File size: 8,919 Bytes
48001d3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
{
  "op": "ai.onnx.GlobalAveragePool",
  "cases": [
    {
      "name": "nchw_1x2048x7x7",
      "preset": "smoke",
      "inputs": { "x": { "dtype": "float32", "shape": [1, 2048, 7, 7] } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 2048, 1, 1] } }
    },
    {
      "name": "nchw_16x512x64x64_wg_parallel",
      "preset": "smoke",
      "vars": { "batch": 16, "channels": 512, "spatial": 4096 },
      "inputs": { "x": { "dtype": "float32", "shape": [16, 512, 64, 64], "dist": "normal", "seed": 201 } },
      "outputs": { "y": { "dtype": "float32", "shape": [16, 512, 1, 1], "dist": "empty" } },
      "bench": {
        "primary": true,
        "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }]
      }
    },
    {
      "name": "nchw_f16_8x256x32x32_wg_parallel",
      "preset": "smoke",
      "vars": { "dtype": "float16", "batch": 8, "channels": 256, "spatial": 1024 },
      "inputs": { "x": { "dtype": "float16", "shape": [8, 256, 32, 32], "dist": "normal", "seed": 203 } },
      "outputs": { "y": { "dtype": "float16", "shape": [8, 256, 1, 1], "dist": "empty" } },
      "bench": {
        "metrics": [
          { "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * dtypeBytes(args.dtype)" }
        ]
      }
    },
    {
      "name": "nchw_1x2048x7x7_smallspatial_scalar",
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 2048, "spatial": 49 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 2048, 7, 7], "dist": "normal", "seed": 211 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 2048, 1, 1], "dist": "empty" } },
      "bench": {
        "primary": true,
        "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }]
      }
    },
    {
      "name": "nchw_1x2048x8x8_spatial64_vec4_planeparallel",
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 2048, "spatial": 64 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 2048, 8, 8], "dist": "normal", "seed": 212 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 2048, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_1x1024x65x65_planeparallel_scalar_unaligned",
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 1024, "spatial": 4225 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 1024, 65, 65], "dist": "normal", "seed": 213 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 1024, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_1x1024x66x66_planeparallel_vec4_aligned",
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 1024, "spatial": 4356 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 1024, 66, 66], "dist": "normal", "seed": 214 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 1024, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_1x8x256x256_lowoccupancy_8planes",
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 8, "spatial": 65536 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 8, 256, 256], "dist": "normal", "seed": 215 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 8, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "f32_inner63_below_subgroup_gate_1x2048x7x9",
      "preset": "edge",
      "vars": { "batch": 1, "channels": 2048, "spatial": 63 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 2048, 7, 9], "dist": "normal", "seed": 301 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 2048, 1, 1], "dist": "empty" } },
      "bench": {
        "primary": true,
        "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }]
      }
    },
    {
      "name": "nchw_1x1024x14x14_portable_gate_window",
      "provenance": {
        "notes": "Spatial 196 sits in the window the portable plane-parallel gate excludes (65 with subgroups, 1024 without), so a subgroupless device falls back to the serial kernel here."
      },
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 1024, "spatial": 196 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 1024, 14, 14], "dist": "normal", "seed": 216 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 1024, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_1x512x28x28_portable_gate_window",
      "provenance": { "notes": "Spatial 784, the upper half of the same excluded window." },
      "preset": "smoke",
      "vars": { "batch": 1, "channels": 512, "spatial": 784 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 512, 28, 28], "dist": "normal", "seed": 217 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 512, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_32x128x16x16_manyplanes_spatial256",
      "preset": "smoke",
      "provenance": {
        "notes": "4096 planes of 256 elements: the serial kernel already has an invocation per plane, so this is the regime where the plane-parallel workgroup has the least to add. GlobalLpPool's same shape prefers the serial route."
      },
      "vars": { "batch": 32, "channels": 128, "spatial": 256 },
      "inputs": { "x": { "dtype": "float32", "shape": [32, 128, 16, 16], "dist": "normal", "seed": 218 } },
      "outputs": { "y": { "dtype": "float32", "shape": [32, 128, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_32x128x28x28_manyplanes_spatial784",
      "preset": "smoke",
      "provenance": {
        "notes": "Separates the two candidate discriminators: 4096 planes (where the serial kernel is saturated) at 784 elements per plane (where the workgroup tree amortizes)."
      },
      "vars": { "batch": 32, "channels": 128, "spatial": 784 },
      "inputs": { "x": { "dtype": "float32", "shape": [32, 128, 28, 28], "dist": "normal", "seed": 219 } },
      "outputs": { "y": { "dtype": "float32", "shape": [32, 128, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_32x64x16x16_planes2048_spatial256",
      "preset": "smoke",
      "provenance": {
        "notes": "Halves the plane count of the 4096-plane spatial-256 case: brackets where the serial kernel stops being saturated."
      },
      "vars": { "batch": 32, "channels": 64, "spatial": 256 },
      "inputs": { "x": { "dtype": "float32", "shape": [32, 64, 16, 16], "dist": "normal", "seed": 220 } },
      "outputs": { "y": { "dtype": "float32", "shape": [32, 64, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_32x128x16x32_manyplanes_spatial512",
      "preset": "smoke",
      "provenance": {
        "notes": "Doubles the spatial extent of the 4096-plane spatial-256 case: brackets where the workgroup tree starts amortizing."
      },
      "vars": { "batch": 32, "channels": 128, "spatial": 512 },
      "inputs": { "x": { "dtype": "float32", "shape": [32, 128, 16, 32], "dist": "normal", "seed": 221 } },
      "outputs": { "y": { "dtype": "float32", "shape": [32, 128, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    },
    {
      "name": "nchw_1x65536x9x9_planes65536_spatial81",
      "preset": "smoke",
      "provenance": {
        "notes": "65536 planes: four outputs per invocation still leaves 16384 invocations, which is the regime the vec4-output serial kernel is written for."
      },
      "vars": { "batch": 1, "channels": 65536, "spatial": 81 },
      "inputs": { "x": { "dtype": "float32", "shape": [1, 65536, 9, 9], "dist": "normal", "seed": 222 } },
      "outputs": { "y": { "dtype": "float32", "shape": [1, 65536, 1, 1], "dist": "empty" } },
      "bench": { "metrics": [{ "type": "bandwidth", "value": "args.batch * args.channels * args.spatial * 4" }] }
    }
  ],
  "tunableSpace": { "WORKGROUP_SIZE": [64, 128, 256] }
}