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
|
// CNN v3 — Decoder level 1
// NearestUp2x(bottleneck) + cat(enc1_skip) -> Conv(32->8, 3x3) + FiLM + ReLU
//
// Inputs: bn_tex_lo (rgba32uint, 8xf16) quarter-res ch 0-7
// bn_tex_hi (rgba32uint, 8xf16) quarter-res ch 8-15
// enc1_tex_lo (rgba32uint, 8xf16) half-res skip ch 0-7
// enc1_tex_hi (rgba32uint, 8xf16) half-res skip ch 8-15
// Output: dec1_out (rgba32uint, 8xf16) half-res
//
// Weight layout (f16, OIHW + bias):
// [0 .. 32*8*9) conv: w[out][in][ky][kx] (in=32: 16 bn + 16 enc1 skip)
// [2304 .. +8) bias: b[out]
#include "cnn_v3/common"
const DEC1_IN: u32 = 32u;
const DEC1_OUT: u32 = 8u;
struct Params {
weight_offset: u32,
_pad: vec3u,
gamma_lo: vec4f,
gamma_hi: vec4f,
beta_lo: vec4f,
beta_hi: vec4f,
}
@group(0) @binding(0) var bn_tex_lo: texture_2d<u32>;
@group(0) @binding(1) var bn_tex_hi: texture_2d<u32>;
@group(0) @binding(2) var enc1_tex_lo: texture_2d<u32>;
@group(0) @binding(3) var enc1_tex_hi: texture_2d<u32>;
@group(0) @binding(4) var<storage, read> weights: array<u32>;
@group(0) @binding(5) var<uniform> params: Params;
@group(0) @binding(6) var dec1_out: texture_storage_2d<rgba32uint, write>;
fn film_gamma(o: u32) -> f32 {
if (o < 4u) { return params.gamma_lo[o]; }
return params.gamma_hi[o - 4u];
}
fn film_beta(o: u32) -> f32 {
if (o < 4u) { return params.beta_lo[o]; }
return params.beta_hi[o - 4u];
}
// Load 32ch: [bn_nearest_up(16ch), enc1_skip(16ch)]
fn load_dec1_concat(hcoord: vec2i, half_dims: vec2i) -> array<f32, 32> {
var r: array<f32, 32>;
if (hcoord.x < 0 || hcoord.y < 0 || hcoord.x >= half_dims.x || hcoord.y >= half_dims.y) {
return r;
}
let quart_dims = half_dims / 2;
let qc = clamp(hcoord / 2, vec2i(0), quart_dims - vec2i(1));
let blo = unpack_8ch(bn_tex_lo, qc);
let bhi = unpack_8ch(bn_tex_hi, qc);
let slo = unpack_8ch(enc1_tex_lo, hcoord);
let shi = unpack_8ch(enc1_tex_hi, hcoord);
for (var i: u32 = 0u; i < 8u; i++) {
r[i] = blo[i];
r[i + 8u] = bhi[i];
r[i + 16u] = slo[i];
r[i + 24u] = shi[i];
}
return r;
}
@compute @workgroup_size(8, 8)
fn dec1_main(@builtin(global_invocation_id) id: vec3u) {
let half_dims = vec2i(textureDimensions(enc1_tex_lo));
let coord = vec2i(id.xy);
if (coord.x >= half_dims.x || coord.y >= half_dims.y) { return; }
let wo = params.weight_offset;
var out: array<f32, DEC1_OUT>;
for (var o: u32 = 0u; o < DEC1_OUT; o++) {
var sum = get_w(wo, DEC1_OUT * DEC1_IN * 9u + o); // bias
for (var ky: i32 = -1; ky <= 1; ky++) {
for (var kx: i32 = -1; kx <= 1; kx++) {
let feat = load_dec1_concat(coord + vec2i(kx, ky), half_dims);
let ki = u32(ky + 1) * 3u + u32(kx + 1);
for (var i: u32 = 0u; i < DEC1_IN; i++) {
sum += get_w(wo, o * DEC1_IN * 9u + i * 9u + ki) * feat[i];
}
}
}
out[o] = max(0.0, film_gamma(o) * sum + film_beta(o));
}
textureStore(dec1_out, coord, vec4u(
pack2x16float(vec2f(out[0], out[1])),
pack2x16float(vec2f(out[2], out[3])),
pack2x16float(vec2f(out[4], out[5])),
pack2x16float(vec2f(out[6], out[7]))
));
}
|
