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
|
//TEST:SIMPLE(filecheck=CUDA): -target cuda -line-directive-mode none
// This test just needs to compile without crashing.
// CUDA: __device__ void s_bwd_myKernel_0
[Differentiable]
uint3 load_foo(int32_t g_idx, TensorView<int32_t> indices)
{
return {
indices[uint2(g_idx, 0)],
indices[uint2(g_idx, 1)],
indices[uint2(g_idx, 2)]
};
}
[Differentiable]
Triangle load_triangle(DiffTensorView vertices, uint3 index_set)
{
float3[3] triangle = {
read_t3_float3(index_set.x, vertices),
read_t3_float3(index_set.y, vertices),
read_t3_float3(index_set.z, vertices)
};
return { triangle };
}
struct Triangle : IDifferentiable
{
float3[3] verts;
}
[Differentiable]
float3 read_t3_float3(uint32_t idx, DiffTensorView t3)
{
return float3(t3[uint2(idx, 0)],
t3[uint2(idx, 1)],
t3[uint2(idx, 2)]);
}
[Differentiable] TriangleFoo load_triangle_foo(int32_t g_idx, DiffTensorView vertices, TensorView<int32_t> indices, DiffTensorView vertex_color)
{
uint3 index_set = load_foo(g_idx, indices);
Triangle triangle = load_triangle(vertices, index_set);
float3 c0 = read_t3_float3(index_set.x, vertex_color);
float3 c1 = read_t3_float3(index_set.y, vertex_color);
float3 c2 = read_t3_float3(index_set.z, vertex_color);
TriangleFoo result = { triangle, 0.f, {c0, c1, c2} };
return result;
}
struct TriangleFoo : IDifferentiable
{
Triangle triangle;
float density;
float3 vertex_color[3];
};
[AutoPyBindCUDA]
[Differentiable]
[CudaKernel]
void myKernel(DiffTensorView vertices, TensorView<int32_t> indices, DiffTensorView vertex_color)
{
if (cudaThreadIdx().x > 0)
return;
TriangleFoo.Differential dp_g = TriangleFoo.dzero();
bwd_diff(load_triangle_foo)(cudaThreadIdx().x, vertices, indices, vertex_color, dp_g);
}
|
