OSDN Git Service

perf bench numa: Show more stats of particular threads in verbose mode
authorPetr Holasek <pholasek@redhat.com>
Thu, 16 Apr 2015 15:38:18 +0000 (17:38 +0200)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Mon, 4 May 2015 15:43:41 +0000 (12:43 -0300)
In verbose mode perf bench numa shows also GB/s speed, system and user cpu
time for each particular thread. Using of getrusage() can provide much more
per process or per thread stats in future.

Signed-off-by: Petr Holasek <pholasek@redhat.com>
Reviewed-by: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Link: http://lkml.kernel.org/r/1429198699-25039-3-git-send-email-pholasek@redhat.com
[ Rename 'usage' variable to not shadow util.h's usage() ]
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/bench/numa.c

index ebfa163..0b704c5 100644 (file)
@@ -23,6 +23,7 @@
 #include <pthread.h>
 #include <sys/mman.h>
 #include <sys/time.h>
+#include <sys/resource.h>
 #include <sys/wait.h>
 #include <sys/prctl.h>
 #include <sys/types.h>
@@ -51,6 +52,9 @@ struct thread_data {
        unsigned int            loops_done;
        u64                     val;
        u64                     runtime_ns;
+       u64                     system_time_ns;
+       u64                     user_time_ns;
+       double                  speed_gbs;
        pthread_mutex_t         *process_lock;
 };
 
@@ -1034,6 +1038,7 @@ static void *worker_thread(void *__tdata)
        u64 bytes_done;
        long work_done;
        u32 l;
+       struct rusage rusage;
 
        bind_to_cpumask(td->bind_cpumask);
        bind_to_memnode(td->bind_node);
@@ -1186,6 +1191,13 @@ static void *worker_thread(void *__tdata)
        timersub(&stop, &start0, &diff);
        td->runtime_ns = diff.tv_sec * 1000000000ULL;
        td->runtime_ns += diff.tv_usec * 1000ULL;
+       td->speed_gbs = bytes_done / (td->runtime_ns / 1e9) / 1e9;
+
+       getrusage(RUSAGE_THREAD, &rusage);
+       td->system_time_ns = rusage.ru_stime.tv_sec * 1000000000ULL;
+       td->system_time_ns += rusage.ru_stime.tv_usec * 1000ULL;
+       td->user_time_ns = rusage.ru_utime.tv_sec * 1000000000ULL;
+       td->user_time_ns += rusage.ru_utime.tv_usec * 1000ULL;
 
        free_data(thread_data, g->p.bytes_thread);
 
@@ -1412,7 +1424,7 @@ static int __bench_numa(const char *name)
        double runtime_sec_min;
        int wait_stat;
        double bytes;
-       int i, t;
+       int i, t, p;
 
        if (init())
                return -1;
@@ -1548,6 +1560,24 @@ static int __bench_numa(const char *name)
        print_res(name, bytes / runtime_sec_max / 1e9,
                "GB/sec,", "total-speed",       "GB/sec total speed");
 
+       if (g->p.show_details >= 2) {
+               char tname[32];
+               struct thread_data *td;
+               for (p = 0; p < g->p.nr_proc; p++) {
+                       for (t = 0; t < g->p.nr_threads; t++) {
+                               memset(tname, 0, 32);
+                               td = g->threads + p*g->p.nr_threads + t;
+                               snprintf(tname, 32, "process%d:thread%d", p, t);
+                               print_res(tname, td->speed_gbs,
+                                       "GB/sec",       "thread-speed", "GB/sec/thread speed");
+                               print_res(tname, td->system_time_ns / 1e9,
+                                       "secs", "thread-system-time", "system CPU time/thread");
+                               print_res(tname, td->user_time_ns / 1e9,
+                                       "secs", "thread-user-time", "user CPU time/thread");
+                       }
+               }
+       }
+
        free(pids);
 
        deinit();