fast-ta 0.2.0

High-performance technical analysis indicators with batch, prepared, and streaming APIs
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

//! ARM NEON SIMD implementation for AArch64

use crate::simd::scalar;
use crate::types::Float;
use crate::Result;

#[inline(never)]
#[target_feature(enable = "neon")]
#[allow(dead_code)]
pub unsafe fn sum(data: &[Float]) -> Float {
    data.iter().copied().sum()
}

#[inline(never)]
#[target_feature(enable = "neon")]
#[allow(dead_code)]
pub unsafe fn dot_product(a: &[Float], b: &[Float]) -> Result<Float> {
    if a.len() != b.len() {
        return Err(crate::TalibError::InvalidInput {
            message: "Dot product requires vectors of equal length".into(),
        });
    }
    let mut sum = 0.0 as Float;
    for (&x, &y) in a.iter().zip(b.iter()) {
        sum += x * y;
    }
    Ok(sum)
}

#[cfg(not(feature = "f32"))]
#[inline(never)]
#[target_feature(enable = "neon")]
pub unsafe fn first_non_finite(values: &[Float]) -> Option<usize> {
    use core::arch::aarch64::{
        uint64x2_t, vandq_u64, vceqq_u64, vdupq_n_u64, vgetq_lane_u64, vld1q_f64, vorrq_u64,
        vreinterpretq_u64_f64,
    };

    const LANES: usize = 2;
    const UNROLLED_LANES: usize = LANES * 4;
    const EXPONENT_MASK: u64 = 0x7ff0_0000_0000_0000;

    #[inline(always)]
    unsafe fn invalid_lanes(values: *const f64, exponent_mask: uint64x2_t) -> uint64x2_t {
        let bits = vreinterpretq_u64_f64(vld1q_f64(values));
        vceqq_u64(vandq_u64(bits, exponent_mask), exponent_mask)
    }

    let exponent_mask = vdupq_n_u64(EXPONENT_MASK);
    let mut index = 0;
    while index + UNROLLED_LANES <= values.len() {
        let ptr = values.as_ptr().add(index);
        let first = invalid_lanes(ptr, exponent_mask);
        let second = invalid_lanes(ptr.add(LANES), exponent_mask);
        let third = invalid_lanes(ptr.add(LANES * 2), exponent_mask);
        let fourth = invalid_lanes(ptr.add(LANES * 3), exponent_mask);
        let invalid = vorrq_u64(vorrq_u64(first, second), vorrq_u64(third, fourth));
        if (vgetq_lane_u64::<0>(invalid) | vgetq_lane_u64::<1>(invalid)) != 0 {
            return scalar::first_non_finite(&values[index..index + UNROLLED_LANES])
                .map(|offset| index + offset);
        }
        index += UNROLLED_LANES;
    }

    while index + LANES <= values.len() {
        let invalid = invalid_lanes(values.as_ptr().add(index), exponent_mask);
        if (vgetq_lane_u64::<0>(invalid) | vgetq_lane_u64::<1>(invalid)) != 0 {
            return scalar::first_non_finite(&values[index..index + LANES])
                .map(|offset| index + offset);
        }
        index += LANES;
    }

    scalar::first_non_finite(&values[index..]).map(|offset| index + offset)
}

#[cfg(feature = "f32")]
#[inline(never)]
#[target_feature(enable = "neon")]
pub unsafe fn first_non_finite(values: &[Float]) -> Option<usize> {
    use core::arch::aarch64::{
        uint32x4_t, vandq_u32, vceqq_u32, vdupq_n_u32, vld1q_f32, vmaxvq_u32, vorrq_u32,
        vreinterpretq_u32_f32,
    };

    const LANES: usize = 4;
    const UNROLLED_LANES: usize = LANES * 4;
    const EXPONENT_MASK: u32 = 0x7f80_0000;

    #[inline(always)]
    unsafe fn invalid_lanes(values: *const f32, exponent_mask: uint32x4_t) -> uint32x4_t {
        let bits = vreinterpretq_u32_f32(vld1q_f32(values));
        vceqq_u32(vandq_u32(bits, exponent_mask), exponent_mask)
    }

    let exponent_mask = vdupq_n_u32(EXPONENT_MASK);
    let mut index = 0;
    while index + UNROLLED_LANES <= values.len() {
        let ptr = values.as_ptr().add(index);
        let first = invalid_lanes(ptr, exponent_mask);
        let second = invalid_lanes(ptr.add(LANES), exponent_mask);
        let third = invalid_lanes(ptr.add(LANES * 2), exponent_mask);
        let fourth = invalid_lanes(ptr.add(LANES * 3), exponent_mask);
        let invalid = vorrq_u32(vorrq_u32(first, second), vorrq_u32(third, fourth));
        if vmaxvq_u32(invalid) != 0 {
            return scalar::first_non_finite(&values[index..index + UNROLLED_LANES])
                .map(|offset| index + offset);
        }
        index += UNROLLED_LANES;
    }

    while index + LANES <= values.len() {
        if vmaxvq_u32(invalid_lanes(values.as_ptr().add(index), exponent_mask)) != 0 {
            return scalar::first_non_finite(&values[index..index + LANES])
                .map(|offset| index + offset);
        }
        index += LANES;
    }

    scalar::first_non_finite(&values[index..]).map(|offset| index + offset)
}

#[cfg(not(feature = "f32"))]
#[inline(never)]
#[target_feature(enable = "neon")]
pub unsafe fn typical_price(high: &[Float], low: &[Float], close: &[Float], output: &mut [Float]) {
    use core::arch::aarch64::{vaddq_f64, vdivq_f64, vdupq_n_f64, vld1q_f64, vst1q_f64};

    const LANES: usize = 2;
    let denominator = vdupq_n_f64(3.0);
    let mut index = 0;
    while index + LANES <= high.len() {
        let high_values = vld1q_f64(high.as_ptr().add(index));
        let low_values = vld1q_f64(low.as_ptr().add(index));
        let close_values = vld1q_f64(close.as_ptr().add(index));
        let values = vdivq_f64(
            vaddq_f64(vaddq_f64(high_values, low_values), close_values),
            denominator,
        );
        vst1q_f64(output.as_mut_ptr().add(index), values);
        index += LANES;
    }

    while index < high.len() {
        output[index] = (high[index] + low[index] + close[index]) / 3.0;
        index += 1;
    }
}

#[cfg(feature = "f32")]
#[inline(never)]
#[target_feature(enable = "neon")]
pub unsafe fn typical_price(high: &[Float], low: &[Float], close: &[Float], output: &mut [Float]) {
    use core::arch::aarch64::{vaddq_f32, vdivq_f32, vdupq_n_f32, vld1q_f32, vst1q_f32};

    const LANES: usize = 4;
    let denominator = vdupq_n_f32(3.0);
    let mut index = 0;
    while index + LANES <= high.len() {
        let high_values = vld1q_f32(high.as_ptr().add(index));
        let low_values = vld1q_f32(low.as_ptr().add(index));
        let close_values = vld1q_f32(close.as_ptr().add(index));
        let values = vdivq_f32(
            vaddq_f32(vaddq_f32(high_values, low_values), close_values),
            denominator,
        );
        vst1q_f32(output.as_mut_ptr().add(index), values);
        index += LANES;
    }

    while index < high.len() {
        output[index] = (high[index] + low[index] + close[index]) / 3.0;
        index += 1;
    }
}