gem_exec_parallel.c 6.46 KB
Newer Older
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
/*
 * Copyright © 2009 Intel Corporation
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice (including the next
 * paragraph) shall be included in all copies or substantial portions of the
 * Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
 * IN THE SOFTWARE.
 *
 */

/** @file gem_exec_parallel.c
 *
 * Exercise using many, many writers into a buffer.
 */

#include <pthread.h>

#include "igt.h"
#include "igt_gt.h"

#define LOCAL_I915_EXEC_NO_RELOC (1<<11)
#define LOCAL_I915_EXEC_HANDLE_LUT (1<<12)

#define LOCAL_I915_EXEC_BSD_SHIFT      (13)
#define LOCAL_I915_EXEC_BSD_MASK       (3 << LOCAL_I915_EXEC_BSD_SHIFT)

#define ENGINE_MASK  (I915_EXEC_RING_MASK | LOCAL_I915_EXEC_BSD_MASK)

#define VERIFY 0

static void check_bo(int fd, uint32_t handle, int pass)
{
	uint32_t *map;
	int i;

	igt_debug("Verifying result (pass=%d, handle=%d)\n", pass, handle);
	map = gem_mmap__cpu(fd, handle, 0, 4096, PROT_READ);
	gem_set_domain(fd, handle, I915_GEM_DOMAIN_CPU, 0);
	for (i = 0; i < 1024; i++)
		igt_assert_eq(map[i], i);
	munmap(map, 4096);
}

#define CONTEXTS 0x1
#define FDS 0x2

struct thread {
	pthread_t thread;
	pthread_mutex_t *mutex;
	pthread_cond_t *cond;
	unsigned flags;
	uint32_t *scratch;
	unsigned id;
	unsigned engine;
	int fd, gen, *go;
};

static void *thread(void *data)
{
	struct thread *t = data;
	struct drm_i915_gem_exec_object2 obj[2];
	struct drm_i915_gem_relocation_entry reloc;
	struct drm_i915_gem_execbuffer2 execbuf;
	uint32_t batch[16];
	int fd, i;

	pthread_mutex_lock(t->mutex);
	while (*t->go == 0)
		pthread_cond_wait(t->cond, t->mutex);
	pthread_mutex_unlock(t->mutex);

	if (t->flags & FDS)
		fd = drm_open_driver(DRIVER_INTEL);
	else
		fd = t->fd;

	i = 0;
	batch[i] = MI_STORE_DWORD_IMM | (t->gen < 6 ? 1 << 22 : 0);
	if (t->gen >= 8) {
		batch[++i] = 4*t->id;
		batch[++i] = 0;
	} else if (t->gen >= 4) {
		batch[++i] = 0;
		batch[++i] = 4*t->id;
	} else {
		batch[i]--;
		batch[++i] = 4*t->id;
	}
	batch[++i] = t->id;
	batch[++i] = MI_BATCH_BUFFER_END;

	memset(obj, 0, sizeof(obj));
	obj[0].flags = EXEC_OBJECT_WRITE;

	memset(&reloc, 0, sizeof(reloc));
	reloc.offset = sizeof(uint32_t);
	if (t->gen < 8 && t->gen >= 4)
		reloc.offset += sizeof(uint32_t);
	reloc.read_domains = I915_GEM_DOMAIN_INSTRUCTION;
	reloc.write_domain = I915_GEM_DOMAIN_INSTRUCTION;
	reloc.delta = 4*t->id;
	obj[1].handle = gem_create(fd, 4096);
117
	obj[1].relocs_ptr = to_user_pointer(&reloc);
118 119 120 121
	obj[1].relocation_count = 1;
	gem_write(fd, obj[1].handle, 0, batch, sizeof(batch));

	memset(&execbuf, 0, sizeof(execbuf));
122
	execbuf.buffers_ptr = to_user_pointer(obj);
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
	execbuf.buffer_count = 2;
	execbuf.flags = t->engine;
	execbuf.flags |= LOCAL_I915_EXEC_HANDLE_LUT;
	execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC;
	if (t->gen < 6)
		execbuf.flags |= I915_EXEC_SECURE;
	if (t->flags & CONTEXTS)
		execbuf.rsvd1 = gem_context_create(fd);

	for (i = 0; i < 16; i++) {
		obj[0].handle = t->scratch[i];
		if (t->flags & FDS)
			obj[0].handle = gem_open(fd, obj[0].handle);

		gem_execbuf(fd, &execbuf);

		if (t->flags & FDS)
			gem_close(fd, obj[0].handle);
	}

	if (t->flags & CONTEXTS)
		gem_context_destroy(fd, execbuf.rsvd1);
	gem_close(fd, obj[1].handle);
	if (t->flags & FDS)
		close(fd);

	return NULL;
}

static void all(int fd, unsigned engine, unsigned flags)
{
	const int gen = intel_gen(intel_get_drm_devid(fd));
	pthread_mutex_t mutex;
	pthread_cond_t cond;
	struct thread *threads;
	unsigned engines[16];
	unsigned nengine;
	uint32_t scratch[16], handle[16];
	int go;
	int i;

	if (flags & CONTEXTS)
165
		gem_require_contexts(fd);
166

167 168 169
	if (flags & FDS)
		igt_require(gen > 5);

170
	nengine = 0;
171
	if (engine == ALL_ENGINES) {
172 173
		for_each_physical_engine(fd, engine) {
			if (gem_can_store_dword(fd, engine))
174 175 176 177
				engines[nengine++] = engine;
		}
	} else {
		igt_require(gem_has_ring(fd, engine));
178
		igt_require(gem_can_store_dword(fd, engine));
179 180 181 182 183 184 185 186 187 188 189 190 191
		engines[nengine++] = engine;
	}
	igt_require(nengine);

	for (i = 0; i < 16; i++) {
		scratch[i] = handle[i] = gem_create(fd, 4096);
		if (flags & FDS)
			scratch[i] = gem_flink(fd, handle[i]);
	}

	threads = calloc(1024, sizeof(struct thread));
	igt_assert(threads);

192
	intel_detect_and_clear_missed_interrupts(fd);
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
	pthread_mutex_init(&mutex, 0);
	pthread_cond_init(&cond, 0);
	go = 0;

	for (i = 0; i < 1024; i++) {
		threads[i].id = i;
		threads[i].fd = fd;
		threads[i].gen = gen;
		threads[i].engine = engines[i % nengine];
		threads[i].flags = flags;
		threads[i].scratch = scratch;
		threads[i].mutex = &mutex;
		threads[i].cond = &cond;
		threads[i].go = &go;

		pthread_create(&threads[i].thread, 0, thread, &threads[i]);
	}

	pthread_mutex_lock(&mutex);
	go = 1024;
	pthread_cond_broadcast(&cond);
	pthread_mutex_unlock(&mutex);

	for (i = 0; i < 1024; i++)
		pthread_join(threads[i].thread, NULL);

	for (i = 0; i < 16; i++) {
		check_bo(fd, handle[i], i);
		gem_close(fd, handle[i]);
	}

224
	igt_assert_eq(intel_detect_and_clear_missed_interrupts(fd), 0);
225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
	free(threads);
}

igt_main
{
	const struct mode {
		const char *name;
		unsigned flags;
	} modes[] = {
		{ "", 0 },
		{ "contexts", CONTEXTS },
		{ "fds", FDS },
		{ NULL }
	};
	int fd;

241
	igt_fixture {
242
		fd = drm_open_driver_master(DRIVER_INTEL);
243
		igt_require_gem(fd);
244

245 246
		igt_fork_hang_detector(fd);
	}
247 248 249

	for (const struct mode *m = modes; m->name; m++)
		igt_subtest_f("%s", *m->name ? m->name : "basic")
250
			all(fd, ALL_ENGINES, m->flags);
251 252 253 254 255 256 257 258 259 260 261

	for (const struct intel_execution_engine *e = intel_execution_engines;
	     e->name; e++) {
		for (const struct mode *m = modes; m->name; m++)
			igt_subtest_f("%s%s%s",
				      e->name,
				      *m->name ? "-" : "",
				      m->name)
				all(fd, e->exec_id | e->flags, m->flags);
	}

262 263
	igt_fixture {
		igt_stop_hang_detector();
264
		close(fd);
265
	}
266
}