/*
* Profiling framework for the events library
*
* Copyright (c) 2016 ARM Limited
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include "events/equeue.h"
#include <unistd.h>
#include <stdio.h>
#include <setjmp.h>
#include <stdint.h>
#include <stdlib.h>
#include <inttypes.h>
#include <sys/time.h>
// Performance measurement utils
#define PROF_RUNS 5
#define PROF_INTERVAL 100000000
#define prof_volatile(t) __attribute__((unused)) volatile t
typedef uint64_t prof_cycle_t;
static volatile prof_cycle_t prof_start_cycle;
static volatile prof_cycle_t prof_stop_cycle;
static prof_cycle_t prof_accum_cycle;
static prof_cycle_t prof_baseline_cycle;
static prof_cycle_t prof_iterations;
static const char *prof_units;
#define prof_cycle() ({ \
uint32_t a, b; \
__asm__ volatile ("rdtsc" : "=a" (a), "=d" (b)); \
((uint64_t)b << 32) | (uint64_t)a; \
})
#define prof_loop() \
for (prof_iterations = 0; \
prof_accum_cycle < PROF_INTERVAL; \
prof_iterations++)
#define prof_start() ({ \
prof_start_cycle = prof_cycle(); \
})
#define prof_stop() ({ \
prof_stop_cycle = prof_cycle(); \
prof_accum_cycle += prof_stop_cycle - prof_start_cycle; \
})
#define prof_result(value, units) ({ \
prof_accum_cycle = value+prof_baseline_cycle; \
prof_iterations = 1; \
prof_units = units; \
})
#define prof_measure(func, ...) ({ \
printf("%s: ...", #func); \
fflush(stdout); \
\
prof_units = "cycles"; \
prof_cycle_t runs[PROF_RUNS]; \
for (int i = 0; i < PROF_RUNS; i++) { \
prof_accum_cycle = 0; \
prof_iterations = 0; \
func(__VA_ARGS__); \
runs[i] = prof_accum_cycle / prof_iterations; \
} \
\
prof_cycle_t res = runs[0]; \
for (int i = 0; i < PROF_RUNS; i++) { \
if (runs[i] < res) { \
res = runs[i]; \
} \
} \
res -= prof_baseline_cycle; \
printf("\r%s: %"PRIu64" %s", #func, res, prof_units); \
\
if (!isatty(0)) { \
prof_cycle_t prev; \
while (scanf("%*[^0-9]%"PRIu64, &prev) == 0); \
int64_t perc = 100*((int64_t)prev - (int64_t)res) / (int64_t)prev; \
\
if (perc > 10) { \
printf(" (\e[32m%+"PRId64"%%\e[0m)", perc); \
} else if (perc < -10) { \
printf(" (\e[31m%+"PRId64"%%\e[0m)", perc); \
} else { \
printf(" (%+"PRId64"%%)", perc); \
} \
} \
\
printf("\n"); \
res; \
})
#define prof_baseline(func, ...) ({ \
prof_baseline_cycle = 0; \
prof_baseline_cycle = prof_measure(func, __VA_ARGS__); \
})
// Various test functions
void no_func(void *eh)
{
}
// Actual performance tests
void baseline_prof(void)
{
prof_loop() {
prof_start();
__asm__ volatile("");
prof_stop();
}
}
void equeue_tick_prof(void)
{
prof_volatile(unsigned) res;
prof_loop() {
prof_start();
res = equeue_tick();
prof_stop();
}
}
void equeue_alloc_prof(void)
{
struct equeue q;
equeue_create(&q, 32 * EQUEUE_EVENT_SIZE);
prof_loop() {
prof_start();
void *e = equeue_alloc(&q, 8 * sizeof(int));
prof_stop();
equeue_dealloc(&q, e);
}
equeue_destroy(&q);
}
void equeue_alloc_many_prof(int count)
{
struct equeue q;
equeue_create(&q, count * EQUEUE_EVENT_SIZE);
void *es[count];
for (int i = 0; i < count; i++) {
es[i] = equeue_alloc(&q, (i % 4) * sizeof(int));
}
for (int i = 0; i < count; i++) {
equeue_dealloc(&q, es[i]);
}
prof_loop() {
prof_start();
void *e = equeue_alloc(&q, 8 * sizeof(int));
prof_stop();
equeue_dealloc(&q, e);
}
equeue_destroy(&q);
}
void equeue_post_prof(void)
{
struct equeue q;
equeue_create(&q, EQUEUE_EVENT_SIZE);
prof_loop() {
void *e = equeue_alloc(&q, 0);
prof_start();
int id = equeue_post(&q, no_func, e);
prof_stop();
equeue_cancel(&q, id);
}
equeue_destroy(&q);
}
void equeue_post_many_prof(int count)
{
struct equeue q;
equeue_create(&q, count * EQUEUE_EVENT_SIZE);
for (int i = 0; i < count - 1; i++) {
equeue_call(&q, no_func, 0);
}
prof_loop() {
void *e = equeue_alloc(&q, 0);
prof_start();
int id = equeue_post(&q, no_func, e);
prof_stop();
equeue_cancel(&q, id);
}
equeue_destroy(&q);
}
void equeue_post_future_prof(void)
{
struct equeue q;
equeue_create(&q, EQUEUE_EVENT_SIZE);
prof_loop() {
void *e = equeue_alloc(&q, 0);
equeue_event_delay(e, 1000);
prof_start();
int id = equeue_post(&q, no_func, e);
prof_stop();
equeue_cancel(&q, id);
}
equeue_destroy(&q);
}
void equeue_post_future_many_prof(int count)
{
struct equeue q;
equeue_create(&q, count * EQUEUE_EVENT_SIZE);
for (int i = 0; i < count - 1; i++) {
equeue_call(&q, no_func, 0);
}
prof_loop() {
void *e = equeue_alloc(&q, 0);
equeue_event_delay(e, 1000);
prof_start();
int id = equeue_post(&q, no_func, e);
prof_stop();
equeue_cancel(&q, id);
}
equeue_destroy(&q);
}
void equeue_dispatch_prof(void)
{
struct equeue q;
equeue_create(&q, EQUEUE_EVENT_SIZE);
prof_loop() {
equeue_call(&q, no_func, 0);
prof_start();
equeue_dispatch(&q, 0);
prof_stop();
}
equeue_destroy(&q);
}
void equeue_dispatch_many_prof(int count)
{
struct equeue q;
equeue_create(&q, count * EQUEUE_EVENT_SIZE);
prof_loop() {
for (int i = 0; i < count; i++) {
equeue_call(&q, no_func, 0);
}
prof_start();
equeue_dispatch(&q, 0);
prof_stop();
}
equeue_destroy(&q);
}
void equeue_cancel_prof(void)
{
struct equeue q;
equeue_create(&q, EQUEUE_EVENT_SIZE);
prof_loop() {
int id = equeue_call(&q, no_func, 0);
prof_start();
equeue_cancel(&q, id);
prof_stop();
}
equeue_destroy(&q);
}
void equeue_cancel_many_prof(int count)
{
struct equeue q;
equeue_create(&q, count * EQUEUE_EVENT_SIZE);
for (int i = 0; i < count - 1; i++) {
equeue_call(&q, no_func, 0);
}
prof_loop() {
int id = equeue_call(&q, no_func, 0);
prof_start();
equeue_cancel(&q, id);
prof_stop();
}
equeue_destroy(&q);
}
void equeue_alloc_size_prof(void)
{
size_t size = 32 * EQUEUE_EVENT_SIZE;
struct equeue q;
equeue_create(&q, size);
equeue_alloc(&q, 0);
prof_result(size - q.slab.size, "bytes");
equeue_destroy(&q);
}
void equeue_alloc_many_size_prof(int count)
{
size_t size = count * EQUEUE_EVENT_SIZE;
struct equeue q;
equeue_create(&q, size);
for (int i = 0; i < count; i++) {
equeue_alloc(&q, (i % 4) * sizeof(int));
}
prof_result(size - q.slab.size, "bytes");
equeue_destroy(&q);
}
void equeue_alloc_fragmented_size_prof(int count)
{
size_t size = count * EQUEUE_EVENT_SIZE;
struct equeue q;
equeue_create(&q, size);
void *es[count];
for (int i = 0; i < count; i++) {
es[i] = equeue_alloc(&q, (i % 4) * sizeof(int));
}
for (int i = 0; i < count; i++) {
equeue_dealloc(&q, es[i]);
}
for (int i = count - 1; i >= 0; i--) {
es[i] = equeue_alloc(&q, (i % 4) * sizeof(int));
}
for (int i = count - 1; i >= 0; i--) {
equeue_dealloc(&q, es[i]);
}
for (int i = 0; i < count; i++) {
equeue_alloc(&q, (i % 4) * sizeof(int));
}
prof_result(size - q.slab.size, "bytes");
equeue_destroy(&q);
}
// Entry point
int main()
{
printf("beginning profiling...\n");
prof_baseline(baseline_prof);
prof_measure(equeue_tick_prof);
prof_measure(equeue_alloc_prof);
prof_measure(equeue_post_prof);
prof_measure(equeue_post_future_prof);
prof_measure(equeue_dispatch_prof);
prof_measure(equeue_cancel_prof);
prof_measure(equeue_alloc_many_prof, 1000);
prof_measure(equeue_post_many_prof, 1000);
prof_measure(equeue_post_future_many_prof, 1000);
prof_measure(equeue_dispatch_many_prof, 100);
prof_measure(equeue_cancel_many_prof, 100);
prof_measure(equeue_alloc_size_prof);
prof_measure(equeue_alloc_many_size_prof, 1000);
prof_measure(equeue_alloc_fragmented_size_prof, 1000);
printf("done!\n");
}
