From: christian.babeux@efficios.com (Christian Babeux)
Subject: [lttng-dev] [RFC PATCH lttng-ust] Introduce hash table for lttng_create_event_if_missing()
Date: Wed, 16 Jan 2013 11:19:19 -0500 [thread overview]
Message-ID: <CAGDH53nkNuNsaAPY0oKS4LNOHEeDnAMk4=+EopBMg104vu-2bw@mail.gmail.com> (raw)
In-Reply-To: <20130116151446.GA11082@Krystal>
I was able to reproduce this overhead issue on my machine.
This patch fixed it.
Acked-by: Christian Babeux <christian.babeux at efficios.com>
On Wed, Jan 16, 2013 at 10:14 AM, Mathieu Desnoyers
<mathieu.desnoyers at efficios.com> wrote:
> lttng_create_event_if_missing() takes a lot of CPU time with stress-test
> applications containing 1000 different TRACEPOINT_EVENT() and 1000
> individual tracepoint() call-site.
>
> With tracing disabled:
>
> time ./AppWith1000_lines_TP 0
>
> real 0m2.487s
> user 0m2.424s
> sys 0m0.000s
>
>
> Introducing this hash table cuts the overhead very significantly when
> tracing is enabled:
>
> Before patch:
>
> 72.16% AppWith1000_lin liblttng-ust.so.0.0.0 [.] lttng_create_event_if_missing
> 25.64% AppWith1000_lin AppWith1000_lines_TP [.] a(int)
>
> time ./AppWith1000_lines_TP 0
>
> real 0m7.946s
> user 0m7.864s
> sys 0m0.000s
>
>
> After patch:
>
> 89.13% AppWith1000_lin AppWith1000_lines_TP [.] a(int)
> 3.30% AppWith1000_lin liblttng-ust.so.0.0.0 [.] lttng_create_event_if_missing
>
> time ./AppWith1000_lines_TP 0
>
> real 0m2.762s
> user 0m2.692s
> sys 0m0.004s
>
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
> ---
> diff --git a/include/lttng/ust-events.h b/include/lttng/ust-events.h
> index 095fb1e..1db8e58 100644
> --- a/include/lttng/ust-events.h
> +++ b/include/lttng/ust-events.h
> @@ -362,6 +362,7 @@ struct lttng_event {
> int has_enablers_without_bytecode;
> /* Backward references: list of lttng_enabler_ref (ref to enablers) */
> struct cds_list_head enablers_ref_head;
> + struct cds_hlist_node hlist; /* session ht of events */
> };
>
> struct channel;
> @@ -467,6 +468,7 @@ struct lttng_session {
> /* New UST 2.1 */
> /* List of enablers */
> struct cds_list_head enablers_head;
> + struct lttng_ust_event_ht events_ht; /* ht of events */
> };
>
> struct lttng_transport {
> diff --git a/liblttng-ust/lttng-events.c b/liblttng-ust/lttng-events.c
> index 5e6c90c..f70c65f 100644
> --- a/liblttng-ust/lttng-events.c
> +++ b/liblttng-ust/lttng-events.c
> @@ -127,7 +127,7 @@ void synchronize_trace(void)
> struct lttng_session *lttng_session_create(void)
> {
> struct lttng_session *session;
> - int ret;
> + int ret, i;
>
> session = zmalloc(sizeof(struct lttng_session));
> if (!session)
> @@ -135,6 +135,8 @@ struct lttng_session *lttng_session_create(void)
> CDS_INIT_LIST_HEAD(&session->chan_head);
> CDS_INIT_LIST_HEAD(&session->events_head);
> CDS_INIT_LIST_HEAD(&session->enablers_head);
> + for (i = 0; i < LTTNG_UST_EVENT_HT_SIZE; i++)
> + CDS_INIT_HLIST_HEAD(&session->events_ht.table[i]);
> ret = lttng_ust_uuid_generate(session->uuid);
> if (ret != 0) {
> session->uuid[0] = '\0';
> @@ -329,17 +331,19 @@ int lttng_event_create(const struct lttng_event_desc *desc,
> {
> const char *event_name = desc->name;
> struct lttng_event *event;
> + struct cds_hlist_head *head;
> + struct cds_hlist_node *node;
> int ret = 0;
> + size_t name_len = strlen(event_name);
> + uint32_t hash;
>
> if (chan->used_event_id == -1U) {
> ret = -ENOMEM;
> goto full;
> }
> - /*
> - * This is O(n^2) (for each event, the loop is called at event
> - * creation). Might require a hash if we have lots of events.
> - */
> - cds_list_for_each_entry(event, &chan->session->events_head, node) {
> + hash = jhash(event_name, name_len, 0);
> + head = &chan->session->events_ht.table[hash & (LTTNG_UST_EVENT_HT_SIZE - 1)];
> + cds_hlist_for_each_entry(event, node, head, hlist) {
> assert(event->desc);
> if (!strncmp(event->desc->name,
> desc->name,
> @@ -380,6 +384,7 @@ int lttng_event_create(const struct lttng_event_desc *desc,
> if (ret)
> goto statedump_error;
> cds_list_add(&event->node, &chan->session->events_head);
> + cds_hlist_add_head(&event->hlist, head);
> return 0;
>
> statedump_error:
> @@ -499,17 +504,24 @@ void lttng_create_event_if_missing(struct lttng_enabler *enabler)
> cds_list_for_each_entry(probe_desc, probe_list, head) {
> for (i = 0; i < probe_desc->nr_events; i++) {
> int found = 0, ret;
> + struct cds_hlist_head *head;
> + struct cds_hlist_node *node;
> + const char *event_name;
> + size_t name_len;
> + uint32_t hash;
>
> desc = probe_desc->event_desc[i];
> if (!lttng_desc_match_enabler(desc, enabler))
> continue;
> + event_name = desc->name;
> + name_len = strlen(event_name);
>
> /*
> - * For each event in session event list,
> - * check if already created.
> + * Check if already created.
> */
> - cds_list_for_each_entry(event,
> - &session->events_head, node) {
> + hash = jhash(event_name, name_len, 0);
> + head = &session->events_ht.table[hash & (LTTNG_UST_EVENT_HT_SIZE - 1)];
> + cds_hlist_for_each_entry(event, node, head, hlist) {
> if (event->desc == desc)
> found = 1;
> }
>
> --
> Mathieu Desnoyers
> EfficiOS Inc.
> http://www.efficios.com
>
> _______________________________________________
> lttng-dev mailing list
> lttng-dev at lists.lttng.org
> http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
next prev parent reply other threads:[~2013-01-16 16:19 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-01-16 15:14 Mathieu Desnoyers
2013-01-16 15:53 ` David Goulet
2013-01-16 16:19 ` Christian Babeux [this message]
2013-01-16 17:58 ` Mathieu Desnoyers
2013-01-16 21:37 Anik Mishra
2013-01-21 19:06 ` Mathieu Desnoyers
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAGDH53nkNuNsaAPY0oKS4LNOHEeDnAMk4=+EopBMg104vu-2bw@mail.gmail.com' \
--to=christian.babeux@efficios.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox