lkml.org 
[lkml]   [2015]   [Jan]   [29]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
    Patch in this message
    /
    From
    Subject[PATCH 38/42] perf session: Handle index files generally
    Date
    The current code assumes that the number of index item and cpu are
    matched so it creates that number of threads. But it's not the case
    of non-system-wide session or data came from different machine.

    Just creates threads at most number of online cpus and process data.

    Signed-off-by: Namhyung Kim <namhyung@kernel.org>
    ---
    tools/perf/util/session.c | 68 ++++++++++++++++++++++++++++++++++-------------
    tools/perf/util/tool.h | 1 -
    2 files changed, 50 insertions(+), 19 deletions(-)

    diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
    index 6d34c880010f..ccf9371ef292 100644
    --- a/tools/perf/util/session.c
    +++ b/tools/perf/util/session.c
    @@ -1431,26 +1431,51 @@ static struct ui_progress_ops mt_progress__ops = {
    .update = mt_progress__update,
    };

    +static int perf_session__get_index(struct perf_session *session)
    +{
    + int ret;
    + static unsigned index;
    + static pthread_mutex_t idx_lock = PTHREAD_MUTEX_INITIALIZER;
    +
    + pthread_mutex_lock(&idx_lock);
    + if (index < session->header.nr_index)
    + ret = index++;
    + else
    + ret = -1;
    + pthread_mutex_unlock(&idx_lock);
    +
    + return ret;
    +}
    +
    static void *processing_thread_idx(void *arg)
    {
    struct perf_tool_mt *mt_tool = arg;
    struct perf_session *session = mt_tool->session;
    int fd = perf_data_file__fd(session->file);
    - u64 offset = session->header.index[mt_tool->idx].offset;
    - u64 size = session->header.index[mt_tool->idx].size;
    u64 file_size = perf_data_file__size(session->file);
    + int idx;

    - ui_progress__init(&mt_tool->prog, size, "");
    + while ((idx = perf_session__get_index(session)) >= 0) {
    + u64 offset = session->header.index[idx].offset;
    + u64 size = session->header.index[idx].size;
    + struct perf_tool_mt *mtt = &mt_tool[idx];

    - pr_debug("processing samples using thread [%d]\n", mt_tool->idx);
    - if (__perf_session__process_events(session, &mt_tool->stats,
    - fd, offset, size, file_size,
    - &mt_tool->tool, &mt_tool->prog) < 0) {
    - pr_err("processing samples failed (thread [%d)\n", mt_tool->idx);
    - return NULL;
    + if (size == 0)
    + continue;
    +
    + pr_debug("processing samples [index %d]\n", idx);
    +
    + ui_progress__init(&mtt->prog, size, "");
    +
    + if (__perf_session__process_events(session, &mtt->stats,
    + fd, offset, size, file_size,
    + &mtt->tool, &mtt->prog) < 0) {
    + pr_err("processing samples failed [index %d]\n", idx);
    + return NULL;
    + }
    + pr_debug("processing samples done [index %d]\n", idx);
    }

    - pr_debug("processing samples done for thread [%d]\n", mt_tool->idx);
    return arg;
    }

    @@ -1468,6 +1493,7 @@ int perf_session__process_events_mt(struct perf_session *session,
    int err, i, k;
    int nr_index = session->header.nr_index;
    u64 size = perf_data_file__size(file);
    + int nr_thread = sysconf(_SC_NPROCESSORS_ONLN);

    if (perf_session__register_idle_thread(session) == NULL)
    return -ENOMEM;
    @@ -1491,10 +1517,6 @@ int perf_session__process_events_mt(struct perf_session *session,
    ui_progress__ops = &mt_progress__ops;
    ui_progress__ops->finish = orig_progress__ops->finish;

    - th_id = calloc(nr_index, sizeof(*th_id));
    - if (th_id == NULL)
    - goto out;
    -
    mt_tools = calloc(nr_index, sizeof(*mt_tools));
    if (mt_tools == NULL)
    goto out;
    @@ -1513,20 +1535,30 @@ int perf_session__process_events_mt(struct perf_session *session,

    mt->session = session;
    mt->tool.ordered_events = false;
    - mt->idx = i;
    mt->priv = arg;
    mt->global_prog = &prog;
    -
    - pthread_create(&th_id[i], NULL, processing_thread_idx, mt);
    }

    - for (i = 0; i < nr_index; i++) {
    + if (nr_thread > nr_index)
    + nr_thread = nr_index;
    +
    + th_id = calloc(nr_thread, sizeof(*th_id));
    + if (th_id == NULL)
    + goto out;
    +
    + for (i = 0; i < nr_thread; i++)
    + pthread_create(&th_id[i], NULL, processing_thread_idx, mt_tools);
    +
    + for (i = 0; i < nr_thread; i++) {
    pthread_join(th_id[i], (void **)&mt);
    if (mt == NULL) {
    err = -EINVAL;
    continue;
    }
    + }

    + for (i = 0; i < nr_index; i++) {
    + mt = &mt_tools[i];
    events_stats__add(&session->stats, &mt->stats);

    evlist__for_each(evlist, evsel) {
    diff --git a/tools/perf/util/tool.h b/tools/perf/util/tool.h
    index aa7f110b9425..e52c936d1b9e 100644
    --- a/tools/perf/util/tool.h
    +++ b/tools/perf/util/tool.h
    @@ -55,7 +55,6 @@ struct perf_tool_mt {
    struct perf_session *session;
    struct ui_progress prog;
    struct ui_progress *global_prog;
    - int idx;

    void *priv;
    };
    --
    2.2.2


    \
     
     \ /
      Last update: 2015-01-29 09:21    [W:4.180 / U:0.192 seconds]
    ©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site