perf sched: Import schedbench.c
Ingo Molnar [Fri, 11 Sep 2009 10:12:54 +0000 (12:12 +0200)]
Import the schedbench.c tool that i wrote some time ago to
simulate scheduler behavior but never finished. It's a good
basis for perf sched nevertheless.

Most of its guts are not hooked up to the perf event loop
yet - that will be done in the patches to come.

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

tools/perf/builtin-sched.c
tools/perf/util/trace-event-parse.c
tools/perf/util/trace-event.h

index 60228d9..c66e6a3 100644 (file)
 #include "util/debug.h"
 
 #include "util/trace-event.h"
+#include <sys/types.h>
 
-static char            const *input_name = "perf.data";
-static int             input;
-static unsigned long   page_size;
-static unsigned long   mmap_window = 32;
+static char                    const *input_name = "perf.data";
+static int                     input;
+static unsigned long           page_size;
+static unsigned long           mmap_window = 32;
 
-static unsigned long   total = 0;
-static unsigned long   total_comm = 0;
+static unsigned long           total_comm = 0;
 
-static struct rb_root  threads;
-static struct thread   *last_match;
+static struct rb_root          threads;
+static struct thread           *last_match;
 
-static struct perf_header *header;
-static u64             sample_type;
+static struct perf_header      *header;
+static u64                     sample_type;
 
 
+/*
+ * Scheduler benchmarks
+ */
+#include <sys/resource.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <sys/prctl.h>
+
+#include <linux/unistd.h>
+
+#include <semaphore.h>
+#include <pthread.h>
+#include <signal.h>
+#include <values.h>
+#include <string.h>
+#include <unistd.h>
+#include <stdlib.h>
+#include <assert.h>
+#include <fcntl.h>
+#include <time.h>
+#include <math.h>
+
+#include <stdio.h>
+
+#define PR_SET_NAME    15               /* Set process name */
+
+#define BUG_ON(x)      assert(!(x))
+
+#define DEBUG          1
+
+typedef unsigned long long nsec_t;
+
+#define printk(x...)           do { printf(x); fflush(stdout); } while (0)
+
+nsec_t prev_printk;
+
+#define __dprintk(x,y...) do {                                          \
+       nsec_t __now = get_nsecs(), __delta = __now - prev_printk;       \
+                                                                        \
+       prev_printk = __now;                                             \
+                                                                        \
+       printf("%.3f [%Ld] [%.3f]: " x, (double)__now/1e6, __now, (double)__delta/1e6, y);\
+} while (0)
+
+#if !DEBUG
+# define dprintk(x...) do { } while (0)
+#else
+# define dprintk(x...) __dprintk(x)
+#endif
+
+#define __DP()         __dprintk("parent: line %d\n", __LINE__)
+#define DP()           dprintk("parent: line %d\n", __LINE__)
+#define D()            dprintk("task %ld: line %d\n", this_task->nr, __LINE__)
+
+
+static nsec_t run_measurement_overhead;
+static nsec_t sleep_measurement_overhead;
+
+static nsec_t get_nsecs(void)
+{
+       struct timespec ts;
+
+       clock_gettime(CLOCK_MONOTONIC, &ts);
+
+       return ts.tv_sec * 1000000000ULL + ts.tv_nsec;
+}
+
+static void burn_nsecs(nsec_t nsecs)
+{
+       nsec_t T0 = get_nsecs(), T1;
+
+       do {
+               T1 = get_nsecs();
+       } while (T1 + run_measurement_overhead < T0 + nsecs);
+}
+
+static void sleep_nsecs(nsec_t nsecs)
+{
+       struct timespec ts;
+
+       ts.tv_nsec = nsecs % 999999999;
+       ts.tv_sec = nsecs / 999999999;
+
+       nanosleep(&ts, NULL);
+}
+
+static void calibrate_run_measurement_overhead(void)
+{
+       nsec_t T0, T1, delta, min_delta = 1000000000ULL;
+       int i;
+
+       for (i = 0; i < 10; i++) {
+               T0 = get_nsecs();
+               burn_nsecs(0);
+               T1 = get_nsecs();
+               delta = T1-T0;
+               min_delta = min(min_delta, delta);
+       }
+       run_measurement_overhead = min_delta;
+
+       printk("run measurement overhead: %Ld nsecs\n", min_delta);
+}
+
+static void calibrate_sleep_measurement_overhead(void)
+{
+       nsec_t T0, T1, delta, min_delta = 1000000000ULL;
+       int i;
+
+       for (i = 0; i < 10; i++) {
+               T0 = get_nsecs();
+               sleep_nsecs(10000);
+               T1 = get_nsecs();
+               delta = T1-T0;
+               min_delta = min(min_delta, delta);
+       }
+       min_delta -= 10000;
+       sleep_measurement_overhead = min_delta;
+
+       printk("sleep measurement overhead: %Ld nsecs\n", min_delta);
+}
+
+#define COMM_LEN       20
+#define SYM_LEN                129
+
+#define MAX_PID                65536
+
+static unsigned long nr_tasks;
+
+struct sched_event;
+
+struct task_desc {
+       unsigned long           nr;
+       unsigned long           pid;
+       char                    comm[COMM_LEN];
+
+       unsigned long           nr_events;
+       unsigned long           curr_event;
+       struct sched_event      **events;
+
+       pthread_t               thread;
+       sem_t                   sleep_sem;
+
+       sem_t                   ready_for_work;
+       sem_t                   work_done_sem;
+
+       nsec_t                  cpu_usage;
+};
+
+enum sched_event_type {
+       SCHED_EVENT_RUN,
+       SCHED_EVENT_SLEEP,
+       SCHED_EVENT_WAKEUP,
+};
+
+struct sched_event {
+       enum sched_event_type   type;
+       nsec_t                  timestamp;
+       nsec_t                  duration;
+       unsigned long           nr;
+       int                     specific_wait;
+       sem_t                   *wait_sem;
+       struct task_desc        *wakee;
+};
+
+static struct task_desc                *pid_to_task[MAX_PID];
+
+static struct task_desc                **tasks;
+
+static pthread_mutex_t         start_work_mutex = PTHREAD_MUTEX_INITIALIZER;
+static nsec_t                  start_time;
+
+static pthread_mutex_t         work_done_wait_mutex = PTHREAD_MUTEX_INITIALIZER;
+
+static unsigned long           nr_run_events;
+static unsigned long           nr_sleep_events;
+static unsigned long           nr_wakeup_events;
+
+static unsigned long           nr_sleep_corrections;
+static unsigned long           nr_run_events_optimized;
+
+static struct sched_event *
+get_new_event(struct task_desc *task, nsec_t timestamp)
+{
+       struct sched_event *event = calloc(1, sizeof(*event));
+       unsigned long idx = task->nr_events;
+       size_t size;
+
+       event->timestamp = timestamp;
+       event->nr = idx;
+
+       task->nr_events++;
+       size = sizeof(struct sched_event *) * task->nr_events;
+       task->events = realloc(task->events, size);
+       BUG_ON(!task->events);
+
+       task->events[idx] = event;
+
+       return event;
+}
+
+static struct sched_event *last_event(struct task_desc *task)
+{
+       if (!task->nr_events)
+               return NULL;
+
+       return task->events[task->nr_events - 1];
+}
+
+static void
+add_sched_event_run(struct task_desc *task, nsec_t timestamp,
+                   unsigned long duration)
+{
+       struct sched_event *event, *curr_event = last_event(task);
+
+       /*
+        * optimize an existing RUN event by merging this one
+        * to it:
+        */
+       if (curr_event && curr_event->type == SCHED_EVENT_RUN) {
+               nr_run_events_optimized++;
+               curr_event->duration += duration;
+               return;
+       }
+
+       event = get_new_event(task, timestamp);
+
+       event->type = SCHED_EVENT_RUN;
+       event->duration = duration;
+
+       nr_run_events++;
+}
+
+static unsigned long targetless_wakeups;
+static unsigned long multitarget_wakeups;
+
+static void
+add_sched_event_wakeup(struct task_desc *task, nsec_t timestamp,
+                      struct task_desc *wakee)
+{
+       struct sched_event *event, *wakee_event;
+
+       event = get_new_event(task, timestamp);
+       event->type = SCHED_EVENT_WAKEUP;
+       event->wakee = wakee;
+
+       wakee_event = last_event(wakee);
+       if (!wakee_event || wakee_event->type != SCHED_EVENT_SLEEP) {
+               targetless_wakeups++;
+               return;
+       }
+       if (wakee_event->wait_sem) {
+               multitarget_wakeups++;
+               return;
+       }
+
+       wakee_event->wait_sem = calloc(1, sizeof(*wakee_event->wait_sem));
+       sem_init(wakee_event->wait_sem, 0, 0);
+       wakee_event->specific_wait = 1;
+       event->wait_sem = wakee_event->wait_sem;
+
+       nr_wakeup_events++;
+}
+
+static void
+add_sched_event_sleep(struct task_desc *task, nsec_t timestamp,
+                     unsigned long uninterruptible __used)
+{
+       struct sched_event *event = get_new_event(task, timestamp);
+
+       event->type = SCHED_EVENT_SLEEP;
+
+       nr_sleep_events++;
+}
+
+static struct task_desc *register_pid(unsigned long pid, const char *comm)
+{
+       struct task_desc *task;
+
+       BUG_ON(pid >= MAX_PID);
+
+       task = pid_to_task[pid];
+
+       if (task)
+               return task;
+
+       task = calloc(1, sizeof(*task));
+       task->pid = pid;
+       task->nr = nr_tasks;
+       strcpy(task->comm, comm);
+       /*
+        * every task starts in sleeping state - this gets ignored
+        * if there's no wakeup pointing to this sleep state:
+        */
+       add_sched_event_sleep(task, 0, 0);
+
+       pid_to_task[pid] = task;
+       nr_tasks++;
+       tasks = realloc(tasks, nr_tasks*sizeof(struct task_task *));
+       BUG_ON(!tasks);
+       tasks[task->nr] = task;
+
+       printk("registered task #%ld, PID %ld (%s)\n", nr_tasks, pid, comm);
+
+       return task;
+}
+
+
+static int first_trace_line = 1;
+
+static nsec_t first_timestamp;
+static nsec_t prev_timestamp;
+
+void parse_line(char *line);
+
+void parse_line(char *line)
+{
+       unsigned long param1 = 0, param2 = 0;
+       char comm[COMM_LEN], comm2[COMM_LEN];
+       unsigned long pid, pid2, timestamp0;
+       struct task_desc *task, *task2;
+       char func_str[SYM_LEN];
+       nsec_t timestamp;
+       int ret;
+
+       //"   <idle> 0     0D.s3    0us+: try_to_wake_up <events/0 9> (1 0)"
+       ret = sscanf(line, "%20s %5ld %*s %ldus%*c:"
+                          " %128s <%20s %ld> (%ld %ld)\n",
+               comm, &pid, &timestamp0,
+               func_str, comm2, &pid2, &param1, &param2);
+       dprintk("ret: %d\n", ret);
+       if (ret != 8)
+               return;
+
+       timestamp = timestamp0 * 1000LL;
+
+       if (first_trace_line) {
+               first_trace_line = 0;
+               first_timestamp = timestamp;
+       }
+
+       timestamp -= first_timestamp;
+       BUG_ON(timestamp < prev_timestamp);
+       prev_timestamp = timestamp;
+
+       dprintk("parsed: %s - %ld %Ld: %s - <%s %ld> (%ld %ld)\n",
+               comm,
+               pid,
+               timestamp, 
+               func_str,
+               comm2,
+               pid2,
+               param1,
+               param2);
+
+       task = register_pid(pid, comm);
+       task2 = register_pid(pid2, comm2);
+
+       if (!strcmp(func_str, "update_curr")) {
+               dprintk("%Ld: task %ld runs for %ld nsecs\n",
+                       timestamp, task->nr, param1);
+               add_sched_event_run(task, timestamp, param1);
+       } else if (!strcmp(func_str, "try_to_wake_up")) {
+               dprintk("%Ld: task %ld wakes up task %ld\n",
+                       timestamp, task->nr, task2->nr);
+               add_sched_event_wakeup(task, timestamp, task2);
+       } else if (!strcmp(func_str, "deactivate_task")) {
+               dprintk("%Ld: task %ld goes to sleep (uninterruptible: %ld)\n",
+                       timestamp, task->nr, param1);
+               add_sched_event_sleep(task, timestamp, param1);
+       }
+}
+
+static void print_task_traces(void)
+{
+       struct task_desc *task;
+       unsigned long i;
+
+       for (i = 0; i < nr_tasks; i++) {
+               task = tasks[i];
+               printk("task %6ld (%20s:%10ld), nr_events: %ld\n",
+                       task->nr, task->comm, task->pid, task->nr_events);
+       }
+}
+
+static void add_cross_task_wakeups(void)
+{
+       struct task_desc *task1, *task2;
+       unsigned long i, j;
+
+       for (i = 0; i < nr_tasks; i++) {
+               task1 = tasks[i];
+               j = i + 1;
+               if (j == nr_tasks)
+                       j = 0;
+               task2 = tasks[j];
+               add_sched_event_wakeup(task1, 0, task2);
+       }
+}
+
+static void
+process_sched_event(struct task_desc *this_task, struct sched_event *event)
+{
+       int ret = 0;
+       nsec_t now;
+       long long delta;
+
+       now = get_nsecs();
+       delta = start_time + event->timestamp - now;
+
+       dprintk("task %ld, event #%ld, %Ld, delta: %.3f (%Ld)\n",
+               this_task->nr, event->nr, event->timestamp,
+               (double)delta/1e6, delta);
+
+       if (0 && delta > 0) {
+               dprintk("%.3f: task %ld FIX %.3f\n",
+                       (double)event->timestamp/1e6,
+                       this_task->nr,
+                       (double)delta/1e6);
+               sleep_nsecs(start_time + event->timestamp - now);
+               nr_sleep_corrections++;
+       }
+
+       switch (event->type) {
+               case SCHED_EVENT_RUN:
+                       dprintk("%.3f: task %ld RUN for %.3f\n",
+                               (double)event->timestamp/1e6,
+                               this_task->nr,
+                               (double)event->duration/1e6);
+                       burn_nsecs(event->duration);
+                       break;
+               case SCHED_EVENT_SLEEP:
+                       dprintk("%.3f: task %ld %s SLEEP\n",
+                               (double)event->timestamp/1e6,
+                               this_task->nr, event->wait_sem ? "" : "SKIP");
+                       if (event->wait_sem)
+                               ret = sem_wait(event->wait_sem);
+                       BUG_ON(ret);
+                       break;
+               case SCHED_EVENT_WAKEUP:
+                       dprintk("%.3f: task %ld WAKEUP => task %ld\n",
+                               (double)event->timestamp/1e6,
+                               this_task->nr,
+                               event->wakee->nr);
+                       if (event->wait_sem)
+                               ret = sem_post(event->wait_sem);
+                       BUG_ON(ret);
+                       break;
+               default:
+                       BUG_ON(1);
+       }
+}
+
+static nsec_t get_cpu_usage_nsec_parent(void)
+{
+       struct rusage ru;
+       nsec_t sum;
+       int err;
+
+       err = getrusage(RUSAGE_SELF, &ru);
+       BUG_ON(err);
+
+       sum =  ru.ru_utime.tv_sec*1e9 + ru.ru_utime.tv_usec*1e3;
+       sum += ru.ru_stime.tv_sec*1e9 + ru.ru_stime.tv_usec*1e3;
+
+       return sum;
+}
+
+static nsec_t get_cpu_usage_nsec_self(void)
+{
+       char filename [] = "/proc/1234567890/sched";
+       unsigned long msecs, nsecs;
+       char *line = NULL;
+       nsec_t total = 0;
+       size_t len = 0;
+       ssize_t chars;
+       FILE *file;
+       int ret;
+
+       sprintf(filename, "/proc/%d/sched", getpid());
+       file = fopen(filename, "r");
+       BUG_ON(!file);
+
+       while ((chars = getline(&line, &len, file)) != -1) {
+               dprintk("got line with length %zu :\n", chars);
+               dprintk("%s", line);
+               ret = sscanf(line, "se.sum_exec_runtime : %ld.%06ld\n",
+                       &msecs, &nsecs);
+               if (ret == 2) {
+                       total = msecs*1e6 + nsecs;
+                       dprintk("total: (%ld.%06ld) %Ld\n",
+                               msecs, nsecs, total);
+                       break;
+               }
+       }
+       if (line)
+               free(line);
+       fclose(file);
+
+       return total;
+}
+
+static void *thread_func(void *ctx)
+{
+       struct task_desc *this_task = ctx;
+       nsec_t cpu_usage_0, cpu_usage_1;
+       unsigned long i, ret;
+       char comm2[22];
+
+       dprintk("task %ld started up.\n", this_task->nr);
+       sprintf(comm2, ":%s", this_task->comm);
+       prctl(PR_SET_NAME, comm2);
+
+again:
+       ret = sem_post(&this_task->ready_for_work);
+       BUG_ON(ret);
+       D();
+       ret = pthread_mutex_lock(&start_work_mutex);
+       BUG_ON(ret);
+       ret = pthread_mutex_unlock(&start_work_mutex);
+       BUG_ON(ret);
+       D();
+
+       cpu_usage_0 = get_cpu_usage_nsec_self();
+
+       for (i = 0; i < this_task->nr_events; i++) {
+               this_task->curr_event = i;
+               process_sched_event(this_task, this_task->events[i]);
+       }
+
+       cpu_usage_1 = get_cpu_usage_nsec_self();
+       this_task->cpu_usage = cpu_usage_1 - cpu_usage_0;
+
+       dprintk("task %ld cpu usage: %0.3f msecs\n",
+               this_task->nr, (double)this_task->cpu_usage / 1e6);
+
+       D();
+       ret = sem_post(&this_task->work_done_sem);
+       BUG_ON(ret);
+       D();
+
+       ret = pthread_mutex_lock(&work_done_wait_mutex);
+       BUG_ON(ret);
+       ret = pthread_mutex_unlock(&work_done_wait_mutex);
+       BUG_ON(ret);
+       D();
+
+       goto again;
+}
+
+static void create_tasks(void)
+{
+       struct task_desc *task;
+       pthread_attr_t attr;
+       unsigned long i;
+       int err;
+
+       err = pthread_attr_init(&attr);
+       BUG_ON(err);
+       err = pthread_attr_setstacksize(&attr, (size_t)(16*1024));
+       BUG_ON(err);
+       err = pthread_mutex_lock(&start_work_mutex);
+       BUG_ON(err);
+       err = pthread_mutex_lock(&work_done_wait_mutex);
+       BUG_ON(err);
+       for (i = 0; i < nr_tasks; i++) {
+               task = tasks[i];
+               sem_init(&task->sleep_sem, 0, 0);
+               sem_init(&task->ready_for_work, 0, 0);
+               sem_init(&task->work_done_sem, 0, 0);
+               task->curr_event = 0;
+               err = pthread_create(&task->thread, &attr, thread_func, task);
+               BUG_ON(err);
+       }
+}
+
+static nsec_t cpu_usage;
+static nsec_t runavg_cpu_usage;
+static nsec_t parent_cpu_usage;
+static nsec_t runavg_parent_cpu_usage;
+
+static void wait_for_tasks(void)
+{
+       nsec_t cpu_usage_0, cpu_usage_1;
+       struct task_desc *task;
+       unsigned long i, ret;
+
+       DP();
+       start_time = get_nsecs();
+       DP();
+       cpu_usage = 0;
+       pthread_mutex_unlock(&work_done_wait_mutex);
+
+       for (i = 0; i < nr_tasks; i++) {
+               task = tasks[i];
+               ret = sem_wait(&task->ready_for_work);
+               BUG_ON(ret);
+               sem_init(&task->ready_for_work, 0, 0);
+       }
+       ret = pthread_mutex_lock(&work_done_wait_mutex);
+       BUG_ON(ret);
+
+       cpu_usage_0 = get_cpu_usage_nsec_parent();
+
+       pthread_mutex_unlock(&start_work_mutex);
+
+#if 0
+       for (i = 0; i < nr_tasks; i++) {
+               unsigned long missed;
+
+               task = tasks[i];
+               while (task->curr_event + 1 < task->nr_events) {
+                       dprintk("parent waiting for %ld (%ld != %ld)\n",
+                               i, task->curr_event, task->nr_events);
+                       sleep_nsecs(100000000);
+               }
+               missed = task->nr_events - 1 - task->curr_event;
+               if (missed)
+                       printk("task %ld missed events: %ld\n", i, missed);
+               ret = sem_post(&task->sleep_sem);
+               BUG_ON(ret);
+       }
+#endif
+       DP();
+       for (i = 0; i < nr_tasks; i++) {
+               task = tasks[i];
+               ret = sem_wait(&task->work_done_sem);
+               BUG_ON(ret);
+               sem_init(&task->work_done_sem, 0, 0);
+               cpu_usage += task->cpu_usage;
+               task->cpu_usage = 0;
+       }
+
+       cpu_usage_1 = get_cpu_usage_nsec_parent();
+       if (!runavg_cpu_usage)
+               runavg_cpu_usage = cpu_usage;
+       runavg_cpu_usage = (runavg_cpu_usage*9 + cpu_usage)/10;
+
+       parent_cpu_usage = cpu_usage_1 - cpu_usage_0;
+       if (!runavg_parent_cpu_usage)
+               runavg_parent_cpu_usage = parent_cpu_usage;
+       runavg_parent_cpu_usage = (runavg_parent_cpu_usage*9 +
+                                  parent_cpu_usage)/10;
+
+       ret = pthread_mutex_lock(&start_work_mutex);
+       BUG_ON(ret);
+
+       for (i = 0; i < nr_tasks; i++) {
+               task = tasks[i];
+               sem_init(&task->sleep_sem, 0, 0);
+               task->curr_event = 0;
+       }
+}
+
+static int __cmd_sched(void);
+
+static void parse_trace(void)
+{
+       __cmd_sched();
+
+       printk("nr_run_events:        %ld\n", nr_run_events);
+       printk("nr_sleep_events:      %ld\n", nr_sleep_events);
+       printk("nr_wakeup_events:     %ld\n", nr_wakeup_events);
+
+       if (targetless_wakeups)
+               printk("target-less wakeups:  %ld\n", targetless_wakeups);
+       if (multitarget_wakeups)
+               printk("multi-target wakeups: %ld\n", multitarget_wakeups);
+       if (nr_run_events_optimized)
+               printk("run events optimized: %ld\n",
+                       nr_run_events_optimized);
+}
+
+static unsigned long nr_runs;
+static nsec_t sum_runtime;
+static nsec_t sum_fluct;
+static nsec_t run_avg;
+
+static void run_one_test(void)
+{
+       nsec_t T0, T1, delta, avg_delta, fluct, std_dev;
+
+       T0 = get_nsecs();
+       wait_for_tasks();
+       T1 = get_nsecs();
+
+       delta = T1 - T0;
+       sum_runtime += delta;
+       nr_runs++;
+
+       avg_delta = sum_runtime / nr_runs;
+       if (delta < avg_delta)
+               fluct = avg_delta - delta;
+       else
+               fluct = delta - avg_delta;
+       sum_fluct += fluct;
+       std_dev = sum_fluct / nr_runs / sqrt(nr_runs);
+       if (!run_avg)
+               run_avg = delta;
+       run_avg = (run_avg*9 + delta)/10;
+
+       printk("#%-3ld: %0.3f, ",
+               nr_runs, (double)delta/1000000.0);
+
+#if 0
+       printk("%0.2f +- %0.2f, ",
+               (double)avg_delta/1e6, (double)std_dev/1e6);
+#endif
+       printk("ravg: %0.2f, ",
+               (double)run_avg/1e6);
+
+       printk("cpu: %0.2f / %0.2f",
+               (double)cpu_usage/1e6, (double)runavg_cpu_usage/1e6);
+
+#if 0
+       /*
+        * rusage statistics done by the parent, these are less
+        * accurate than the sum_exec_runtime based statistics:
+        */
+       printk(" [%0.2f / %0.2f]",
+               (double)parent_cpu_usage/1e6,
+               (double)runavg_parent_cpu_usage/1e6);
+#endif
+
+       printk("\n");
+
+       if (nr_sleep_corrections)
+               printk(" (%ld sleep corrections)\n", nr_sleep_corrections);
+       nr_sleep_corrections = 0;
+}
+
+static void test_calibrations(void)
+{
+       nsec_t T0, T1;
+
+       T0 = get_nsecs();
+       burn_nsecs(1e6);
+       T1 = get_nsecs();
+
+       printk("the run test took %Ld nsecs\n", T1-T0);
+
+       T0 = get_nsecs();
+       sleep_nsecs(1e6);
+       T1 = get_nsecs();
+
+       printk("the sleep test took %Ld nsecs\n", T1-T0);
+}
+
 static int
 process_comm_event(event_t *event, unsigned long offset, unsigned long head)
 {
@@ -50,6 +798,46 @@ process_comm_event(event_t *event, unsigned long offset, unsigned long head)
        return 0;
 }
 
+static void process_sched_wakeup_event(struct event *event,
+                 int cpu __used, u64 timestamp __used, struct thread *thread __used)
+{
+       printf("sched_wakeup event %p\n", event);
+}
+
+static void process_sched_switch_event(struct event *event,
+                 int cpu __used, u64 timestamp __used, struct thread *thread __used)
+{
+       printf("sched_switch event %p\n", event);
+}
+
+static void
+process_raw_event(event_t *raw_event, void *more_data,
+                 int cpu, u64 timestamp, struct thread *thread)
+{
+       struct {
+               u32 size;
+               char data[0];
+       } *raw = more_data;
+       struct event *event;
+       int type;
+
+       type = trace_parse_common_type(raw->data);
+       event = trace_find_event(type);
+
+       /*
+        * FIXME: better resolve from pid from the struct trace_entry
+        * field, although it should be the same than this perf
+        * event pid
+        */
+       printf("id %d, type: %d, event: %s\n",
+               raw_event->header.type, type, event->name);
+
+       if (!strcmp(event->name, "sched_switch"))
+               process_sched_switch_event(event, cpu, timestamp, thread);
+       if (!strcmp(event->name, "sched_wakeup"))
+               process_sched_wakeup_event(event, cpu, timestamp, thread);
+}
+
 static int
 process_sample_event(event_t *event, unsigned long offset, unsigned long head)
 {
@@ -122,20 +910,8 @@ process_sample_event(event_t *event, unsigned long offset, unsigned long head)
                dump_printf(" ...... dso: [hypervisor]\n");
        }
 
-       if (sample_type & PERF_SAMPLE_RAW) {
-               struct {
-                       u32 size;
-                       char data[0];
-               } *raw = more_data;
-
-               /*
-                * FIXME: better resolve from pid from the struct trace_entry
-                * field, although it should be the same than this perf
-                * event pid
-                */
-               print_event(cpu, raw->data, raw->size, timestamp, thread->comm);
-       }
-       total += period;
+       if (sample_type & PERF_SAMPLE_RAW)
+               process_raw_event(event, more_data, cpu, timestamp, thread);
 
        return 0;
 }
@@ -277,6 +1053,8 @@ static const struct option options[] = {
 
 int cmd_sched(int argc, const char **argv, const char *prefix __used)
 {
+       long nr_iterations = LONG_MAX, i;
+
        symbol__init();
        page_size = getpagesize();
 
@@ -293,5 +1071,19 @@ int cmd_sched(int argc, const char **argv, const char *prefix __used)
 
        setup_pager();
 
-       return __cmd_sched();
+       calibrate_run_measurement_overhead();
+       calibrate_sleep_measurement_overhead();
+
+       test_calibrations();
+
+       parse_trace();
+       print_task_traces();
+       add_cross_task_wakeups();
+
+       create_tasks();
+       printk("------------------------------------------------------------\n");
+       for (i = 0; i < nr_iterations; i++)
+               run_one_test();
+
+       return 0;
 }
index 629e602..16cf2d5 100644 (file)
@@ -1799,7 +1799,7 @@ static int get_common_info(const char *type, int *offset, int *size)
        return 0;
 }
 
-static int parse_common_type(void *data)
+int trace_parse_common_type(void *data)
 {
        static int type_offset;
        static int type_size;
@@ -1832,7 +1832,7 @@ static int parse_common_pid(void *data)
        return read_size(data + pid_offset, pid_size);
 }
 
-static struct event *find_event(int id)
+struct event *trace_find_event(int id)
 {
        struct event *event;
 
@@ -2420,8 +2420,8 @@ get_return_for_leaf(int cpu, int cur_pid, unsigned long long cur_func,
        int type;
        int pid;
 
-       type = parse_common_type(next->data);
-       event = find_event(type);
+       type = trace_parse_common_type(next->data);
+       event = trace_find_event(type);
        if (!event)
                return NULL;
 
@@ -2502,8 +2502,8 @@ print_graph_entry_leaf(struct event *event, void *data, struct record *ret_rec)
        int type;
        int i;
 
-       type = parse_common_type(ret_rec->data);
-       ret_event = find_event(type);
+       type = trace_parse_common_type(ret_rec->data);
+       ret_event = trace_find_event(type);
 
        field = find_field(ret_event, "rettime");
        if (!field)
@@ -2696,9 +2696,9 @@ void print_event(int cpu, void *data, int size, unsigned long long nsecs,
        nsecs -= secs * NSECS_PER_SEC;
        usecs = nsecs / NSECS_PER_USEC;
 
-       type = parse_common_type(data);
+       type = trace_parse_common_type(data);
 
-       event = find_event(type);
+       event = trace_find_event(type);
        if (!event)
                die("ug! no event found for type %d", type);
 
index 420294a..bc81612 100644 (file)
@@ -234,6 +234,8 @@ extern int header_page_data_offset;
 extern int header_page_data_size;
 
 int parse_header_page(char *buf, unsigned long size);
+int trace_parse_common_type(void *data);
+struct event *trace_find_event(int id);
 
 void read_tracing_data(struct perf_counter_attr *pattrs, int nb_counters);