oxigdal-gpu-advanced 0.1.4

Advanced GPU computing with multi-GPU support, memory pooling, and shader optimization for OxiGDAL
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

// Histogram equalization for image enhancement

@group(0) @binding(0) var<storage, read> input_image: array<f32>;
@group(0) @binding(1) var<storage, read_write> output_image: array<f32>;
@group(0) @binding(2) var<storage, read_write> histogram: array<atomic<u32>>;
@group(0) @binding(3) var<storage, read_write> cdf: array<u32>;
@group(0) @binding(4) var<uniform> params: HistogramParams;

struct HistogramParams {
    width: u32,
    height: u32,
    bins: u32,
    min_val: f32,
    max_val: f32,
    pad: array<u32, 3>,
}

// Step 1: Compute histogram
@compute @workgroup_size(16, 16, 1)
fn compute_histogram(
    @builtin(global_invocation_id) global_id: vec3<u32>,
) {
    let x = global_id.x;
    let y = global_id.y;

    if (x >= params.width || y >= params.height) {
        return;
    }

    let idx = y * params.width + x;
    let value = input_image[idx];

    // Normalize value to bin index
    let normalized = (value - params.min_val) / (params.max_val - params.min_val);
    let bin = u32(clamp(normalized * f32(params.bins - 1u), 0.0, f32(params.bins - 1u)));

    atomicAdd(&histogram[bin], 1u);
}

// Step 2: Compute cumulative distribution function (CDF)
@compute @workgroup_size(256, 1, 1)
fn compute_cdf(
    @builtin(global_invocation_id) global_id: vec3<u32>,
) {
    let bin = global_id.x;

    if (bin >= params.bins) {
        return;
    }

    var sum = 0u;
    for (var i = 0u; i <= bin; i = i + 1u) {
        sum = sum + atomicLoad(&histogram[i]);
    }

    cdf[bin] = sum;
}

// Step 3: Equalize using CDF
@compute @workgroup_size(16, 16, 1)
fn histogram_equalize(
    @builtin(global_invocation_id) global_id: vec3<u32>,
) {
    let x = global_id.x;
    let y = global_id.y;

    if (x >= params.width || y >= params.height) {
        return;
    }

    let idx = y * params.width + x;
    let value = input_image[idx];

    // Find bin
    let normalized = (value - params.min_val) / (params.max_val - params.min_val);
    let bin = u32(clamp(normalized * f32(params.bins - 1u), 0.0, f32(params.bins - 1u)));

    // Apply equalization
    let total_pixels = params.width * params.height;
    let cdf_min = cdf[0];
    let cdf_val = cdf[bin];

    let equalized = f32(cdf_val - cdf_min) / f32(total_pixels - cdf_min);
    output_image[idx] = params.min_val + equalized * (params.max_val - params.min_val);
}

// Adaptive histogram equalization (CLAHE - simplified)
@compute @workgroup_size(8, 8, 1)
fn clahe(
    @builtin(global_invocation_id) global_id: vec3<u32>,
    @builtin(local_invocation_id) local_id: vec3<u32>,
) {
    let x = global_id.x;
    let y = global_id.y;

    if (x >= params.width || y >= params.height) {
        return;
    }

    // Define local region (simplified - would use actual tiles in full implementation)
    let region_size = 32u;
    let local_x_start = (x / region_size) * region_size;
    let local_y_start = (y / region_size) * region_size;

    // Compute local histogram
    var local_hist: array<u32, 256>;
    for (var i = 0u; i < 256u; i = i + 1u) {
        local_hist[i] = 0u;
    }

    for (var dy = 0u; dy < region_size; dy = dy + 1u) {
        for (var dx = 0u; dx < region_size; dx = dx + 1u) {
            let px = local_x_start + dx;
            let py = local_y_start + dy;

            if (px < params.width && py < params.height) {
                let pidx = py * params.width + px;
                let pval = input_image[pidx];
                let normalized = (pval - params.min_val) / (params.max_val - params.min_val);
                let bin = u32(clamp(normalized * 255.0, 0.0, 255.0));
                local_hist[bin] = local_hist[bin] + 1u;
            }
        }
    }

    // Apply local equalization
    let idx = y * params.width + x;
    let value = input_image[idx];
    let normalized = (value - params.min_val) / (params.max_val - params.min_val);
    let bin = u32(clamp(normalized * 255.0, 0.0, 255.0));

    // Compute local CDF
    var local_cdf = 0u;
    for (var i = 0u; i <= bin; i = i + 1u) {
        local_cdf = local_cdf + local_hist[i];
    }

    let total = region_size * region_size;
    let equalized = f32(local_cdf) / f32(total);
    output_image[idx] = params.min_val + equalized * (params.max_val - params.min_val);
}