[PATCH 2/4] benchmark: Measure allocation time for objects

[Dandamudi, Priyanka]

> -----Original Message-----
> From: igt-dev <igt-dev-bounces at lists.freedesktop.org> On Behalf Of Pravalika
> Gurram
> Sent: 02 April 2025 11:41 AM
> To: igt-dev at lists.freedesktop.org
> Cc: Gurram, Pravalika <pravalika.gurram at intel.com>
> Subject: [PATCH 2/4] benchmark: Measure allocation time for objects
> 
> A basic measurement, how fast can we create and populate an object with
> backing storage
> 
> Signed-off-by: Pravalika Gurram <pravalika.gurram at intel.com>
> ---
>  benchmarks/meson.build |   1 +
>  benchmarks/xe_create.c | 233
> +++++++++++++++++++++++++++++++++++++++++
>  2 files changed, 234 insertions(+)
>  create mode 100644 benchmarks/xe_create.c
> 
> diff --git a/benchmarks/meson.build b/benchmarks/meson.build index
> 4421ede86..00203c62e 100644
> --- a/benchmarks/meson.build
> +++ b/benchmarks/meson.build
> @@ -22,6 +22,7 @@ benchmark_progs = [
>  	'prime_lookup',
>  	'vgem_mmap',
>          'xe_blt',
> +        'xe_create',
>  ]
> 
>  benchmarksdir = join_paths(libexecdir, 'benchmarks') diff --git
> a/benchmarks/xe_create.c b/benchmarks/xe_create.c new file mode 100644
> index 000000000..cd97d6920
> --- /dev/null
> +++ b/benchmarks/xe_create.c
> @@ -0,0 +1,233 @@
> +/*
> + * Copyright © 2025 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:
> + *    Pravalika Gurram <pravalika.gurram at intel.com>
> + *
> + */
> +
> +#include <unistd.h>
> +#include <stdlib.h>
> +#include <stdint.h>
> +#include <stdio.h>
> +#include <string.h>
> +#include <fcntl.h>
> +#include <inttypes.h>
> +#include <errno.h>
> +#include <sys/stat.h>
> +#include <sys/ioctl.h>
> +#include <sys/time.h>
> +#include <time.h>
> +
> +#include "drm.h"
> +#include "drmtest.h"
> +#include "i915/gem_create.h"
> +#include "igt_aux.h"
> +#include "igt_stats.h"
> +#include "intel_reg.h"
> +#include "ioctl_wrappers.h"
> +
> +#include "igt.h"
> +#include "igt_core.h"
> +#include "igt_syncobj.h"
> +#include "intel_reg.h"
> +#include "xe/xe_ioctl.h"
> +#include "xe/xe_query.h"
> +
> +
> +#define OBJECT_SIZE (1<<23)
> +
> +struct data {
> +	uint32_t batch[16];
> +	uint64_t pad;
> +	uint32_t data;
> +	uint64_t addr;
> +};
> +
> +static double elapsed(const struct timespec *start,
> +				const struct timespec *end)
> +{
> +	return (end->tv_sec - start->tv_sec) + 1e-9*(end->tv_nsec -
> +start->tv_nsec); }
> +
> +static void store_dword_batch(struct data *data, uint64_t addr, int
> +value) {
> +	int b;
> +	uint64_t batch_offset = (char *)&(data->batch) - (char *)data;
> +	uint64_t batch_addr = addr + batch_offset;
> +	uint64_t sdi_offset = (char *)&(data->data) - (char *)data;
> +	uint64_t sdi_addr = addr + sdi_offset;
> +
> +	b = 0;
> +
> +	data->batch[b++] = MI_BATCH_BUFFER_END;
> +	igt_assert(b <= ARRAY_SIZE(data->batch));
> +
> +	data->addr = batch_addr;
> +}
> +static void test_exec(int fd, int busy) {
> +	uint32_t vm;
> +	size_t bo_size;
> +	uint32_t bo = 0;
> +	struct data *data;
> +	uint32_t exec_queue;
> +	uint32_t syncobj;
> +	int value = 0x123456;
> +	uint64_t addr = 0x100000;
> +
> +	struct drm_xe_sync sync = {
> +		.flags = DRM_XE_SYNC_TYPE_SYNCOBJ |
> DRM_XE_SYNC_FLAG_SIGNAL
> +	};
> +	struct drm_xe_exec exec = {
> +		.num_batch_buffer = 1,
> +		.num_syncs = 1,
> +		.syncs = to_user_pointer(&sync),
> +	};
> +
> +	struct drm_xe_engine_class_instance inst = {
> +		.engine_class = DRM_XE_ENGINE_CLASS_COPY,
> +	};
> +
> +	vm = xe_vm_create(fd, 0, 0);
> +	bo_size = sizeof(*data);
> +	bo_size = xe_bb_size(fd, bo_size);
> +
> +	bo = xe_bo_create(fd, vm, bo_size,
> +			vram_if_possible(fd, 0),
> +
> 	DRM_XE_GEM_CREATE_FLAG_NEEDS_VISIBLE_VRAM);
> +	if (busy) {
Why cant it be all below made into a function and just called. Also, can you make all the changes which I already mentioned to the pr you sent.
> +		syncobj = syncobj_create(fd, 0);
> +		sync.handle = syncobj;
> +
> +		xe_vm_bind_async(fd, vm, 0, bo, 0, addr, bo_size, &sync, 1);
> +		data = xe_bo_map(fd, bo, bo_size);
> +		store_dword_batch(data, addr, value);
> +
> +		igt_assert(syncobj_wait(fd, &syncobj, 1, INT64_MAX, 0,
> NULL));
> +		syncobj_reset(fd, &syncobj, 1);
> +
> +		exec_queue = xe_exec_queue_create(fd, vm, &inst, 0);
> +		exec.exec_queue_id = exec_queue;
> +		exec.address = data->addr;
> +		sync.flags &= DRM_XE_SYNC_FLAG_SIGNAL;
> +		xe_exec(fd, &exec);
> +
> +		igt_assert(syncobj_wait(fd, &syncobj, 1, INT64_MAX, 0,
> NULL));
> +
> +		syncobj_destroy(fd, syncobj);
> +
> +		xe_exec_queue_destroy(fd, exec_queue);
> +	}
> +	munmap(data, bo_size);
> +	gem_close(fd, bo);
> +
> +	xe_vm_destroy(fd, vm);
> +}
> +int main(int argc, char **argv)
> +{
> +	int fd = drm_open_driver(DRIVER_XE);
> +	int size = 0;
> +	int busy = 0;
> +	int reps = 13;
> +	int ncpus = 1;
> +	int c, n, s;
> +
> +	while ((c = getopt (argc, argv, "bs:r:f")) != -1) {
> +		switch (c) {
> +		case 's':
> +			size = atoi(optarg);
> +			break;
> +
> +		case 'r':
> +			reps = atoi(optarg);
> +			if (reps < 1)
> +				reps = 1;
> +			break;
> +
> +		case 'f':
> +			ncpus = sysconf(_SC_NPROCESSORS_ONLN);
> +			break;
> +
> +		case 'b':
> +			busy = true;
> +			break;
> +
> +		default:
> +			break;
> +		}
> +	}
> +
> +	if (size == 0) {
> +		for (s = 4096; s <=  OBJECT_SIZE; s <<= 1) {
> +			igt_stats_t stats;
> +
> +			igt_stats_init_with_size(&stats, reps);
> +			for (n = 0; n < reps; n++) {
> +				struct timespec start, end;
> +				uint64_t count = 0;
> +
> +				clock_gettime(CLOCK_MONOTONIC, &start);
> +				do {
> +					for (c = 0; c < 1000; c++)
> +						test_exec(fd, busy);
> +					count += c;
> +					clock_gettime(CLOCK_MONOTONIC,
> &end);
> +				} while (end.tv_sec - start.tv_sec < 2);
> +
> +				igt_stats_push_float(&stats, count /
> elapsed(&start, &end));
> +			}
> +			printf("%f\n", igt_stats_get_trimean(&stats));
> +			igt_stats_fini(&stats);
> +		}
> +	} else {
> +		double *shared;
> +
> +		shared = mmap(0, 4096, PROT_WRITE, MAP_SHARED |
> MAP_ANON, -1, 0);
> +		for (n = 0; n < reps; n++) {
> +			memset(shared, 0, 4096);
> +
> +			igt_fork(child, ncpus) {
> +				struct timespec start, end;
> +				uint64_t count = 0;
> +
> +				clock_gettime(CLOCK_MONOTONIC, &start);
> +				do {
> +					for (c = 0; c < 1000; c++)
> +						test_exec(fd, busy);
> +					count += c;
> +					clock_gettime(CLOCK_MONOTONIC,
> &end);
> +				} while (end.tv_sec - start.tv_sec < 2);
> +
> +				shared[child] = count / elapsed(&start, &end);
> +			}
> +			igt_waitchildren();
> +
> +			for (int child = 0; child < ncpus; child++)
> +				shared[ncpus] += shared[child];
> +
> +			printf("%7.3f\n", shared[ncpus]);
> +		}
> +	}
> +
> +	return 0;
> +}
> --
> 2.34.1