numam-dpdk/lib/librte_jobstats/rte_jobstats.c
Pawel Wodkowski 597b0f74e2 jobstats: new library
This library provide API to measure time spend in particular parts of
code and to calculate optimal polling time.

To calculate a those statistics application code need to be divided into
parts (called jobs) that do something. It is up to application to decide
what is considered a job.

Series of jobs must be surrounded with the rte_jobstats_context_start()
and rte_jobstats_context_finish() calls. After that, jobs might be
started.  Each job must be surrounded with rte_jobstats_start() and
rte_jobstats_finish() calls.

After job finishes its execution, period in which it should be called
again is adjusted. It might be used to minimize time wasted on
unnecessary polls/calls. Adjustment is based on data provided by job
itself (ex: number of packets it processed).

After all jobs in serie are executed fallowing statistics are updated
and might be used by application. Statistics can be reset. Some of
provided statistic data:
 - total/min/max execution - time spent in executing jobs.
 - total/min/max management - time spent outside execution area. This
value might be used to measure overhead of scheduling jobs. This time
also contains overhead of rte_jobstats library itself.
 - number of loops that executed at least one job
 - executed jobs
 - time when statistics were reset.

Each job provide total/min/max execution time and execution count
statistics.

Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Acked-by: Pablo de Lara <pablo.de.lara.guarch@intel.com>
2015-02-24 22:12:35 +01:00

274 lines
6.9 KiB
C

/*-
* BSD LICENSE
*
* Copyright(c) 2015 Intel Corporation. All rights reserved.
* 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 <string.h>
#include <stdlib.h>
#include <errno.h>
#include <rte_errno.h>
#include <rte_common.h>
#include <rte_eal.h>
#include <rte_log.h>
#include <rte_cycles.h>
#include <rte_branch_prediction.h>
#include "rte_jobstats.h"
#define ADD_TIME_MIN_MAX(obj, type, value) do { \
typeof(value) tmp = (value); \
(obj)->type ## _time += tmp; \
if (tmp < (obj)->min_ ## type ## _time) \
(obj)->min_ ## type ## _time = tmp; \
if (tmp > (obj)->max_ ## type ## _time) \
(obj)->max_ ## type ## _time = tmp; \
} while (0)
#define RESET_TIME_MIN_MAX(obj, type) do { \
(obj)->type ## _time = 0; \
(obj)->min_ ## type ## _time = UINT64_MAX; \
(obj)->max_ ## type ## _time = 0; \
} while (0)
static inline uint64_t
get_time(void)
{
rte_rmb();
return rte_get_timer_cycles();
}
/* Those are steps used to adjust job period.
* Experiments show that for forwarding apps the up step must be less than down
* step to achieve optimal performance.
*/
#define JOB_UPDATE_STEP_UP 1
#define JOB_UPDATE_STEP_DOWN 4
/*
* Default update function that implements simple period adjustment.
*/
static void
default_update_function(struct rte_jobstats *job, int64_t result)
{
int64_t err = job->target - result;
/* Job is happy. Nothing to do */
if (err == 0)
return;
if (err > 0) {
if (job->period + JOB_UPDATE_STEP_UP < job->max_period)
job->period += JOB_UPDATE_STEP_UP;
} else {
if (job->min_period + JOB_UPDATE_STEP_DOWN < job->period)
job->period -= JOB_UPDATE_STEP_DOWN;
}
}
int
rte_jobstats_context_init(struct rte_jobstats_context *ctx)
{
if (ctx == NULL)
return -EINVAL;
/* Init only needed parameters. Zero out everything else. */
memset(ctx, 0, sizeof(struct rte_jobstats_context));
rte_jobstats_context_reset(ctx);
return 0;
}
void
rte_jobstats_context_start(struct rte_jobstats_context *ctx)
{
uint64_t now;
ctx->loop_executed_jobs = 0;
now = get_time();
ADD_TIME_MIN_MAX(ctx, management, now - ctx->state_time);
ctx->state_time = now;
}
void
rte_jobstats_context_finish(struct rte_jobstats_context *ctx)
{
uint64_t now;
if (likely(ctx->loop_executed_jobs))
ctx->loop_cnt++;
now = get_time();
ADD_TIME_MIN_MAX(ctx, management, now - ctx->state_time);
ctx->state_time = now;
}
void
rte_jobstats_context_reset(struct rte_jobstats_context *ctx)
{
RESET_TIME_MIN_MAX(ctx, exec);
RESET_TIME_MIN_MAX(ctx, management);
ctx->start_time = get_time();
ctx->state_time = ctx->start_time;
ctx->job_exec_cnt = 0;
ctx->loop_cnt = 0;
}
void
rte_jobstats_set_target(struct rte_jobstats *job, int64_t target)
{
job->target = target;
}
int
rte_jobstats_start(struct rte_jobstats_context *ctx, struct rte_jobstats *job)
{
uint64_t now;
/* Some sanity check. */
if (unlikely(ctx == NULL || job == NULL || job->context != NULL))
return -EINVAL;
/* Link job with context object. */
job->context = ctx;
now = get_time();
ADD_TIME_MIN_MAX(ctx, management, now - ctx->state_time);
ctx->state_time = now;
return 0;
}
int
rte_jobstats_finish(struct rte_jobstats *job, int64_t job_value)
{
struct rte_jobstats_context *ctx;
uint64_t now, exec_time;
int need_update;
/* Some sanity check. */
if (unlikely(job == NULL || job->context == NULL))
return -EINVAL;
need_update = job->target != job_value;
/* Adjust period only if job is unhappy of its current period. */
if (need_update)
(*job->update_period_cb)(job, job_value);
ctx = job->context;
/* Update execution time is considered as runtime so get time after it is
* executed. */
now = get_time();
exec_time = now - ctx->state_time;
ADD_TIME_MIN_MAX(job, exec, exec_time);
ADD_TIME_MIN_MAX(ctx, exec, exec_time);
ctx->state_time = now;
ctx->loop_executed_jobs++;
ctx->job_exec_cnt++;
job->exec_cnt++;
job->context = NULL;
return need_update;
}
void
rte_jobstats_set_period(struct rte_jobstats *job, uint64_t period,
uint8_t saturate)
{
if (saturate != 0) {
if (period < job->min_period)
period = job->min_period;
else if (period > job->max_period)
period = job->max_period;
}
job->period = period;
}
void
rte_jobstats_set_min(struct rte_jobstats *job, uint64_t period)
{
job->min_period = period;
if (job->period < period)
job->period = period;
}
void
rte_jobstats_set_max(struct rte_jobstats *job, uint64_t period)
{
job->max_period = period;
if (job->period > period)
job->period = period;
}
int
rte_jobstats_init(struct rte_jobstats *job, const char *name,
uint64_t min_period, uint64_t max_period, uint64_t initial_period,
int64_t target)
{
if (job == NULL)
return -EINVAL;
job->period = initial_period;
job->min_period = min_period;
job->max_period = max_period;
job->target = target;
job->update_period_cb = &default_update_function;
rte_jobstats_reset(job);
snprintf(job->name, RTE_DIM(job->name), "%s", name == NULL ? "" : name);
job->context = NULL;
return 0;
}
void
rte_jobstats_set_update_period_function(struct rte_jobstats *job,
rte_job_update_period_cb_t update_period_cb)
{
if (update_period_cb == NULL)
update_period_cb = default_update_function;
job->update_period_cb = update_period_cb;
}
void
rte_jobstats_reset(struct rte_jobstats *job)
{
RESET_TIME_MIN_MAX(job, exec);
job->exec_cnt = 0;
}