runmat-accelerate 0.6.0

Pluggable GPU acceleration layer for RunMat (CUDA, ROCm, Metal, Vulkan/Spir-V)
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

pub const LINEAR_GATHER_SHADER_F64: &str = r#"
struct Tensor {
  data: array<f64>
};
struct Indices {
  data: array<u32>
};
struct Params {
  count: u32,
  lane_factor: u32,
  _pad1: u32,
  _pad2: u32,
};
@group(0) @binding(0) var<storage, read> Source: Tensor;
@group(0) @binding(1) var<storage, read> Ind: Indices;
@group(0) @binding(2) var<storage, read_write> Out: Tensor;
@group(0) @binding(3) var<uniform> Pm: Params;
@compute @workgroup_size(@WG@)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
  let idx = gid.x;
  if (idx >= Pm.count) {
    return;
  }
  let lin = Ind.data[idx];
  let src = lin * Pm.lane_factor;
  let dst = idx * Pm.lane_factor;
  var lane = 0u;
  loop {
    if (lane >= Pm.lane_factor) {
      break;
    }
    Out.data[dst + lane] = Source.data[src + lane];
    lane = lane + 1u;
  }
}
"#;

pub const LINEAR_GATHER_SHADER_F32: &str = r#"
struct Tensor {
  data: array<f32>
};
struct Indices {
  data: array<u32>
};
struct Params {
  count: u32,
  lane_factor: u32,
  _pad1: u32,
  _pad2: u32,
};
@group(0) @binding(0) var<storage, read> Source: Tensor;
@group(0) @binding(1) var<storage, read> Ind: Indices;
@group(0) @binding(2) var<storage, read_write> Out: Tensor;
@group(0) @binding(3) var<uniform> Pm: Params;
@compute @workgroup_size(@WG@)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
  let idx = gid.x;
  if (idx >= Pm.count) {
    return;
  }
  let lin = Ind.data[idx];
  let src = lin * Pm.lane_factor;
  let dst = idx * Pm.lane_factor;
  var lane = 0u;
  loop {
    if (lane >= Pm.lane_factor) {
      break;
    }
    Out.data[dst + lane] = Source.data[src + lane];
    lane = lane + 1u;
  }
}
"#;

pub const LINEAR_SCATTER_SHADER_F64: &str = r#"
struct Tensor {
  data: array<f64>
};
struct Indices {
  data: array<u32>
};
struct Params {
  count: u32,
  lane_factor: u32,
  _pad1: u32,
  _pad2: u32,
};
@group(0) @binding(0) var<storage, read_write> Target: Tensor;
@group(0) @binding(1) var<storage, read> Values: Tensor;
@group(0) @binding(2) var<storage, read> Ind: Indices;
@group(0) @binding(3) var<uniform> Pm: Params;
@compute @workgroup_size(@WG@)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
  let idx = gid.x;
  if (idx >= Pm.count) {
    return;
  }
  let lin = Ind.data[idx];
  let dst = lin * Pm.lane_factor;
  let src = idx * Pm.lane_factor;
  var lane = 0u;
  loop {
    if (lane >= Pm.lane_factor) {
      break;
    }
    Target.data[dst + lane] = Values.data[src + lane];
    lane = lane + 1u;
  }
}
"#;

pub const LINEAR_SCATTER_SHADER_F32: &str = r#"
struct Tensor {
  data: array<f32>
};
struct Indices {
  data: array<u32>
};
struct Params {
  count: u32,
  lane_factor: u32,
  _pad1: u32,
  _pad2: u32,
};
@group(0) @binding(0) var<storage, read_write> Target: Tensor;
@group(0) @binding(1) var<storage, read> Values: Tensor;
@group(0) @binding(2) var<storage, read> Ind: Indices;
@group(0) @binding(3) var<uniform> Pm: Params;
@compute @workgroup_size(@WG@)
fn main(@builtin(global_invocation_id) gid: vec3<u32>) {
  let idx = gid.x;
  if (idx >= Pm.count) {
    return;
  }
  let lin = Ind.data[idx];
  let dst = lin * Pm.lane_factor;
  let src = idx * Pm.lane_factor;
  var lane = 0u;
  loop {
    if (lane >= Pm.lane_factor) {
      break;
    }
    Target.data[dst + lane] = Values.data[src + lane];
    lane = lane + 1u;
  }
}
"#;