benchmarks: Add a microbenchmark for relocation overhead

Allow specification of the many different busyness modes and relocation
interfaces, along with the number of buffers to use and relocations.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

1 files changed, 257 insertions, 0 deletions

diff --git a/benchmarks/gem_exec_reloc.c b/benchmarks/gem_exec_reloc.c
new file mode 100644
index 00000000..f9e9c84c
--- /dev/null
+++ b/benchmarks/gem_exec_reloc.c
@@ -0,0 +1,257 @@
+/*
+ * Copyright © 2012 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.
+ *
+ * Authors:
+ *    Chris Wilson <chris@chris-wilson.co.uk>
+ *
+ */
+
+/* Exercises the basic execbuffer using the handle LUT interface */
+
+#include <stdlib.h>
+#include <stdio.h>
+#include <string.h>
+#include <fcntl.h>
+#include <inttypes.h>
+#include <errno.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include "drm.h"
+#include "ioctl_wrappers.h"
+#include "igt_debugfs.h"
+#include "drmtest.h"
+
+#define LOCAL_I915_EXEC_NO_RELOC (1<<11)
+#define LOCAL_I915_EXEC_HANDLE_LUT (1<<12)
+
+#define SKIP_RELOC 0x1
+#define NO_RELOC 0x2
+#define CYCLE_BATCH 0x4
+#define FAULT 0x8
+#define LUT 0x10
+#define SEQUENTIAL_OFFSET 0x20
+#define REVERSE_OFFSET 0x40
+#define RANDOM_OFFSET 0x80
+
+static uint32_t
+hars_petruska_f54_1_random (void)
+{
+	static uint32_t state = 0x12345678;
+
+#define rol(x,k) ((x << k) | (x >> (32-k)))
+	return state = (state ^ rol (state, 5) ^ rol (state, 24)) + 0x37798849;
+#undef rol
+}
+
+#define ELAPSED(a,b) (1e6*((b)->tv_sec - (a)->tv_sec) + ((b)->tv_usec - (a)->tv_usec))
+static int run(unsigned batch_size,
+	       unsigned flags,
+	       int num_objects,
+	       int num_relocs, int reps)
+{
+	uint32_t batch[2] = {MI_BATCH_BUFFER_END};
+	uint32_t cycle[16];
+	int fd, n, count, c, size;
+	struct drm_i915_gem_relocation_entry *reloc;
+	struct drm_i915_gem_execbuffer2 execbuf;
+	struct drm_i915_gem_exec_object2 *objects;
+	struct timeval start, end;
+	uint32_t reloc_handle;
+	struct drm_i915_gem_exec_object2 *gem_exec;
+	struct drm_i915_gem_relocation_entry *mem_reloc;
+	int *target;
+
+	gem_exec = calloc(sizeof(*gem_exec), num_objects + 1);
+	mem_reloc = calloc(sizeof(*mem_reloc), num_relocs);
+	target = calloc(sizeof(*target), num_relocs);
+
+	fd = drm_open_any();
+
+	for (n = 0; n < num_objects; n++)
+		gem_exec[n].handle = gem_create(fd, 4096);
+
+	for (n = 0; n < 16; n++) {
+		cycle[n] = gem_create(fd, batch_size);
+		gem_write(fd, cycle[n], 0, batch, sizeof(batch));
+	}
+	gem_exec[num_objects].handle = cycle[c = 0];
+
+	for (n = 0; n < num_relocs; n++) {
+		mem_reloc[n].offset = 1024;
+		mem_reloc[n].read_domains = I915_GEM_DOMAIN_RENDER;
+	}
+	for (n = 0; n < num_relocs; n++) {
+		if (flags & SEQUENTIAL_OFFSET)
+			mem_reloc[n].offset = 8 + (8*n % (batch_size - 16));
+		else if (flags & REVERSE_OFFSET)
+			mem_reloc[n].offset = batch_size - 8 - (8*n % (batch_size - 16));
+		else if (flags & RANDOM_OFFSET)
+			mem_reloc[n].offset = 8 +
+				8*hars_petruska_f54_1_random() % (batch_size - 16);
+		else
+			mem_reloc[n].offset = 1024;
+		mem_reloc[n].read_domains = I915_GEM_DOMAIN_RENDER;
+	}
+
+	size = ALIGN(sizeof(*mem_reloc)*num_relocs, 4096);
+	reloc_handle = gem_create(fd, size);
+	reloc = gem_mmap__cpu(fd, reloc_handle, 0, size, PROT_READ | PROT_WRITE);
+	memcpy(reloc, mem_reloc, sizeof(*mem_reloc)*num_relocs);
+	munmap(reloc, size);
+
+	if (flags & FAULT) {
+		igt_disable_prefault();
+		reloc = gem_mmap__cpu(fd, reloc_handle, 0, size, PROT_READ | PROT_WRITE);
+	} else
+		reloc = mem_reloc;
+
+	gem_exec[num_objects].relocation_count = num_relocs;
+	gem_exec[num_objects].relocs_ptr = (uintptr_t)reloc;
+	objects = gem_exec;
+
+	memset(&execbuf, 0, sizeof(execbuf));
+	execbuf.buffers_ptr = (uintptr_t)objects;
+	execbuf.buffer_count = num_objects + 1;
+	if (flags & LUT)
+		execbuf.flags |= LOCAL_I915_EXEC_HANDLE_LUT;
+	if (flags & NO_RELOC)
+		execbuf.flags |= LOCAL_I915_EXEC_NO_RELOC;
+
+	for (n = 0; n < num_relocs; n++) {
+		target[n] = hars_petruska_f54_1_random() % num_objects;
+		if (flags & LUT)
+			reloc[n].target_handle = target[n];
+		else
+			reloc[n].target_handle = objects[target[n]].handle;
+		reloc[n].presumed_offset = 0;
+	}
+
+	gem_execbuf(fd, &execbuf);
+
+	while (reps--) {
+		gettimeofday(&start, NULL);
+		for (count = 0; count < 1000; count++) {
+			if ((flags & SKIP_RELOC) == 0) {
+				for (n = 0; n < num_relocs; n++)
+					reloc[n].presumed_offset = 0;
+				if (flags & CYCLE_BATCH) {
+					c = (c + 1) % 16;
+					gem_exec[num_objects].handle = cycle[c];
+				}
+			}
+			if (flags & FAULT) {
+				munmap(reloc, size);
+				reloc = gem_mmap__cpu(fd, reloc_handle, 0, size, PROT_READ | PROT_WRITE);
+				gem_exec[num_objects].relocs_ptr = (uintptr_t)reloc;
+			}
+			gem_execbuf(fd, &execbuf);
+		}
+		gettimeofday(&end, NULL);
+		printf("%.3f\n", ELAPSED(&start, &end));
+	}
+
+	if (flags & FAULT) {
+		munmap(reloc, size);
+		igt_enable_prefault();
+	}
+
+	return 0;
+}
+
+int main(int argc, char **argv)
+{
+	unsigned num_objects = 1, num_relocs = 0, flags = 0;
+	unsigned size = 4096;
+	int reps = 13;
+	int c;
+
+	while ((c = getopt (argc, argv, "b:r:s:e:l:m:o:")) != -1) {
+		switch (c) {
+		case 'l':
+			reps = atoi(optarg);
+			if (reps < 1)
+				reps = 1;
+			break;
+
+		case 's':
+			size = atoi(optarg);
+			if (size < 4096)
+				size = 4096;
+			size = ALIGN(size, 4096);
+			break;
+
+		case 'e':
+			if (strcmp(optarg, "busy") == 0) {
+				flags |= 0;
+			} else if (strcmp(optarg, "cyclic") == 0) {
+				flags |= CYCLE_BATCH;
+			} else if (strcmp(optarg, "fault") == 0) {
+				flags |= FAULT;
+			} else if (strcmp(optarg, "skip") == 0) {
+				flags |= SKIP_RELOC;
+			} else if (strcmp(optarg, "none") == 0) {
+				flags |= SKIP_RELOC | NO_RELOC;
+			} else {
+				abort();
+			}
+			break;
+
+		case 'm':
+			if (strcmp(optarg, "old") == 0) {
+				flags |= 0;
+			} else if (strcmp(optarg, "lut") == 0) {
+				flags |= LUT;
+			} else {
+				abort();
+			}
+			break;
+
+		case 'o':
+			if (strcmp(optarg, "constant") == 0) {
+				flags |= 0;
+			} else if (strcmp(optarg, "sequential") == 0) {
+				flags |= SEQUENTIAL_OFFSET;
+			} else if (strcmp(optarg, "reverse") == 0) {
+				flags |= REVERSE_OFFSET;
+			} else if (strcmp(optarg, "random") == 0) {
+				flags |= RANDOM_OFFSET;
+			} else {
+				abort();
+			}
+			break;
+
+		case 'b':
+			num_objects = atoi(optarg);
+			if (num_objects < 1)
+				num_objects = 1;
+			break;
+
+		case 'r':
+			num_relocs = atoi(optarg);
+			if (num_relocs < 0)
+				num_relocs = 0;
+			break;
+		}
+	}
+
+	return run(size, flags, num_objects, num_relocs, reps);
+}