LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com>
To: Jin Yao <yao.jin@linux.intel.com>
Cc: jolsa@kernel.org, peterz@infradead.org, mingo@redhat.com,
	alexander.shishkin@linux.intel.com, Linux-kernel@vger.kernel.org,
	ak@linux.intel.com, kan.liang@intel.com, yao.jin@intel.com
Subject: Re: [PATCH v1 02/14] perf util: Create streams for managing top N hottest callchains
Date: Tue, 10 Mar 2020 12:11:17 -0300	[thread overview]
Message-ID: <20200310151117.GG15931@kernel.org> (raw)
In-Reply-To: <20200310070245.16314-3-yao.jin@linux.intel.com>

Em Tue, Mar 10, 2020 at 03:02:33PM +0800, Jin Yao escreveu:
> We think the stream is a callchain which is aggregated by the LBR
> records from samples. By browsing the stream, we can understand
> the code flow.
> 
> The struct callchain_node represents one callchain and we use the
> callchain_node->hit to measure the hot level of this callchain.
> Higher is hotter.
> 
> Since in perf data file, there may be many callchains so we just
> need to focus on the top N hottest callchains. N is a user defined
> parameter or just a predefined default value.
> 
> This patch saves the top N hottest callchains in 'struct stream_node'
> type array, which is defined in a per event 'struct callchain_streams'.
> 
> So now we can get the per-event top N hottest callchains.
> 
> Signed-off-by: Jin Yao <yao.jin@linux.intel.com>
> ---
>  tools/perf/util/callchain.c | 125 ++++++++++++++++++++++++++++++++++++
>  tools/perf/util/callchain.h |  16 +++++
>  2 files changed, 141 insertions(+)
> 
> diff --git a/tools/perf/util/callchain.c b/tools/perf/util/callchain.c
> index 818aa4efd386..d9c68a8e7619 100644
> --- a/tools/perf/util/callchain.c
> +++ b/tools/perf/util/callchain.c
> @@ -31,6 +31,7 @@
>  #include "callchain.h"
>  #include "branch.h"
>  #include "symbol.h"
> +#include "evlist.h"
>  #include "../perf.h"
>  
>  #define CALLCHAIN_PARAM_DEFAULT			\
> @@ -1599,3 +1600,127 @@ void callchain_cursor_reset(struct callchain_cursor *cursor)
>  	for (node = cursor->first; node != NULL; node = node->next)
>  		map__zput(node->ms.map);
>  }
> +
> +static void free_evsel_streams(struct callchain_streams *callchain_streams,
> +			       int nr_evsel)
> +{
> +	for (int i = 0; i < nr_evsel; i++) {
> +		if (callchain_streams[i].streams)
> +			free(callchain_streams[i].streams);

free(NULL) is valid, so remove that extra check and use zfree() to reset
that entry to NULL, i.e.:

	for ()
		zfree(&callchain_streams[i].streams);

> +	}
> +
> +	free(callchain_streams);
> +}
> +
> +static struct callchain_streams *create_evsel_streams(int nr_evsel,
> +						      int nr_streams_max)
> +{
> +	struct callchain_streams *callchain_streams;
> +
> +	callchain_streams = calloc(nr_evsel, sizeof(struct callchain_streams));

calloc is the right thing here, as this is an array

> +	if (!callchain_streams)
> +		return NULL;
> +
> +	for (int i = 0; i < nr_evsel; i++) {
> +		struct callchain_streams *s = &callchain_streams[i];
> +
> +		s->streams = calloc(nr_streams_max, sizeof(struct stream_node));
> +		if (!s->streams)
> +			goto err;
> +
> +		s->nr_streams_max = nr_streams_max;
> +		s->evsel_idx = -1;
> +	}
> +
> +	return callchain_streams;
> +
> +err:
> +	free_evsel_streams(callchain_streams, nr_evsel);
> +	return NULL;
> +}
> +
> +/*
> + * The cnodes with high hit number are hot callchains.
> + */
> +static void set_hot_cnode(struct callchain_streams *s,
> +			  struct callchain_node *cnode)
> +{
> +	int i, idx = 0;
> +	u64 hit;
> +
> +	if (s->nr_streams < s->nr_streams_max) {
> +		i = s->nr_streams;
> +		s->streams[i].cnode = cnode;
> +		s->nr_streams++;
> +		return;
> +	}
> +
> +	/*
> +	 * Since only a few number of hot streams, so only use simple
> +	 * way to find the cnode with smallest hit number and replace.
> +	 */
> +	hit = (s->streams[0].cnode)->hit;
> +	for (i = 1; i < s->nr_streams; i++) {
> +		if ((s->streams[i].cnode)->hit < hit) {
> +			hit = (s->streams[i].cnode)->hit;
> +			idx = i;
> +		}
> +	}
> +
> +	if (cnode->hit > hit)
> +		s->streams[idx].cnode = cnode;
> +}
> +
> +static void update_hot_streams(struct hist_entry *he,
> +			       struct callchain_streams *s)
> +{
> +	struct rb_root *root = &he->sorted_chain;
> +	struct rb_node *rb_node = rb_first(root);
> +	struct callchain_node *node;
> +
> +	while (rb_node) {
> +		node = rb_entry(rb_node, struct callchain_node, rb_node);
> +		set_hot_cnode(s, node);
> +		rb_node = rb_next(rb_node);
> +	}
> +}
> +
> +static void get_hot_streams(struct hists *hists,
> +			    struct callchain_streams *s)
> +{
> +	struct rb_node *next;
> +
> +	next = rb_first_cached(&hists->entries);
> +	while (next) {
> +		struct hist_entry *he;
> +
> +		he = rb_entry(next, struct hist_entry, rb_node);
> +		update_hot_streams(he, s);
> +		next = rb_next(&he->rb_node);
> +	}
> +}
> +
> +struct callchain_streams *callchain_evsel_streams_create(struct evlist *evlist,
> +							 int nr_streams_max,
> +							 int *nr_evsel_streams)
> +{
> +	int nr_evsel = evlist->core.nr_entries, i = 0;
> +	struct callchain_streams *callchain_streams;
> +	struct evsel *pos;
> +
> +	callchain_streams = create_evsel_streams(nr_evsel, nr_streams_max);
> +	if (!callchain_streams)
> +		return NULL;
> +
> +	evlist__for_each_entry(evlist, pos) {
> +		struct hists *hists = evsel__hists(pos);
> +
> +		hists__output_resort(hists, NULL);
> +		get_hot_streams(hists, &callchain_streams[i]);
> +		callchain_streams[i].evsel_idx = pos->idx;
> +		i++;
> +	}
> +
> +	*nr_evsel_streams = nr_evsel;
> +	return callchain_streams;
> +}
> diff --git a/tools/perf/util/callchain.h b/tools/perf/util/callchain.h
> index 706bb7bbe1e1..5852990cdf60 100644
> --- a/tools/perf/util/callchain.h
> +++ b/tools/perf/util/callchain.h
> @@ -13,6 +13,7 @@ struct ip_callchain;
>  struct map;
>  struct perf_sample;
>  struct thread;
> +struct evlist;
>  
>  #define HELP_PAD "\t\t\t\t"
>  
> @@ -159,6 +160,17 @@ struct callchain_cursor {
>  	struct callchain_cursor_node	*curr;
>  };
>  
> +struct stream_node {
> +	struct callchain_node	*cnode;
> +};
> +
> +struct callchain_streams {
> +	struct stream_node	*streams;
> +	int			nr_streams_max;
> +	int			nr_streams;
> +	int			evsel_idx;
> +};
> +
>  extern __thread struct callchain_cursor callchain_cursor;
>  
>  static inline void callchain_init(struct callchain_root *root)
> @@ -289,4 +301,8 @@ int callchain_branch_counts(struct callchain_root *root,
>  			    u64 *branch_count, u64 *predicted_count,
>  			    u64 *abort_count, u64 *cycles_count);
>  
> +struct callchain_streams *callchain_evsel_streams_create(struct evlist *evlist,
> +							 int nr_streams_max,
> +							 int *nr_evsel_streams);
> +
>  #endif	/* __PERF_CALLCHAIN_H */
> -- 
> 2.17.1
> 

-- 

- Arnaldo

  reply	other threads:[~2020-03-10 15:11 UTC|newest]

Thread overview: 23+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-03-10  7:02 [PATCH v1 00/14] perf: Stream comparison Jin Yao
2020-03-10  7:02 ` [PATCH v1 01/14] perf util: Create source line mapping table Jin Yao
2020-03-10 15:08   ` Arnaldo Carvalho de Melo
2020-03-11  5:33     ` Jin, Yao
2020-03-10  7:02 ` [PATCH v1 02/14] perf util: Create streams for managing top N hottest callchains Jin Yao
2020-03-10 15:11   ` Arnaldo Carvalho de Melo [this message]
2020-03-11  5:38     ` Jin, Yao
2020-03-10  7:02 ` [PATCH v1 03/14] perf util: Return per-event callchain streams Jin Yao
2020-03-10  7:02 ` [PATCH v1 04/14] perf util: Compare two streams Jin Yao
2020-03-10  7:02 ` [PATCH v1 05/14] perf util: Calculate the sum of all streams hits Jin Yao
2020-03-10 15:14   ` Arnaldo Carvalho de Melo
2020-03-11  5:44     ` Jin, Yao
2020-03-10  7:02 ` [PATCH v1 06/14] perf util: Report hot streams Jin Yao
2020-03-10  7:02 ` [PATCH v1 07/14] perf diff: Support hot streams comparison Jin Yao
2020-03-10  7:02 ` [PATCH v1 08/14] perf util: Add new block info functions for top N hot blocks comparison Jin Yao
2020-03-10 15:17   ` Arnaldo Carvalho de Melo
2020-03-11  5:47     ` Jin, Yao
2020-03-10  7:02 ` [PATCH v1 09/14] perf util: Add new block info fmts for showing " Jin Yao
2020-03-10  7:02 ` [PATCH v1 10/14] perf util: Enable block source line comparison Jin Yao
2020-03-10  7:02 ` [PATCH v1 11/14] perf diff: support hot blocks comparison Jin Yao
2020-03-10  7:02 ` [PATCH v1 12/14] perf util: Filter out streams by name of changed functions Jin Yao
2020-03-10  7:02 ` [PATCH v1 13/14] perf util: Filter out blocks " Jin Yao
2020-03-10  7:02 ` [PATCH v1 14/14] perf diff: Filter out streams by " Jin Yao

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=20200310151117.GG15931@kernel.org \
    --to=arnaldo.melo@gmail.com \
    --cc=Linux-kernel@vger.kernel.org \
    --cc=ak@linux.intel.com \
    --cc=alexander.shishkin@linux.intel.com \
    --cc=jolsa@kernel.org \
    --cc=kan.liang@intel.com \
    --cc=mingo@redhat.com \
    --cc=peterz@infradead.org \
    --cc=yao.jin@intel.com \
    --cc=yao.jin@linux.intel.com \
    --subject='Re: [PATCH v1 02/14] perf util: Create streams for managing top N hottest callchains' \
    /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

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).