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
|
#version 450 core
#extension GL_KHR_memory_scope_semantics : enable
#extension GL_KHR_cooperative_matrix : enable
#extension GL_EXT_shader_explicit_arithmetic_types : enable
#extension GL_NV_cooperative_matrix2 : enable
#extension GL_EXT_buffer_reference : enable
layout (local_size_x = 64, local_size_y = 1, local_size_z = 1) in;
buffer BufType {
float16_t arr[256][256];
float16_t x[];
} Buf;
layout(buffer_reference, std430, buffer_reference_align = 2) buffer fp16Buf {
float16_t f;
};
float16_t decode(const in fp16Buf b, const in uint32_t blockCoords[2], const in uint32_t coordInBlock[2])
{
return b.f;
}
struct S {
f16vec2 x;
};
layout(std430, binding = 0) buffer SBuf {
S s[];
} sbuf;
layout(constant_id = 0) const uint32_t Clamp = gl_CooperativeMatrixClampModeConstantNV;
layout(constant_id = 1) const int32_t Clamp_signed = gl_CooperativeMatrixClampModeConstantNV;
void main()
{
coopmat<float16_t, gl_ScopeWorkgroup, 64, 32, gl_MatrixUseA> A;
tensorLayoutNV<2> t = createTensorLayoutNV(2);
tensorLayoutNV<3, 1> t2 = createTensorLayoutNV(3, 1);
t = setTensorLayoutBlockSizeNV(t, 4, 8);
t = setTensorLayoutDimensionNV(t, 256, 512);
t = sliceTensorLayoutNV(t, 128, 32, 256, 32);
tensorViewNV<5> v = createTensorViewNV(5);
v = setTensorViewDimensionsNV(v, 10, 11, 12, 13, 14);
v = setTensorViewStrideNV(v, 10, 11, 12, 13, 15);
v = setTensorViewClipNV(v, 0, 16, 0, 16);
tensorViewNV<5, true> v2 = createTensorViewNV(5, true);
tensorViewNV<2, true, 1, 0> v3 = createTensorViewNV(2, true, 1, 0);
coopMatLoadTensorNV(A, Buf.x, 0, t);
coopMatStoreTensorNV(A, Buf.x, 0, t);
coopMatLoadTensorNV(A, Buf.x, 0, t, v);
coopMatStoreTensorNV(A, Buf.x, 0, t, v);
coopMatLoadTensorNV(A, Buf.x, 0, t, decode);
coopMatLoadTensorNV(A, Buf.x, 0, t, v, decode);
coopMatLoadTensorNV(A, sbuf.s, 1, t, v, decode);
tensorLayoutNV<2, Clamp> tc = createTensorLayoutNV(2, Clamp);
coopmat<float16_t, gl_ScopeWorkgroup, 64, 32, gl_MatrixUseA> Arr[2];
coopMatLoadTensorNV(Arr[1], Buf.x, 0, t);
coopMatLoadTensorNV(A, Buf.arr[8], 0, t);
coopMatStoreTensorNV(A, Buf.arr[16], 0, t);
tensorLayoutNV<2, Clamp_signed> tc_s = createTensorLayoutNV(2, Clamp_signed);
tensorViewNV<Clamp, true> v4 = createTensorViewNV(Clamp, true);
}
|
