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
//use crate::common::validate_inputs;
use crate::indicators::simd_indicators::road_train::{Asset, Driver, PrimeMover};
use crate::indicators::simd_indicators::stochrsi_simd::assets::SimdState;
use crate::indicators::{
rsi::{multiplier, output_length as rsi_output_length},
stochrsi::{min_data, output_length, IndicatorState, State, INPUTS_WIDTH, OPTIONS_WIDTH},
};
use crate::types::IndicatorError;
use crate::{common::validate_options, common_simd::assets::validate_inputs};
use std::simd::Simd;
/// SIMD driver that advances the Stochastic RSI (STOCHRSI) across `N` asset lanes per scheduling epoch.
struct StochrsiDriver {
want_optional_outputs: bool,
period: usize,
multipliers: (f64, f64),
}
impl Driver<State> for StochrsiDriver {
/// Processes one epoch of bars for `N` assets simultaneously using SIMD.
fn next_run<const N: usize>(
&mut self,
inputs: Vec<Vec<&[f64]>>,
mut outputs: Vec<Vec<&mut [f64]>>,
mut states: Vec<&mut State>,
_options: Vec<Option<&()>>,
) {
let mut state = SimdState::<N>::new(&mut states);
let len = inputs[0][0].len();
let multipliers_simd = (
Simd::splat(self.multipliers.0),
Simd::splat(self.multipliers.1),
);
let want_rsi = self.want_optional_outputs;
// Optimization 1: Direct array construction instead of collect+try_into
//collect outputs
let (stochrsi_line_ptr, rsi_line_ptr) =
crate::extract_output_ptrs!(outputs, N, stochrsi_line_ptr, rsi_line_ptr);
// Optimization 2: Pre-compute all input and output pointers
let real_ptrs = crate::extract_input_ptrs!(inputs, N, real_ptrs);
match self.period {
1..=14 => {
for i in 0..len {
// Get inputs arrays for stocks
let real = crate::extract_simd_inputs_at_index!(
i,
N,
real @ real_ptrs
);
let (stochrsi, rsi) = state.calc_simd::<1>(real, multipliers_simd, self.period);
// Store results using pre-computed pointers
crate::write_simd_at_indices!(N, i,
stochrsi_line_ptr => stochrsi
);
crate::store_simd_optional_outputs!(i, N,
want_rsi, rsi_line_ptr => rsi
);
}
}
_ => {
for i in 0..len {
// Get inputs arrays for stocks
let real = crate::extract_simd_inputs_at_index!(
i,
N,
real @ real_ptrs
);
let (stochrsi, rsi) = state.calc_simd::<8>(real, multipliers_simd, self.period);
// Store results using pre-computed pointers
crate::write_simd_at_indices!(N, i,
stochrsi_line_ptr => stochrsi
);
crate::store_simd_optional_outputs!(i, N,
want_rsi, rsi_line_ptr => rsi
);
}
}
}
// Update states efficiently
state.write_states(&mut states);
}
}
/// Calculates the Stochastic RSI (STOCHRSI) for `N` assets simultaneously using SIMD
/// parallelism.
///
/// STOCHRSI applies the Stochastic Oscillator formula to RSI values, normalising RSI
/// relative to its own high/low range over the look-back period.
/// Uses the [`PrimeMover`] scheduler to batch assets into SIMD-width groups.
///
/// # Arguments
/// * `inputs` - An array of `N` asset input sets; `inputs[i]` is `[&[f64]; INPUTS_WIDTH]`
/// containing `[real]` for asset `i`.
/// * `options` - `[period]` — the look-back period for both the inner RSI and
/// the Stochastic normalisation.
/// * `optional_outputs` - Optional slice of booleans enabling extra outputs:
/// `[0]` → `rsi`.
///
/// # Returns
/// `Ok((outputs, states))` where `outputs[i][0]` is the STOCHRSI line and
/// `outputs[i][1]` is the RSI line (empty unless requested) for asset `i`.
/// `states[i]` is the final [`IndicatorState`] for asset `i`.
/// Returns `Err(IndicatorError)` if any input slice is too short or options are invalid.
pub fn indicator_by_assets<const N: usize>(
inputs: &[&[&[f64]; INPUTS_WIDTH]; N], //stock[ fields [ field [f64] ] ]
options: &[f64; OPTIONS_WIDTH],
optional_outputs: Option<&[bool]>,
) -> Result<(Vec<Vec<Vec<f64>>>, Vec<IndicatorState>), IndicatorError> {
validate_inputs::<INPUTS_WIDTH>(inputs, min_data(options))?;
validate_options(options)?;
let (period, multipliers) = {
let period = options[0] as usize;
(period, multiplier(period))
};
let mut road_train = PrimeMover::<N, State>::new();
let mut want_optional_outputs = false;
let mut output_buffers = Vec::with_capacity(N);
for i in 0..N {
let asset_inputs = vec![
inputs[i][0], // real
];
let (stochrsi_line, mut rsi_line);
{
let len = inputs[i][0].len();
let capacity = output_length(len, options);
stochrsi_line = crate::uninit_vec!(f64, capacity);
let rsi_capacity = rsi_output_length(len, options);
rsi_line = crate::init_optional_outputs_eff!(
optional_outputs, &[false],
rsi_line: rsi_capacity
);
}
let state = State::init_state(&inputs[i][0], period, &mut rsi_line);
if i == 0 {
(want_optional_outputs, _) = crate::calc_want_flags!(rsi_line);
}
let mut starts = [0; 2];
starts[1] = crate::slice_outputs_start!(stochrsi_line.len(), rsi_line);
let mut output_buffer = vec![stochrsi_line, rsi_line];
//let adosc_len = output_buffer[0].len();
let mut asset_outputs = Vec::with_capacity(output_buffer.len());
for j in 0..output_buffer.len() {
unsafe {
//let slice_len = output_buffer.len() - starts[j];
// Get a mutable reference to the output buffer for this asset
let output_buffer = &mut output_buffer[j];
asset_outputs.push(std::slice::from_raw_parts_mut(
output_buffer.as_mut_ptr().add(starts[j]), //slice from
output_buffer.len() - starts[j], // slice to
));
}
}
road_train.add_asset(Asset::new(
asset_inputs,
asset_outputs,
i,
period * 2,
0,
state,
None,
));
output_buffers.push(output_buffer);
}
let mut driver = StochrsiDriver {
period,
want_optional_outputs,
multipliers,
};
let states_vec = road_train.drive(&mut driver);
let mut states = Vec::with_capacity(N);
for state in states_vec.into_iter() {
states.push(IndicatorState::new(state, period, multipliers));
}
Ok((output_buffers, states))
}