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
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
|
export const description = `
Stress tests covering GPUComputePassEncoder usage.
`;
import { makeTestGroup } from '../../common/framework/test_group.js';
import { assert, iterRange } from '../../common/util/util.js';
import { GPUTest } from '../../webgpu/gpu_test.js';
export const g = makeTestGroup(GPUTest);
g.test('many')
.desc(
`Tests execution of a huge number of compute passes using the same
GPUComputePipeline.`
)
.fn(async t => {
const kNumElements = 64;
const data = new Uint32Array([...iterRange(kNumElements, x => x)]);
const buffer = t.makeBufferWithContents(data, GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC);
const pipeline = t.device.createComputePipeline({
layout: 'auto',
compute: {
module: t.device.createShaderModule({
code: `
struct Buffer { data: array<u32>, };
@group(0) @binding(0) var<storage, read_write> buffer: Buffer;
@compute @workgroup_size(1) fn main(
@builtin(global_invocation_id) id: vec3<u32>) {
buffer.data[id.x] = buffer.data[id.x] + 1u;
}
`,
}),
entryPoint: 'main',
},
});
const bindGroup = t.device.createBindGroup({
layout: pipeline.getBindGroupLayout(0),
entries: [{ binding: 0, resource: { buffer } }],
});
const kNumIterations = 250_000;
for (let i = 0; i < kNumIterations; ++i) {
const encoder = t.device.createCommandEncoder();
const pass = encoder.beginComputePass();
pass.setPipeline(pipeline);
pass.setBindGroup(0, bindGroup);
pass.dispatchWorkgroups(kNumElements);
pass.end();
t.device.queue.submit([encoder.finish()]);
}
t.expectGPUBufferValuesEqual(
buffer,
new Uint32Array([...iterRange(kNumElements, x => x + kNumIterations)])
);
});
g.test('pipeline_churn')
.desc(
`Tests execution of a huge number of compute passes which each use a different
GPUComputePipeline.`
)
.fn(async t => {
const buffer = t.makeBufferWithContents(
new Uint32Array([0]),
GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC
);
const kNumIterations = 10_000;
const stages = iterRange(kNumIterations, i => ({
module: t.device.createShaderModule({
code: `
struct Buffer { data: u32, };
@group(0) @binding(0) var<storage, read_write> buffer: Buffer;
@compute @workgroup_size(1) fn main${i}() {
buffer.data = buffer.data + 1u;
}
`,
}),
entryPoint: `main${i}`,
}));
for (const compute of stages) {
const encoder = t.device.createCommandEncoder();
const pipeline = t.device.createComputePipeline({ layout: 'auto', compute });
const bindGroup = t.device.createBindGroup({
layout: pipeline.getBindGroupLayout(0),
entries: [{ binding: 0, resource: { buffer } }],
});
const pass = encoder.beginComputePass();
pass.setPipeline(pipeline);
pass.setBindGroup(0, bindGroup);
pass.dispatchWorkgroups(1);
pass.end();
t.device.queue.submit([encoder.finish()]);
}
t.expectGPUBufferValuesEqual(buffer, new Uint32Array([kNumIterations]));
});
g.test('bind_group_churn')
.desc(
`Tests execution of compute passes which switch between a huge number of bind
groups.`
)
.fn(async t => {
const kNumElements = 64;
const data = new Uint32Array([...iterRange(kNumElements, x => x)]);
const buffer1 = t.makeBufferWithContents(
data,
GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC
);
const buffer2 = t.makeBufferWithContents(
data,
GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC
);
const module = t.device.createShaderModule({
code: `
struct Buffer { data: array<u32>, };
@group(0) @binding(0) var<storage, read_write> buffer1: Buffer;
@group(0) @binding(1) var<storage, read_write> buffer2: Buffer;
@compute @workgroup_size(1) fn main(
@builtin(global_invocation_id) id: vec3<u32>) {
buffer1.data[id.x] = buffer1.data[id.x] + 1u;
buffer2.data[id.x] = buffer2.data[id.x] + 2u;
}
`,
});
const kNumIterations = 250_000;
const pipeline = t.device.createComputePipeline({
layout: 'auto',
compute: { module, entryPoint: 'main' },
});
const encoder = t.device.createCommandEncoder();
const pass = encoder.beginComputePass();
pass.setPipeline(pipeline);
for (let i = 0; i < kNumIterations; ++i) {
const buffer1Binding = i % 2;
const buffer2Binding = buffer1Binding ^ 1;
const bindGroup = t.device.createBindGroup({
layout: pipeline.getBindGroupLayout(0),
entries: [
{ binding: buffer1Binding, resource: { buffer: buffer1 } },
{ binding: buffer2Binding, resource: { buffer: buffer2 } },
],
});
pass.setBindGroup(0, bindGroup);
pass.dispatchWorkgroups(kNumElements);
}
pass.end();
t.device.queue.submit([encoder.finish()]);
const kTotalAddition = (kNumIterations / 2) * 3;
t.expectGPUBufferValuesEqual(
buffer1,
new Uint32Array([...iterRange(kNumElements, x => x + kTotalAddition)])
);
t.expectGPUBufferValuesEqual(
buffer2,
new Uint32Array([...iterRange(kNumElements, x => x + kTotalAddition)])
);
});
g.test('many_dispatches')
.desc(`Tests execution of compute passes with a huge number of dispatch calls`)
.fn(async t => {
const kNumElements = 64;
const data = new Uint32Array([...iterRange(kNumElements, x => x)]);
const buffer = t.makeBufferWithContents(data, GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC);
const module = t.device.createShaderModule({
code: `
struct Buffer { data: array<u32>, };
@group(0) @binding(0) var<storage, read_write> buffer: Buffer;
@compute @workgroup_size(1) fn main(
@builtin(global_invocation_id) id: vec3<u32>) {
buffer.data[id.x] = buffer.data[id.x] + 1u;
}
`,
});
const kNumIterations = 1_000_000;
const pipeline = t.device.createComputePipeline({
layout: 'auto',
compute: { module, entryPoint: 'main' },
});
const encoder = t.device.createCommandEncoder();
const pass = encoder.beginComputePass();
pass.setPipeline(pipeline);
const bindGroup = t.device.createBindGroup({
layout: pipeline.getBindGroupLayout(0),
entries: [{ binding: 0, resource: { buffer } }],
});
pass.setBindGroup(0, bindGroup);
for (let i = 0; i < kNumIterations; ++i) {
pass.dispatchWorkgroups(kNumElements);
}
pass.end();
t.device.queue.submit([encoder.finish()]);
t.expectGPUBufferValuesEqual(
buffer,
new Uint32Array([...iterRange(kNumElements, x => x + kNumIterations)])
);
});
g.test('huge_dispatches')
.desc(`Tests execution of compute passes with huge dispatch calls`)
.fn(async t => {
const kDimensions = [512, 512, 128];
kDimensions.forEach(x => {
assert(x <= t.device.limits.maxComputeWorkgroupsPerDimension);
});
const kNumElements = kDimensions[0] * kDimensions[1] * kDimensions[2];
const data = new Uint32Array([...iterRange(kNumElements, x => x)]);
const buffer = t.makeBufferWithContents(data, GPUBufferUsage.STORAGE | GPUBufferUsage.COPY_SRC);
const module = t.device.createShaderModule({
code: `
struct Buffer { data: array<u32>, };
@group(0) @binding(0) var<storage, read_write> buffer: Buffer;
@compute @workgroup_size(1) fn main(
@builtin(global_invocation_id) id: vec3<u32>) {
let index = (id.z * 512u + id.y) * 512u + id.x;
buffer.data[index] = buffer.data[index] + 1u;
}
`,
});
const kNumIterations = 16;
const pipeline = t.device.createComputePipeline({
layout: 'auto',
compute: { module, entryPoint: 'main' },
});
const bindGroup = t.device.createBindGroup({
layout: pipeline.getBindGroupLayout(0),
entries: [{ binding: 0, resource: { buffer } }],
});
for (let i = 0; i < kNumIterations; ++i) {
const encoder = t.device.createCommandEncoder();
const pass = encoder.beginComputePass();
pass.setBindGroup(0, bindGroup);
pass.setPipeline(pipeline);
pass.dispatchWorkgroups(kDimensions[0], kDimensions[1], kDimensions[2]);
pass.end();
t.device.queue.submit([encoder.finish()]);
await t.device.queue.onSubmittedWorkDone();
}
t.expectGPUBufferValuesEqual(
buffer,
new Uint32Array([...iterRange(kNumElements, x => x + kNumIterations)])
);
});
|
