summaryrefslogtreecommitdiffstats
path: root/app/test-crypto-perf/cperf_test_pmd_cyclecount.c
blob: 9b41724af3eabc73a36c5ad8ed588ae002a77f8a (plain)
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
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
/*-
 *   BSD LICENSE
 *
 *   Copyright(c) 2017 Intel Corporation. All rights reserved.
 *
 *   Redistribution and use in source and binary forms, with or without
 *   modification, are permitted provided that the following conditions
 *   are met:
 *
 *     * Redistributions of source code must retain the above copyright
 *       notice, this list of conditions and the following disclaimer.
 *     * Redistributions in binary form must reproduce the above copyright
 *       notice, this list of conditions and the following disclaimer in
 *       the documentation and/or other materials provided with the
 *       distribution.
 *     * Neither the name of Intel Corporation nor the names of its
 *       contributors may be used to endorse or promote products derived
 *       from this software without specific prior written permission.
 *
 *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 */

#include <stdbool.h>

#include <rte_crypto.h>
#include <rte_cryptodev.h>
#include <rte_cycles.h>
#include <rte_malloc.h>

#include "cperf_ops.h"
#include "cperf_test_pmd_cyclecount.h"
#include "cperf_test_common.h"

#define PRETTY_HDR_FMT "%12s%12s%12s%12s%12s%12s%12s%12s%12s%12s\n\n"
#define PRETTY_LINE_FMT "%12u%12u%12u%12u%12u%12u%12u%12.0f%12.0f%12.0f\n"
#define CSV_HDR_FMT "%s,%s,%s,%s,%s,%s,%s,%s,%s,%s\n"
#define CSV_LINE_FMT "%10u;%10u;%u;%u;%u;%u;%u;%.f3;%.f3;%.f3\n"

struct cperf_pmd_cyclecount_ctx {
	uint8_t dev_id;
	uint16_t qp_id;
	uint8_t lcore_id;

	struct rte_mempool *pool;
	struct rte_crypto_op **ops;
	struct rte_crypto_op **ops_processed;

	struct rte_cryptodev_sym_session *sess;

	cperf_populate_ops_t populate_ops;

	uint32_t src_buf_offset;
	uint32_t dst_buf_offset;

	const struct cperf_options *options;
	const struct cperf_test_vector *test_vector;
};

struct pmd_cyclecount_state {
	struct cperf_pmd_cyclecount_ctx *ctx;
	const struct cperf_options *opts;
	uint32_t lcore;
	uint64_t delay;
	int linearize;
	uint32_t ops_enqd;
	uint32_t ops_deqd;
	uint32_t ops_enq_retries;
	uint32_t ops_deq_retries;
	double cycles_per_build;
	double cycles_per_enq;
	double cycles_per_deq;
};

static const uint16_t iv_offset =
		sizeof(struct rte_crypto_op) + sizeof(struct rte_crypto_sym_op);

static void
cperf_pmd_cyclecount_test_free(struct cperf_pmd_cyclecount_ctx *ctx)
{
	if (ctx) {
		if (ctx->sess) {
			rte_cryptodev_sym_session_clear(ctx->dev_id, ctx->sess);
			rte_cryptodev_sym_session_free(ctx->sess);
		}

		if (ctx->pool)
			rte_mempool_free(ctx->pool);

		if (ctx->ops)
			rte_free(ctx->ops);

		if (ctx->ops_processed)
			rte_free(ctx->ops_processed);

		rte_free(ctx);
	}
}

void *
cperf_pmd_cyclecount_test_constructor(struct rte_mempool *sess_mp,
		uint8_t dev_id, uint16_t qp_id,
		const struct cperf_options *options,
		const struct cperf_test_vector *test_vector,
		const struct cperf_op_fns *op_fns)
{
	struct cperf_pmd_cyclecount_ctx *ctx = NULL;

	/* preallocate buffers for crypto ops as they can get quite big */
	size_t alloc_sz = sizeof(struct rte_crypto_op *) *
			options->nb_descriptors;

	ctx = rte_malloc(NULL, sizeof(struct cperf_pmd_cyclecount_ctx), 0);
	if (ctx == NULL)
		goto err;

	ctx->dev_id = dev_id;
	ctx->qp_id = qp_id;

	ctx->populate_ops = op_fns->populate_ops;
	ctx->options = options;
	ctx->test_vector = test_vector;

	/* IV goes at the end of the crypto operation */
	uint16_t iv_offset = sizeof(struct rte_crypto_op) +
			sizeof(struct rte_crypto_sym_op);

	ctx->sess = op_fns->sess_create(
			sess_mp, dev_id, options, test_vector, iv_offset);
	if (ctx->sess == NULL)
		goto err;

	if (cperf_alloc_common_memory(options, test_vector, dev_id, qp_id, 0,
			&ctx->src_buf_offset, &ctx->dst_buf_offset,
			&ctx->pool) < 0)
		goto err;

	ctx->ops = rte_malloc("ops", alloc_sz, 0);
	if (!ctx->ops)
		goto err;

	ctx->ops_processed = rte_malloc("ops_processed", alloc_sz, 0);
	if (!ctx->ops_processed)
		goto err;

	return ctx;

err:
	cperf_pmd_cyclecount_test_free(ctx);

	return NULL;
}

/* benchmark alloc-build-free of ops */
static inline int
pmd_cyclecount_bench_ops(struct pmd_cyclecount_state *state, uint32_t cur_op,
		uint16_t test_burst_size)
{
	uint32_t iter_ops_left = state->opts->total_ops - cur_op;
	uint32_t iter_ops_needed =
			RTE_MIN(state->opts->nb_descriptors, iter_ops_left);
	uint32_t cur_iter_op;

	for (cur_iter_op = 0; cur_iter_op < iter_ops_needed;
			cur_iter_op += test_burst_size) {
		uint32_t burst_size = RTE_MIN(state->opts->total_ops - cur_op,
				test_burst_size);
		struct rte_crypto_op **ops = &state->ctx->ops[cur_iter_op];

		/* Allocate objects containing crypto operations and mbufs */
		if (rte_mempool_get_bulk(state->ctx->pool, (void **)ops,
					burst_size) != 0) {
			RTE_LOG(ERR, USER1,
					"Failed to allocate more crypto operations "
					"from the the crypto operation pool.\n"
					"Consider increasing the pool size "
					"with --pool-sz\n");
				return -1;
		}

		/* Setup crypto op, attach mbuf etc */
		(state->ctx->populate_ops)(ops,
				state->ctx->src_buf_offset,
				state->ctx->dst_buf_offset,
				burst_size,
				state->ctx->sess, state->opts,
				state->ctx->test_vector, iv_offset);

#ifdef CPERF_LINEARIZATION_ENABLE
		/* Check if source mbufs require coalescing */
		if (state->linearize) {
			uint8_t i;
			for (i = 0; i < burst_size; i++) {
				struct rte_mbuf *src = ops[i]->sym->m_src;
				rte_pktmbuf_linearize(src);
			}
		}
#endif /* CPERF_LINEARIZATION_ENABLE */
		rte_mempool_put_bulk(state->ctx->pool, (void **)ops,
				burst_size);
	}

	return 0;
}

/* allocate and build ops (no free) */
static int
pmd_cyclecount_build_ops(struct pmd_cyclecount_state *state,
		uint32_t iter_ops_needed, uint16_t test_burst_size)
{
	uint32_t cur_iter_op;

	for (cur_iter_op = 0; cur_iter_op < iter_ops_needed;
			cur_iter_op += test_burst_size) {
		uint32_t burst_size = RTE_MIN(
				iter_ops_needed - cur_iter_op, test_burst_size);
		struct rte_crypto_op **ops = &state->ctx->ops[cur_iter_op];

		/* Allocate objects containing crypto operations and mbufs */
		if (rte_mempool_get_bulk(state->ctx->pool, (void **)ops,
					burst_size) != 0) {
			RTE_LOG(ERR, USER1,
					"Failed to allocate more crypto operations "
					"from the the crypto operation pool.\n"
					"Consider increasing the pool size "
					"with --pool-sz\n");
				return -1;
		}

		/* Setup crypto op, attach mbuf etc */
		(state->ctx->populate_ops)(ops,
				state->ctx->src_buf_offset,
				state->ctx->dst_buf_offset,
				burst_size,
				state->ctx->sess, state->opts,
				state->ctx->test_vector, iv_offset);
	}
	return 0;
}

/* benchmark enqueue, returns number of ops enqueued */
static uint32_t
pmd_cyclecount_bench_enq(struct pmd_cyclecount_state *state,
		uint32_t iter_ops_needed, uint16_t test_burst_size)
{
	/* Enqueue full descriptor ring of ops on crypto device */
	uint32_t cur_iter_op = 0;
	while (cur_iter_op < iter_ops_needed) {
		uint32_t burst_size = RTE_MIN(iter_ops_needed - cur_iter_op,
				test_burst_size);
		struct rte_crypto_op **ops = &state->ctx->ops[cur_iter_op];
		uint32_t burst_enqd;

		burst_enqd = rte_cryptodev_enqueue_burst(state->ctx->dev_id,
				state->ctx->qp_id, ops, burst_size);

		/* if we couldn't enqueue anything, the queue is full */
		if (!burst_enqd) {
			/* don't try to dequeue anything we didn't enqueue */
			return cur_iter_op;
		}

		if (burst_enqd < burst_size)
			state->ops_enq_retries++;
		state->ops_enqd += burst_enqd;
		cur_iter_op += burst_enqd;
	}
	return iter_ops_needed;
}

/* benchmark dequeue */
static void
pmd_cyclecount_bench_deq(struct pmd_cyclecount_state *state,
		uint32_t iter_ops_needed, uint16_t test_burst_size)
{
	/* Dequeue full descriptor ring of ops on crypto device */
	uint32_t cur_iter_op = 0;
	while (cur_iter_op < iter_ops_needed) {
		uint32_t burst_size = RTE_MIN(iter_ops_needed - cur_iter_op,
				test_burst_size);
		struct rte_crypto_op **ops_processed =
				&state->ctx->ops[cur_iter_op];
		uint32_t burst_deqd;

		burst_deqd = rte_cryptodev_dequeue_burst(state->ctx->dev_id,
				state->ctx->qp_id, ops_processed, burst_size);

		if (burst_deqd < burst_size)
			state->ops_deq_retries++;
		state->ops_deqd += burst_deqd;
		cur_iter_op += burst_deqd;
	}
}

/* run benchmark per burst size */
static inline int
pmd_cyclecount_bench_burst_sz(
		struct pmd_cyclecount_state *state, uint16_t test_burst_size)
{
	uint64_t tsc_start;
	uint64_t tsc_end;
	uint64_t tsc_op;
	uint64_t tsc_enq;
	uint64_t tsc_deq;
	uint32_t cur_op;

	/* reset all counters */
	tsc_enq = 0;
	tsc_deq = 0;
	state->ops_enqd = 0;
	state->ops_enq_retries = 0;
	state->ops_deqd = 0;
	state->ops_deq_retries = 0;

	/*
	 * Benchmark crypto op alloc-build-free separately.
	 */
	tsc_start = rte_rdtsc_precise();

	for (cur_op = 0; cur_op < state->opts->total_ops;
			cur_op += state->opts->nb_descriptors) {
		if (unlikely(pmd_cyclecount_bench_ops(
				state, cur_op, test_burst_size)))
			return -1;
	}

	tsc_end = rte_rdtsc_precise();
	tsc_op = tsc_end - tsc_start;


	/*
	 * Hardware acceleration cyclecount benchmarking loop.
	 *
	 * We're benchmarking raw enq/deq performance by filling up the device
	 * queue, so we never get any failed enqs unless the driver won't accept
	 * the exact number of descriptors we requested, or the driver won't
	 * wrap around the end of the TX ring. However, since we're only
	 * dequeueing once we've filled up the queue, we have to benchmark it
	 * piecemeal and then average out the results.
	 */
	cur_op = 0;
	while (cur_op < state->opts->total_ops) {
		uint32_t iter_ops_left = state->opts->total_ops - cur_op;
		uint32_t iter_ops_needed = RTE_MIN(
				state->opts->nb_descriptors, iter_ops_left);
		uint32_t iter_ops_allocd = iter_ops_needed;

		/* allocate and build ops */
		if (unlikely(pmd_cyclecount_build_ops(state, iter_ops_needed,
				test_burst_size)))
			return -1;

		tsc_start = rte_rdtsc_precise();

		/* fill up TX ring */
		iter_ops_needed = pmd_cyclecount_bench_enq(state,
				iter_ops_needed, test_burst_size);

		tsc_end = rte_rdtsc_precise();

		tsc_enq += tsc_end - tsc_start;

		/* allow for HW to catch up */
		if (state->delay)
			rte_delay_us_block(state->delay);

		tsc_start = rte_rdtsc_precise();

		/* drain RX ring */
		pmd_cyclecount_bench_deq(state, iter_ops_needed,
				test_burst_size);

		tsc_end = rte_rdtsc_precise();

		tsc_deq += tsc_end - tsc_start;

		cur_op += iter_ops_needed;

		/*
		 * we may not have processed all ops that we allocated, so
		 * free everything we've allocated.
		 */
		rte_mempool_put_bulk(state->ctx->pool,
				(void **)state->ctx->ops, iter_ops_allocd);
	}

	state->cycles_per_build = (double)tsc_op / state->opts->total_ops;
	state->cycles_per_enq = (double)tsc_enq / state->ops_enqd;
	state->cycles_per_deq = (double)tsc_deq / state->ops_deqd;

	return 0;
}

int
cperf_pmd_cyclecount_test_runner(void *test_ctx)
{
	struct pmd_cyclecount_state state = {0};
	const struct cperf_options *opts;
	uint16_t test_burst_size;
	uint8_t burst_size_idx = 0;

	state.ctx = test_ctx;
	opts = state.ctx->options;
	state.opts = opts;
	state.lcore = rte_lcore_id();
	state.linearize = 0;

	static int only_once;
	static bool warmup = true;

	/*
	 * We need a small delay to allow for hardware to process all the crypto
	 * operations. We can't automatically figure out what the delay should
	 * be, so we leave it up to the user (by default it's 0).
	 */
	state.delay = 1000 * opts->pmdcc_delay;

#ifdef CPERF_LINEARIZATION_ENABLE
	struct rte_cryptodev_info dev_info;

	/* Check if source mbufs require coalescing */
	if (opts->segments_sz < ctx->options->max_buffer_size) {
		rte_cryptodev_info_get(state.ctx->dev_id, &dev_info);
		if ((dev_info.feature_flags &
				    RTE_CRYPTODEV_FF_MBUF_SCATTER_GATHER) ==
				0) {
			state.linearize = 1;
		}
	}
#endif /* CPERF_LINEARIZATION_ENABLE */

	state.ctx->lcore_id = state.lcore;

	/* Get first size from range or list */
	if (opts->inc_burst_size != 0)
		test_burst_size = opts->min_burst_size;
	else
		test_burst_size = opts->burst_size_list[0];

	while (test_burst_size <= opts->max_burst_size) {
		/* do a benchmark run */
		if (pmd_cyclecount_bench_burst_sz(&state, test_burst_size))
			return -1;

		/*
		 * First run is always a warm up run.
		 */
		if (warmup) {
			warmup = false;
			continue;
		}

		if (!opts->csv) {
			if (!only_once)
				printf(PRETTY_HDR_FMT, "lcore id", "Buf Size",
						"Burst Size", "Enqueued",
						"Dequeued", "Enq Retries",
						"Deq Retries", "Cycles/Op",
						"Cycles/Enq", "Cycles/Deq");
			only_once = 1;

			printf(PRETTY_LINE_FMT, state.ctx->lcore_id,
					opts->test_buffer_size, test_burst_size,
					state.ops_enqd, state.ops_deqd,
					state.ops_enq_retries,
					state.ops_deq_retries,
					state.cycles_per_build,
					state.cycles_per_enq,
					state.cycles_per_deq);
		} else {
			if (!only_once)
				printf(CSV_HDR_FMT, "# lcore id", "Buf Size",
						"Burst Size", "Enqueued",
						"Dequeued", "Enq Retries",
						"Deq Retries", "Cycles/Op",
						"Cycles/Enq", "Cycles/Deq");
			only_once = 1;

			printf(CSV_LINE_FMT, state.ctx->lcore_id,
					opts->test_buffer_size, test_burst_size,
					state.ops_enqd, state.ops_deqd,
					state.ops_enq_retries,
					state.ops_deq_retries,
					state.cycles_per_build,
					state.cycles_per_enq,
					state.cycles_per_deq);
		}

		/* Get next size from range or list */
		if (opts->inc_burst_size != 0)
			test_burst_size += opts->inc_burst_size;
		else {
			if (++burst_size_idx == opts->burst_size_count)
				break;
			test_burst_size = opts->burst_size_list[burst_size_idx];
		}
	}

	return 0;
}

void
cperf_pmd_cyclecount_test_destructor(void *arg)
{
	struct cperf_pmd_cyclecount_ctx *ctx = arg;

	if (ctx == NULL)
		return;

	cperf_pmd_cyclecount_test_free(ctx);
}