LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org,
	Stephane Eranian <eranian@google.com>,
	Andi Kleen <ak@linux.intel.com>, David Ahern <dsahern@gmail.com>,
	Don Zickus <dzickus@redhat.com>, Ingo Molnar <mingo@elte.hu>,
	Jiri Olsa <jolsa@redhat.com>, Joe Mario <jmario@redhat.com>,
	Namhyung Kim <namhyung@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Richard Fowles <rfowles@redhat.com>,
	Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: [PATCH 01/26] perf mem: Enable sampling loads and stores simultaneously
Date: Mon, 26 Jan 2015 14:47:04 -0300	[thread overview]
Message-ID: <1422294449-12735-2-git-send-email-acme@kernel.org> (raw)
In-Reply-To: <1422294449-12735-1-git-send-email-acme@kernel.org>

From: Stephane Eranian <eranian@google.com>

This patch modifies perf mem to default to sampling loads and stores
simultaneously. It could only do one or the other before yet there was
no hardware restriction preventing simultaneous collection. With this
patch, one run is sufficient to collect both.

It is still possible to sample only loads or stores by using the
-t option:
 $ perf mem -t load rec
 $ perf mem -t load rep
Or
 $ perf mem -t store rec
 $ perf mem -t store rep

The perf report TUI will show one event at a time. The store output will
contain a Weight column which will be empty.

In V2, we updated the man pages to reflect the change and also simplify
the initialization of the argv vector passed to the cmd_*() functions as
per LKML feedback.

In V3, we fixed typos in the changelog.

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Don Zickus <dzickus@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Joe Mario <jmario@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Fowles <rfowles@redhat.com>
Link: http://lkml.kernel.org/r/20141217152355.GA10053@thinkpad
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Documentation/perf-mem.txt |   9 +--
 tools/perf/builtin-mem.c              | 123 +++++++++++++++++++++++++++-------
 2 files changed, 105 insertions(+), 27 deletions(-)

diff --git a/tools/perf/Documentation/perf-mem.txt b/tools/perf/Documentation/perf-mem.txt
index 1d78a4064da4..43310d8661fe 100644
--- a/tools/perf/Documentation/perf-mem.txt
+++ b/tools/perf/Documentation/perf-mem.txt
@@ -12,11 +12,12 @@ SYNOPSIS
 
 DESCRIPTION
 -----------
-"perf mem -t <TYPE> record" runs a command and gathers memory operation data
+"perf mem record" runs a command and gathers memory operation data
 from it, into perf.data. Perf record options are accepted and are passed through.
 
-"perf mem -t <TYPE> report" displays the result. It invokes perf report with the
-right set of options to display a memory access profile.
+"perf mem report" displays the result. It invokes perf report with the
+right set of options to display a memory access profile. By default, loads
+and stores are sampled. Use the -t option to limit to loads or stores.
 
 Note that on Intel systems the memory latency reported is the use-latency,
 not the pure load (or store latency). Use latency includes any pipeline
@@ -29,7 +30,7 @@ OPTIONS
 
 -t::
 --type=::
-	Select the memory operation type: load or store (default: load)
+	Select the memory operation type: load or store (default: load,store)
 
 -D::
 --dump-raw-samples=::
diff --git a/tools/perf/builtin-mem.c b/tools/perf/builtin-mem.c
index 24db6ffe2957..1eded0a3a509 100644
--- a/tools/perf/builtin-mem.c
+++ b/tools/perf/builtin-mem.c
@@ -7,10 +7,13 @@
 #include "util/session.h"
 #include "util/data.h"
 
-#define MEM_OPERATION_LOAD	"load"
-#define MEM_OPERATION_STORE	"store"
+#define MEM_OPERATION_LOAD	0x1
+#define MEM_OPERATION_STORE	0x2
 
-static const char	*mem_operation		= MEM_OPERATION_LOAD;
+/*
+ * default to both load an store sampling
+ */
+static int mem_operation = MEM_OPERATION_LOAD | MEM_OPERATION_STORE;
 
 struct perf_mem {
 	struct perf_tool	tool;
@@ -25,26 +28,30 @@ static int __cmd_record(int argc, const char **argv)
 {
 	int rec_argc, i = 0, j;
 	const char **rec_argv;
-	char event[64];
 	int ret;
 
-	rec_argc = argc + 4;
+	rec_argc = argc + 7; /* max number of arguments */
 	rec_argv = calloc(rec_argc + 1, sizeof(char *));
 	if (!rec_argv)
 		return -1;
 
-	rec_argv[i++] = strdup("record");
-	if (!strcmp(mem_operation, MEM_OPERATION_LOAD))
-		rec_argv[i++] = strdup("-W");
-	rec_argv[i++] = strdup("-d");
-	rec_argv[i++] = strdup("-e");
+	rec_argv[i++] = "record";
 
-	if (strcmp(mem_operation, MEM_OPERATION_LOAD))
-		sprintf(event, "cpu/mem-stores/pp");
-	else
-		sprintf(event, "cpu/mem-loads/pp");
+	if (mem_operation & MEM_OPERATION_LOAD)
+		rec_argv[i++] = "-W";
+
+	rec_argv[i++] = "-d";
+
+	if (mem_operation & MEM_OPERATION_LOAD) {
+		rec_argv[i++] = "-e";
+		rec_argv[i++] = "cpu/mem-loads/pp";
+	}
+
+	if (mem_operation & MEM_OPERATION_STORE) {
+		rec_argv[i++] = "-e";
+		rec_argv[i++] = "cpu/mem-stores/pp";
+	}
 
-	rec_argv[i++] = strdup(event);
 	for (j = 1; j < argc; j++, i++)
 		rec_argv[i] = argv[j];
 
@@ -162,17 +169,17 @@ static int report_events(int argc, const char **argv, struct perf_mem *mem)
 	if (!rep_argv)
 		return -1;
 
-	rep_argv[i++] = strdup("report");
-	rep_argv[i++] = strdup("--mem-mode");
-	rep_argv[i++] = strdup("-n"); /* display number of samples */
+	rep_argv[i++] = "report";
+	rep_argv[i++] = "--mem-mode";
+	rep_argv[i++] = "-n"; /* display number of samples */
 
 	/*
 	 * there is no weight (cost) associated with stores, so don't print
 	 * the column
 	 */
-	if (strcmp(mem_operation, MEM_OPERATION_LOAD))
-		rep_argv[i++] = strdup("--sort=mem,sym,dso,symbol_daddr,"
-				       "dso_daddr,tlb,locked");
+	if (!(mem_operation & MEM_OPERATION_LOAD))
+		rep_argv[i++] = "--sort=mem,sym,dso,symbol_daddr,"
+				"dso_daddr,tlb,locked";
 
 	for (j = 1; j < argc; j++, i++)
 		rep_argv[i] = argv[j];
@@ -182,6 +189,75 @@ static int report_events(int argc, const char **argv, struct perf_mem *mem)
 	return ret;
 }
 
+struct mem_mode {
+	const char *name;
+	int mode;
+};
+
+#define MEM_OPT(n, m) \
+	{ .name = n, .mode = (m) }
+
+#define MEM_END { .name = NULL }
+
+static const struct mem_mode mem_modes[]={
+	MEM_OPT("load", MEM_OPERATION_LOAD),
+	MEM_OPT("store", MEM_OPERATION_STORE),
+	MEM_END
+};
+
+static int
+parse_mem_ops(const struct option *opt, const char *str, int unset)
+{
+	int *mode = (int *)opt->value;
+	const struct mem_mode *m;
+	char *s, *os = NULL, *p;
+	int ret = -1;
+
+	if (unset)
+		return 0;
+
+	/* str may be NULL in case no arg is passed to -t */
+	if (str) {
+		/* because str is read-only */
+		s = os = strdup(str);
+		if (!s)
+			return -1;
+
+		/* reset mode */
+		*mode = 0;
+
+		for (;;) {
+			p = strchr(s, ',');
+			if (p)
+				*p = '\0';
+
+			for (m = mem_modes; m->name; m++) {
+				if (!strcasecmp(s, m->name))
+					break;
+			}
+			if (!m->name) {
+				fprintf(stderr, "unknown sampling op %s,"
+					    " check man page\n", s);
+				goto error;
+			}
+
+			*mode |= m->mode;
+
+			if (!p)
+				break;
+
+			s = p + 1;
+		}
+	}
+	ret = 0;
+
+	if (*mode == 0)
+		*mode = MEM_OPERATION_LOAD;
+error:
+	free(os);
+	return ret;
+}
+
 int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
 {
 	struct stat st;
@@ -199,8 +275,9 @@ int cmd_mem(int argc, const char **argv, const char *prefix __maybe_unused)
 		.input_name		 = "perf.data",
 	};
 	const struct option mem_options[] = {
-	OPT_STRING('t', "type", &mem_operation,
-		   "type", "memory operations(load/store)"),
+	OPT_CALLBACK('t', "type", &mem_operation,
+		   "type", "memory operations(load,store) Default load,store",
+		    parse_mem_ops),
 	OPT_BOOLEAN('D', "dump-raw-samples", &mem.dump_raw,
 		    "dump raw samples in ASCII"),
 	OPT_BOOLEAN('U', "hide-unresolved", &mem.hide_unresolved,
-- 
1.9.3


  reply	other threads:[~2015-01-26 17:47 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-01-26 17:47 [GIT PULL 00/26] perf/core improvements and fixes Arnaldo Carvalho de Melo
2015-01-26 17:47 ` Arnaldo Carvalho de Melo [this message]
2015-01-26 17:47 ` [PATCH 02/26] perf mem: Move the mem_operations global to struct perf_mem Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 03/26] perf tools: Remove EOL whitespaces Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 04/26] perf hists: Rename hist_entry__free to __delete Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 05/26] perf hists: Introduce function for deleting/removing hist_entry Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 06/26] perf report: Get rid of report__inc_stat() Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 07/26] perf tools: Remove some unused functions from color.c Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 08/26] perf tools: Support parsing parameterized events Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 09/26] perf tools: Extend format_alias() to include event parameters Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 10/26] perf Documentation: Add " Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 11/26] perf tools: Document parameterized and symbolic events Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 12/26] perf tools: Allow use of an exclusive option more than once Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 13/26] perf diff: Get rid of hists__compute_resort() Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 14/26] perf diff: Print diff result more precisely Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 15/26] perf diff: Introduce fmt_to_data_file() helper Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 16/26] perf tools: Pass struct perf_hpp_fmt to its callbacks Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 17/26] perf diff: Fix output ordering to honor next column Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 18/26] perf diff: Fix -o/--order option behavior Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 19/26] tools lib fs: Adopt debugfs open strerrno method Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 20/26] tools lib fs: Pass filename to debugfs__strerror_open Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 21/26] perf trace: Fix error reporting for evsel pgfault constructor Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 22/26] tools lib fs debugfs: Introduce debugfs__strerror_open_tp Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 23/26] tools lib fs debugfs: Check if debugfs is mounted when handling ENOENT Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 24/26] perf tests: Fix typo in sample-parsing.c Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 25/26] perf ui/tui: Show fatal error message only if exists Arnaldo Carvalho de Melo
2015-01-26 17:47 ` [PATCH 26/26] tools lib traceevent: Add support for IP address formats Arnaldo Carvalho de Melo
2015-01-28 14:50 ` [GIT PULL 00/26] perf/core improvements and fixes Ingo Molnar
2015-01-28 14:55   ` Arnaldo Carvalho de Melo

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=1422294449-12735-2-git-send-email-acme@kernel.org \
    --to=acme@kernel.org \
    --cc=acme@redhat.com \
    --cc=ak@linux.intel.com \
    --cc=dsahern@gmail.com \
    --cc=dzickus@redhat.com \
    --cc=eranian@google.com \
    --cc=jmario@redhat.com \
    --cc=jolsa@redhat.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=mingo@elte.hu \
    --cc=mingo@kernel.org \
    --cc=namhyung@kernel.org \
    --cc=peterz@infradead.org \
    --cc=rfowles@redhat.com \
    --subject='Re: [PATCH 01/26] perf mem: Enable sampling loads and stores simultaneously' \
    /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).