostack, preserves_flags)
                );
            }
            $x
        }
    };
}

impl_op_round!(f64, sqrt1_rd(x), "fsqrt {x:d}, {x:d}", rd);
impl_op_round!(f64, sqrt1_ru(x), "fsqrt {x:d}, {x:d}", ru);
impl_op_round!(f64, sub1_ru(x, y), "fsub {x:d}, {x:d}, {y:d}", ru);
impl_op_round!(F64X2, add_ru(x, y), "fadd.2d {x:v}, {x:v}, {y:v}", ru);
impl_op_round!(F64X2, sub_ru(x, y), "fsub.2d {x:v}, {x:v}, {y:v}", ru);
impl_op_round!(F64X2, mul_ru(x, y), "fmul.2d {x:v}, {x:v}, {y:v}", ru);
impl_op_round!(F64X2, div_ru(x, y), "fdiv.2d {x:v}, {x:v}, {y:v}", ru);

pub(crate) fn mul_add_ru(x: F64X2, y: F64X2, mut z: F64X2) -> F64X2 {
    unsafe {
        asm!(
            "mrs {fpcr_bak}, fpcr",
            "msr fpcr, {fpcr:x}",
            "fmla.2d {z:v}, {x:v}, {y:v}",
            "msr fpcr, {fpcr_bak}",
            x = in(vreg) x,
            y = in(vreg) y,
            z = inout(vreg) z,
            fpcr = in(reg) 0x400000,
            fpcr_bak = out(reg) _,
            options(pure, nomem, nostack, preserves_flags)
        );
    }
    z
}