splimes 0.1.0

Spline interpolation over irregularly sampled time series, on SIMD/parallel CPU or GPU (wgpu), with BigDecimal values.
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

/// GPU compute shader for linear interpolation with f64 precision
pub const LINEAR_INTERPOLATION_SHADER_F64: &str = r"
@group(0) @binding(0) var<storage, read> input_times: array<f64>;
@group(0) @binding(1) var<storage, read> input_values: array<f64>;
@group(0) @binding(2) var<storage, read> target_times: array<f64>;
@group(0) @binding(3) var<storage, read_write> output_values: array<f64>;

fn is_near_zero(val: f64) -> bool {
    return abs(val) < f64(1e-12);
}

fn extrapolate(t: f64, t0: f64, v0: f64, t1: f64, v1: f64) -> f64 {
    let dt = t1 - t0;
    if (is_near_zero(dt)) {
        return v0;
    }
    let slope = (v1 - v0) / dt;
    return v0 + slope * (t - t0);
}

@compute @workgroup_size(256)
fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
    let index = global_id.x;
    let target_count = arrayLength(&target_times);
    
    if (index >= target_count) {
        return;
    }
    
    let target_time = target_times[index];
    let input_count = arrayLength(&input_times);
    
    if (input_count == 0u) {
        output_values[index] = f64(0.0);
        return;
    }
    if (input_count == 1u) {
        output_values[index] = input_values[0];
        return;
    }
    
    if (target_time < input_times[0]) {
        output_values[index] = extrapolate(target_time, input_times[0], input_values[0], input_times[1], input_values[1]);
        return;
    }
    
    if (target_time > input_times[input_count - 1u]) {
        output_values[index] = extrapolate(target_time, input_times[input_count - 2u], input_values[input_count - 2u], input_times[input_count - 1u], input_values[input_count - 1u]);
        return;
    }
    
    var left = 0u;
    var right = input_count - 1u;
    
    while (right - left > 1u) {
        let mid = (left + right) / 2u;
        if (target_time < input_times[mid]) {
            right = mid;
        } else {
            left = mid;
        }
    }
    
    let left_idx = left;
    let right_idx = right;
    
    if (abs(target_time - input_times[left_idx]) < f64(1e-12)) {
        output_values[index] = input_values[left_idx];
        return;
    }
    if (abs(target_time - input_times[right_idx]) < f64(1e-12)) {
        output_values[index] = input_values[right_idx];
        return;
    }
    
    let t0 = input_times[left_idx];
    let t1 = input_times[right_idx];
    let v0 = input_values[left_idx];
    let v1 = input_values[right_idx];
    
    let dt = t1 - t0;
    if (is_near_zero(dt)) {
        output_values[index] = v0;
        return;
    }
    
    let alpha = (target_time - t0) / dt;
    output_values[index] = v0 + alpha * (v1 - v0);
}
";

/// GPU compute shader for linear interpolation with f32 precision (fallback)
pub const LINEAR_INTERPOLATION_SHADER_F32: &str = r"
@group(0) @binding(0) var<storage, read> input_times: array<f32>;
@group(0) @binding(1) var<storage, read> input_values: array<f32>;
@group(0) @binding(2) var<storage, read> target_times: array<f32>;
@group(0) @binding(3) var<storage, read_write> output_values: array<f32>;

fn is_near_zero(val: f32) -> bool {
    return abs(val) < 1e-8;
}

fn extrapolate(t: f32, t0: f32, v0: f32, t1: f32, v1: f32) -> f32 {
    let dt = t1 - t0;
    if (is_near_zero(dt)) {
        return v0;
    }
    let slope = (v1 - v0) / dt;
    let result = v0 + slope * (t - t0);
    return clamp(result, -1e6, 1e6); // Prevent overflow
}

@compute @workgroup_size(256)
fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
    let index = global_id.x;
    let target_count = arrayLength(&target_times);
    
    if (index >= target_count) {
        return;
    }
    
    let target_time = target_times[index];
    let input_count = arrayLength(&input_times);
    
    if (input_count == 0u) {
        output_values[index] = 0.0;
        return;
    }
    if (input_count == 1u) {
        output_values[index] = input_values[0];
        return;
    }
    
    if (target_time < input_times[0]) {
        output_values[index] = extrapolate(target_time, input_times[0], input_values[0], input_times[1], input_values[1]);
        return;
    }
    
    if (target_time > input_times[input_count - 1u]) {
        output_values[index] = extrapolate(target_time, input_times[input_count - 2u], input_values[input_count - 2u], input_times[input_count - 1u], input_values[input_count - 1u]);
        return;
    }
    
    var left = 0u;
    var right = input_count - 1u;
    
    while (right - left > 1u) {
        let mid = (left + right) / 2u;
        if (target_time < input_times[mid]) {
            right = mid;
        } else {
            left = mid;
        }
    }
    
    let left_idx = left;
    let right_idx = right;
    
    if (abs(target_time - input_times[left_idx]) < 1e-8) {
        output_values[index] = input_values[left_idx];
        return;
    }
    if (abs(target_time - input_times[right_idx]) < 1e-8) {
        output_values[index] = input_values[right_idx];
        return;
    }
    
    let t0 = input_times[left_idx];
    let t1 = input_times[right_idx];
    let v0 = input_values[left_idx];
    let v1 = input_values[right_idx];
    
    let dt = t1 - t0;
    if (is_near_zero(dt)) {
        output_values[index] = (v0 + v1) / 2.0;
        return;
    }
    
    let alpha = (target_time - t0) / dt;
    let result = v0 + alpha * (v1 - v0);
    output_values[index] = clamp(result, -1e6, 1e6); // Prevent overflow
}
";