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
|
// SPDX-License-Identifier: MIT
/*
* Copyright © 2022 Intel Corporation
*/
/**
* TEST: Check compute-related preemption functionality
* Category: Core
* Mega feature: WMTP
* Sub-category: wmtp tests
* Functionality: OpenCL kernel
* Test category: functionality test
*/
#include <inttypes.h>
#include <string.h>
#include "igt.h"
#include "intel_compute.h"
#include "xe/xe_query.h"
/**
* SUBTEST: compute-preempt
* GPU requirement: LNL, PTL
* Description:
* Exercise compute walker mid thread preemption scenario
*
* SUBTEST: compute-preempt-many
* GPU requirement: LNL, PTL
* Description:
* Exercise multiple walker mid thread preemption scenario
*
* SUBTEST: compute-preempt-many-all-ram
* GPU requirement: LNL, PTL
* Description:
* Exercise multiple walker mid thread preemption scenario consuming
* whole ram only when there's swap on the machine
*
* SUBTEST: compute-preempt-many-vram
* GPU requirement: LNL, PTL
* Description:
* Exercise multiple walker mid thread preemption scenario on half of vram
*
* SUBTEST: compute-preempt-many-vram-evict
* GPU requirement: LNL, PTL
* Description:
* Exercise multiple walker mid thread preemption scenario on 120% of vram size
*
* SUBTEST: compute-threadgroup-preempt
* GPU requirement: LNL, PTL
* Description:
* Exercise compute walker threadgroup preemption scenario
*/
static void
test_compute_preempt(int fd, struct drm_xe_engine_class_instance *hwe, bool threadgroup_preemption,
enum execenv_alloc_prefs alloc_prefs)
{
igt_require_f(run_intel_compute_kernel_preempt(fd, hwe, threadgroup_preemption,
alloc_prefs),
"GPU not supported\n");
}
#define CONTEXT_MB 100
igt_main
{
int xe;
struct drm_xe_engine_class_instance *hwe;
uint64_t ram_mb, swap_mb, vram_mb;
igt_fixture {
xe = drm_open_driver(DRIVER_XE);
ram_mb = igt_get_avail_ram_mb();
swap_mb = igt_get_total_swap_mb();
vram_mb = xe_visible_vram_size(xe, 0) >> 20;
}
igt_subtest_with_dynamic("compute-preempt") {
xe_for_each_engine(xe, hwe) {
if (hwe->engine_class != DRM_XE_ENGINE_CLASS_COMPUTE)
continue;
igt_dynamic_f("engine-%s", xe_engine_class_string(hwe->engine_class))
test_compute_preempt(xe, hwe, false, EXECENV_PREF_SYSTEM);
}
}
igt_subtest_with_dynamic("compute-preempt-many") {
xe_for_each_engine(xe, hwe) {
if (hwe->engine_class != DRM_XE_ENGINE_CLASS_COMPUTE)
continue;
igt_dynamic_f("engine-%s", xe_engine_class_string(hwe->engine_class)) {
int child_count;
/*
* Get half of ram, then divide by
* CONTEXT_MB * 2 (long and short) job
*/
child_count = ram_mb / 2 / CONTEXT_MB / 2;
igt_debug("RAM: %" PRIu64 ", child count: %d\n",
ram_mb, child_count);
igt_fork(child, child_count)
test_compute_preempt(xe, hwe, false, EXECENV_PREF_SYSTEM);
igt_waitchildren();
}
}
}
igt_subtest_with_dynamic("compute-preempt-many-all-ram") {
igt_require(swap_mb > CONTEXT_MB * 10);
xe_for_each_engine(xe, hwe) {
if (hwe->engine_class != DRM_XE_ENGINE_CLASS_COMPUTE)
continue;
igt_dynamic_f("engine-%s", xe_engine_class_string(hwe->engine_class)) {
int child_count;
/*
* Get whole ram, then divide by
* CONTEXT_MB * 2 (long and short) job
*/
child_count = ram_mb / CONTEXT_MB / 2;
igt_debug("RAM: %" PRIu64 ", child count: %d\n",
ram_mb, child_count);
igt_fork(child, child_count)
test_compute_preempt(xe, hwe, false, EXECENV_PREF_SYSTEM);
igt_waitchildren();
}
}
}
igt_subtest_with_dynamic("compute-preempt-many-vram") {
igt_require(xe_has_vram(xe));
xe_for_each_engine(xe, hwe) {
if (hwe->engine_class != DRM_XE_ENGINE_CLASS_COMPUTE)
continue;
igt_dynamic_f("engine-%s", xe_engine_class_string(hwe->engine_class)) {
int child_count;
/*
* Get half of vram, then divide by
* CONTEXT_MB * 2 (long and short) job
*/
child_count = vram_mb / 2 / CONTEXT_MB / 2;
igt_debug("VRAM: %" PRIu64 ", child count: %d\n",
vram_mb, child_count);
igt_fork(child, child_count)
test_compute_preempt(xe, hwe, false, EXECENV_PREF_VRAM);
igt_waitchildren();
}
}
}
igt_subtest_with_dynamic("compute-preempt-many-vram-evict") {
igt_require(xe_has_vram(xe));
xe_for_each_engine(xe, hwe) {
if (hwe->engine_class != DRM_XE_ENGINE_CLASS_COMPUTE)
continue;
igt_dynamic_f("engine-%s", xe_engine_class_string(hwe->engine_class)) {
int child_count;
/*
* Get all vram + 20%, then divide by
* CONTEXT_MB * 2 (long and short) job
*/
child_count = vram_mb * 1.2 / 2 / CONTEXT_MB;
igt_debug("VRAM: %" PRIu64 ", child count: %d\n",
vram_mb, child_count);
igt_fork(child, child_count)
test_compute_preempt(xe, hwe, false, EXECENV_PREF_VRAM);
igt_waitchildren();
}
}
}
igt_subtest_with_dynamic("compute-threadgroup-preempt") {
xe_for_each_engine(xe, hwe) {
if (hwe->engine_class != DRM_XE_ENGINE_CLASS_COMPUTE)
continue;
igt_dynamic_f("engine-%s", xe_engine_class_string(hwe->engine_class))
test_compute_preempt(xe, hwe, true, EXECENV_PREF_SYSTEM);
}
}
igt_fixture
drm_close_driver(xe);
}
|
