[GIT PULL 0/3] perf tools improvements

[GIT PULL 0/3] perf tools improvements

[Arnaldo Carvalho de Melo]

From: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>

Hi Ingo,

        Please pull from:

git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 perf/core

Regards,

- Arnaldo

Arnaldo Carvalho de Melo (1):
  perf tools: Change my maintainer address

Robert Morell (1):
  perf tools: Remove hardcoded include paths for elfutils

Stephane Eranian (1):
  perf stat: Add no-aggregation mode to -a

 CREDITS                                |    2 -
 MAINTAINERS                            |    2 +-
 tools/perf/Documentation/perf-stat.txt |    5 +
 tools/perf/Makefile                    |    4 +-
 tools/perf/builtin-stat.c              |  169 +++++++++++++++++++++++++++-----
 tools/perf/feature-tests.mak           |    4 +-
 tools/perf/util/probe-finder.h         |    6 +-
 7 files changed, 157 insertions(+), 35 deletions(-)

[PATCH 1/4] perf record: Remove newline character from perror() argument

[Arnaldo Carvalho de Melo]

From: Matt Fleming <matt@console-pimps.org>

If we include a newline character in the string argument to perror()
then the output will be split across two lines like so,

    Unable to read perf file descriptor
    : No space left on device

Deleting the newline character prints a much more readable error,

    Unable to read perf file descriptor: No space left on device

Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <89e77b54659bc3798b23a5596c2debb7f6f4cf27.1283010281.git.matt@console-pimps.org>
Signed-off-by: Matt Fleming <matt@console-pimps.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@felicio.ghostprotocols.net>
---
 tools/perf/builtin-record.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index ff77b80..b530bee 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -353,7 +353,7 @@ try_again:
 		}
 
 		if (read(fd[nr_cpu][counter][thread_index], &read_data, sizeof(read_data)) == -1) {
-			perror("Unable to read perf file descriptor\n");
+			perror("Unable to read perf file descriptor");
 			exit(-1);
 		}
 
@@ -626,7 +626,7 @@ static int __cmd_record(int argc, const char **argv)
 
 	nr_cpus = read_cpu_map(cpu_list);
 	if (nr_cpus < 1) {
-		perror("failed to collect number of CPUs\n");
+		perror("failed to collect number of CPUs");
 		return -1;
 	}
 
-- 
1.6.2.5

[PATCH 1/3] perf stat: Add no-aggregation mode to -a

[Arnaldo Carvalho de Melo]

From: Stephane Eranian <eranian@google.com>

This patch adds a new -A option to perf stat. If specified then perf stat does
not aggregate counts across all monitored CPUs in system-wide mode, i.e., when
using -a. This option is not supported in per-thread mode.

Being able to get a per-cpu breakdown is useful to detect imbalances between
CPUs when running a uniform workload than spans all monitored CPUs.

The second version corrects the missing cpumap[] support, so that it works when
the -C option is used.

The third version fixes a missing cpumap[] in print_counter() and removes a
stray patch in builtin-trace.c.

Examples on a 4-way system:

# perf stat -a   -e cycles,instructions -- sleep 1
 Performance counter stats for 'sleep 1':
         9592808135  cycles
         3490380006  instructions             #      0.364 IPC
        1.001584632  seconds time elapsed

# perf stat -a -A -e cycles,instructions -- sleep 1
 Performance counter stats for 'sleep 1':
CPU0            2398163767  cycles
CPU1            2398180817  cycles
CPU2            2398217115  cycles
CPU3            2398247483  cycles
CPU0             872282046  instructions             #      0.364 IPC
CPU1             873481776  instructions             #      0.364 IPC
CPU2             872638127  instructions             #      0.364 IPC
CPU3             872437789  instructions             #      0.364 IPC
        1.001556052  seconds time elapsed

Cc: David S. Miller <davem@davemloft.net>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Robert Richter <robert.richter@amd.com>
LKML-Reference: <4ce257b5.1e07e30a.7b6b.3aa9@mx.google.com>
Signed-off-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Documentation/perf-stat.txt |    5 +
 tools/perf/builtin-stat.c              |  169 +++++++++++++++++++++++++++-----
 2 files changed, 149 insertions(+), 25 deletions(-)

diff --git a/tools/perf/Documentation/perf-stat.txt b/tools/perf/Documentation/perf-stat.txt
index 4b3a2d4..c405bca 100644
--- a/tools/perf/Documentation/perf-stat.txt
+++ b/tools/perf/Documentation/perf-stat.txt
@@ -53,6 +53,11 @@ comma-sperated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2
 In per-thread mode, this option is ignored. The -a option is still necessary
 to activate system-wide monitoring. Default is to count on all CPUs.
 
+-A::
+--no-aggr::
+Do not aggregate counts across all monitored CPUs in system-wide mode (-a).
+This option is only valid in system-wide mode.
+
 EXAMPLES
 --------
 
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index a6b4d44..b3e568f 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -75,6 +75,7 @@ static int			run_idx				=  0;
 static int			run_count			=  1;
 static bool			no_inherit			= false;
 static bool			scale				=  true;
+static bool			no_aggr				= false;
 static pid_t			target_pid			= -1;
 static pid_t			target_tid			= -1;
 static pid_t			*all_tids			=  NULL;
@@ -89,6 +90,12 @@ static int			*fd[MAX_NR_CPUS][MAX_COUNTERS];
 
 static int			event_scaled[MAX_COUNTERS];
 
+static struct {
+	u64 val;
+	u64 ena;
+	u64 run;
+} cpu_counts[MAX_NR_CPUS][MAX_COUNTERS];
+
 static volatile int done = 0;
 
 struct stats
@@ -136,10 +143,10 @@ static double stddev_stats(struct stats *stats)
 }
 
 struct stats			event_res_stats[MAX_COUNTERS][3];
-struct stats			runtime_nsecs_stats;
+struct stats			runtime_nsecs_stats[MAX_NR_CPUS];
+struct stats			runtime_cycles_stats[MAX_NR_CPUS];
+struct stats			runtime_branches_stats[MAX_NR_CPUS];
 struct stats			walltime_nsecs_stats;
-struct stats			runtime_cycles_stats;
-struct stats			runtime_branches_stats;
 
 #define MATCH_EVENT(t, c, counter)			\
 	(attrs[counter].type == PERF_TYPE_##t &&	\
@@ -205,8 +212,9 @@ static inline int nsec_counter(int counter)
 
 /*
  * Read out the results of a single counter:
+ * aggregate counts across CPUs in system-wide mode
  */
-static void read_counter(int counter)
+static void read_counter_aggr(int counter)
 {
 	u64 count[3], single_count[3];
 	int cpu;
@@ -264,11 +272,58 @@ static void read_counter(int counter)
 	 * Save the full runtime - to allow normalization during printout:
 	 */
 	if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter))
-		update_stats(&runtime_nsecs_stats, count[0]);
+		update_stats(&runtime_nsecs_stats[0], count[0]);
 	if (MATCH_EVENT(HARDWARE, HW_CPU_CYCLES, counter))
-		update_stats(&runtime_cycles_stats, count[0]);
+		update_stats(&runtime_cycles_stats[0], count[0]);
 	if (MATCH_EVENT(HARDWARE, HW_BRANCH_INSTRUCTIONS, counter))
-		update_stats(&runtime_branches_stats, count[0]);
+		update_stats(&runtime_branches_stats[0], count[0]);
+}
+
+/*
+ * Read out the results of a single counter:
+ * do not aggregate counts across CPUs in system-wide mode
+ */
+static void read_counter(int counter)
+{
+	u64 count[3];
+	int cpu;
+	size_t res, nv;
+
+	count[0] = count[1] = count[2] = 0;
+
+	nv = scale ? 3 : 1;
+
+	for (cpu = 0; cpu < nr_cpus; cpu++) {
+
+		if (fd[cpu][counter][0] < 0)
+			continue;
+
+		res = read(fd[cpu][counter][0], count, nv * sizeof(u64));
+
+		assert(res == nv * sizeof(u64));
+
+		close(fd[cpu][counter][0]);
+		fd[cpu][counter][0] = -1;
+
+		if (scale) {
+			if (count[2] == 0) {
+				count[0] = 0;
+			} else if (count[2] < count[1]) {
+				count[0] = (unsigned long long)
+				((double)count[0] * count[1] / count[2] + 0.5);
+			}
+		}
+		cpu_counts[cpu][counter].val = count[0]; /* scaled count */
+		cpu_counts[cpu][counter].ena = count[1];
+		cpu_counts[cpu][counter].run = count[2];
+
+		if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter))
+			update_stats(&runtime_nsecs_stats[cpu], count[0]);
+		if (MATCH_EVENT(HARDWARE, HW_CPU_CYCLES, counter))
+			update_stats(&runtime_cycles_stats[cpu], count[0]);
+		if (MATCH_EVENT(HARDWARE, HW_BRANCH_INSTRUCTIONS, counter))
+			update_stats(&runtime_branches_stats[cpu], count[0]);
+	}
 }
 
 static int run_perf_stat(int argc __used, const char **argv)
@@ -362,9 +417,13 @@ static int run_perf_stat(int argc __used, const char **argv)
 
 	update_stats(&walltime_nsecs_stats, t1 - t0);
 
-	for (counter = 0; counter < nr_counters; counter++)
-		read_counter(counter);
-
+	if (no_aggr) {
+		for (counter = 0; counter < nr_counters; counter++)
+			read_counter(counter);
+	} else {
+		for (counter = 0; counter < nr_counters; counter++)
+			read_counter_aggr(counter);
+	}
 	return WEXITSTATUS(status);
 }
 
@@ -377,11 +436,15 @@ static void print_noise(int counter, double avg)
 			100 * stddev_stats(&event_res_stats[counter][0]) / avg);
 }
 
-static void nsec_printout(int counter, double avg)
+static void nsec_printout(int cpu, int counter, double avg)
 {
 	double msecs = avg / 1e6;
 
-	fprintf(stderr, " %18.6f  %-24s", msecs, event_name(counter));
+	if (no_aggr)
+		fprintf(stderr, "CPU%-4d %18.6f  %-24s",
+			cpumap[cpu], msecs, event_name(counter));
+	else
+		fprintf(stderr, " %18.6f  %-24s", msecs, event_name(counter));
 
 	if (MATCH_EVENT(SOFTWARE, SW_TASK_CLOCK, counter)) {
 		fprintf(stderr, " # %10.3f CPUs ",
@@ -389,33 +452,41 @@ static void nsec_printout(int counter, double avg)
 	}
 }
 
-static void abs_printout(int counter, double avg)
+static void abs_printout(int cpu, int counter, double avg)
 {
 	double total, ratio = 0.0;
+	char cpustr[16] = { '\0', };
+
+	if (no_aggr)
+		sprintf(cpustr, "CPU%-4d", cpumap[cpu]);
+	else
+		cpu = 0;
 
 	if (big_num)
-		fprintf(stderr, " %'18.0f  %-24s", avg, event_name(counter));
+		fprintf(stderr, "%s %'18.0f  %-24s",
+			cpustr, avg, event_name(counter));
 	else
-		fprintf(stderr, " %18.0f  %-24s", avg, event_name(counter));
+		fprintf(stderr, "%s %18.0f  %-24s",
+			cpustr, avg, event_name(counter));
 
 	if (MATCH_EVENT(HARDWARE, HW_INSTRUCTIONS, counter)) {
-		total = avg_stats(&runtime_cycles_stats);
+		total = avg_stats(&runtime_cycles_stats[cpu]);
 
 		if (total)
 			ratio = avg / total;
 
 		fprintf(stderr, " # %10.3f IPC  ", ratio);
 	} else if (MATCH_EVENT(HARDWARE, HW_BRANCH_MISSES, counter) &&
-			runtime_branches_stats.n != 0) {
-		total = avg_stats(&runtime_branches_stats);
+			runtime_branches_stats[cpu].n != 0) {
+		total = avg_stats(&runtime_branches_stats[cpu]);
 
 		if (total)
 			ratio = avg * 100 / total;
 
 		fprintf(stderr, " # %10.3f %%    ", ratio);
 
-	} else if (runtime_nsecs_stats.n != 0) {
-		total = avg_stats(&runtime_nsecs_stats);
+	} else if (runtime_nsecs_stats[cpu].n != 0) {
+		total = avg_stats(&runtime_nsecs_stats[cpu]);
 
 		if (total)
 			ratio = 1000.0 * avg / total;
@@ -426,8 +497,9 @@ static void abs_printout(int counter, double avg)
 
 /*
  * Print out the results of a single counter:
+ * aggregated counts in system-wide mode
  */
-static void print_counter(int counter)
+static void print_counter_aggr(int counter)
 {
 	double avg = avg_stats(&event_res_stats[counter][0]);
 	int scaled = event_scaled[counter];
@@ -439,9 +511,9 @@ static void print_counter(int counter)
 	}
 
 	if (nsec_counter(counter))
-		nsec_printout(counter, avg);
+		nsec_printout(-1, counter, avg);
 	else
-		abs_printout(counter, avg);
+		abs_printout(-1, counter, avg);
 
 	print_noise(counter, avg);
 
@@ -458,6 +530,42 @@ static void print_counter(int counter)
 	fprintf(stderr, "\n");
 }
 
+/*
+ * Print out the results of a single counter:
+ * does not use aggregated count in system-wide
+ */
+static void print_counter(int counter)
+{
+	u64 ena, run, val;
+	int cpu;
+
+	for (cpu = 0; cpu < nr_cpus; cpu++) {
+		val = cpu_counts[cpu][counter].val;
+		ena = cpu_counts[cpu][counter].ena;
+		run = cpu_counts[cpu][counter].run;
+		if (run == 0 || ena == 0) {
+			fprintf(stderr, "CPU%-4d %18s  %-24s", cpumap[cpu],
+					"<not counted>", event_name(counter));
+
+			fprintf(stderr, "\n");
+			continue;
+		}
+
+		if (nsec_counter(counter))
+			nsec_printout(cpu, counter, val);
+		else
+			abs_printout(cpu, counter, val);
+
+		print_noise(counter, 1.0);
+
+		if (run != ena) {
+			fprintf(stderr, "  (scaled from %.2f%%)",
+					100.0 * run / ena);
+		}
+		fprintf(stderr, "\n");
+	}
+}
+
 static void print_stat(int argc, const char **argv)
 {
 	int i, counter;
@@ -480,8 +588,13 @@ static void print_stat(int argc, const char **argv)
 		fprintf(stderr, " (%d runs)", run_count);
 	fprintf(stderr, ":\n\n");
 
-	for (counter = 0; counter < nr_counters; counter++)
-		print_counter(counter);
+	if (no_aggr) {
+		for (counter = 0; counter < nr_counters; counter++)
+			print_counter(counter);
+	} else {
+		for (counter = 0; counter < nr_counters; counter++)
+			print_counter_aggr(counter);
+	}
 
 	fprintf(stderr, "\n");
 	fprintf(stderr, " %18.9f  seconds time elapsed",
@@ -545,6 +658,8 @@ static const struct option options[] = {
 		    "print large numbers with thousands\' separators"),
 	OPT_STRING('C', "cpu", &cpu_list, "cpu",
 		    "list of cpus to monitor in system-wide"),
+	OPT_BOOLEAN('A', "no-aggr", &no_aggr,
+		    "disable CPU count aggregation"),
 	OPT_END()
 };
 
@@ -562,6 +677,10 @@ int cmd_stat(int argc, const char **argv, const char *prefix __used)
 	if (run_count <= 0)
 		usage_with_options(stat_usage, options);
 
+	/* no_aggr is for system-wide only */
+	if (no_aggr && !system_wide)
+		usage_with_options(stat_usage, options);
+
 	/* Set attrs and nr_counters if no event is selected and !null_run */
 	if (!null_run && !nr_counters) {
 		memcpy(attrs, default_attrs, sizeof(default_attrs));
-- 
1.6.2.5

[PATCH 2/4] perf scripting: Shut up 'perf record' final status

[Arnaldo Carvalho de Melo]

From: Arnaldo Carvalho de Melo <acme@redhat.com>

We want just the script output, not internal details about the record phase.

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Documentation/perf-record.txt |    4 ++++
 tools/perf/builtin-record.c              |    4 ++++
 tools/perf/builtin-trace.c               |    9 +++++----
 tools/perf/util/debug.c                  |    4 ++--
 tools/perf/util/debug.h                  |    2 +-
 5 files changed, 16 insertions(+), 7 deletions(-)

diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index 3ee27dc..a91f9f9 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -83,6 +83,10 @@ OPTIONS
 --call-graph::
 	Do call-graph (stack chain/backtrace) recording.
 
+-q::
+--quiet::
+	Don't print any message, useful for scripting.
+
 -v::
 --verbose::
 	Be more verbose (show counter open errors, etc).
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index b530bee..4e75583 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -761,6 +761,9 @@ static int __cmd_record(int argc, const char **argv)
 		}
 	}
 
+	if (quiet)
+		return 0;
+
 	fprintf(stderr, "[ perf record: Woken up %ld times to write data ]\n", waking);
 
 	/*
@@ -820,6 +823,7 @@ static const struct option options[] = {
 		    "do call-graph (stack chain/backtrace) recording"),
 	OPT_INCR('v', "verbose", &verbose,
 		    "be more verbose (show counter open errors, etc)"),
+	OPT_BOOLEAN('q', "quiet", &quiet, "don't print any message"),
 	OPT_BOOLEAN('s', "stat", &inherit_stat,
 		    "per thread counts"),
 	OPT_BOOLEAN('d', "data", &sample_address,
diff --git a/tools/perf/builtin-trace.c b/tools/perf/builtin-trace.c
index deda1a9..2f8df45 100644
--- a/tools/perf/builtin-trace.c
+++ b/tools/perf/builtin-trace.c
@@ -625,12 +625,13 @@ int cmd_trace(int argc, const char **argv, const char *prefix __used)
 			dup2(live_pipe[1], 1);
 			close(live_pipe[0]);
 
-			__argv = malloc(5 * sizeof(const char *));
+			__argv = malloc(6 * sizeof(const char *));
 			__argv[0] = "/bin/sh";
 			__argv[1] = record_script_path;
-			__argv[2] = "-o";
-			__argv[3] = "-";
-			__argv[4] = NULL;
+			__argv[2] = "-q";
+			__argv[3] = "-o";
+			__argv[4] = "-";
+			__argv[5] = NULL;
 
 			execvp("/bin/sh", (char **)__argv);
 			exit(-1);
diff --git a/tools/perf/util/debug.c b/tools/perf/util/debug.c
index f9c7e3a..c8d81b0 100644
--- a/tools/perf/util/debug.c
+++ b/tools/perf/util/debug.c
@@ -12,8 +12,8 @@
 #include "debug.h"
 #include "util.h"
 
-int verbose = 0;
-bool dump_trace = false;
+int verbose;
+bool dump_trace = false, quiet = false;
 
 int eprintf(int level, const char *fmt, ...)
 {
diff --git a/tools/perf/util/debug.h b/tools/perf/util/debug.h
index 7a17ee0..7b51408 100644
--- a/tools/perf/util/debug.h
+++ b/tools/perf/util/debug.h
@@ -6,7 +6,7 @@
 #include "event.h"
 
 extern int verbose;
-extern bool dump_trace;
+extern bool quiet, dump_trace;
 
 int dump_printf(const char *fmt, ...) __attribute__((format(printf, 1, 2)));
 void trace_event(event_t *event);
-- 
1.6.2.5

[PATCH 2/3] perf tools: Remove hardcoded include paths for elfutils

[Arnaldo Carvalho de Melo]

From: Robert Morell <rmorell@nvidia.com>

This change removes the use of hardcoded absolute "/usr/include/elfutils" paths
from the perf build.  The problem with hardcoded paths is that it prevents them
from being overridden by $prefix or by -I in CFLAGS (e.g., for cross-compiling
purposes).

Instead, just include the "elfutils/" subdirectory as a relative path when
files are needed from that directory.

Tested by building perf:
- Cross-compiled for ARM on x86_64
- Built natively on x86_64
- Built on x86_64 with /usr/include/elfutils moved to another location
  and manually included in CFLAGS

Acked-by: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>
LKML-Reference: <1289945793-31441-1-git-send-email-rmorell@nvidia.com>
Signed-off-by: Robert Morell <rmorell@nvidia.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/Makefile            |    4 ++--
 tools/perf/feature-tests.mak   |    4 ++--
 tools/perf/util/probe-finder.h |    6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/tools/perf/Makefile b/tools/perf/Makefile
index d1db0f6..74b684d 100644
--- a/tools/perf/Makefile
+++ b/tools/perf/Makefile
@@ -507,7 +507,7 @@ PERFLIBS = $(LIB_FILE)
 -include config.mak
 
 ifndef NO_DWARF
-FLAGS_DWARF=$(ALL_CFLAGS) -I/usr/include/elfutils -ldw -lelf $(ALL_LDFLAGS) $(EXTLIBS)
+FLAGS_DWARF=$(ALL_CFLAGS) -ldw -lelf $(ALL_LDFLAGS) $(EXTLIBS)
 ifneq ($(call try-cc,$(SOURCE_DWARF),$(FLAGS_DWARF)),y)
 	msg := $(warning No libdw.h found or old libdw.h found or elfutils is older than 0.138, disables dwarf support. Please install new elfutils-devel/libdw-dev);
 	NO_DWARF := 1
@@ -554,7 +554,7 @@ ifndef NO_DWARF
 ifeq ($(origin PERF_HAVE_DWARF_REGS), undefined)
 	msg := $(warning DWARF register mappings have not been defined for architecture $(ARCH), DWARF support disabled);
 else
-	BASIC_CFLAGS += -I/usr/include/elfutils -DDWARF_SUPPORT
+	BASIC_CFLAGS += -DDWARF_SUPPORT
 	EXTLIBS += -lelf -ldw
 	LIB_OBJS += $(OUTPUT)util/probe-finder.o
 endif # PERF_HAVE_DWARF_REGS
diff --git a/tools/perf/feature-tests.mak b/tools/perf/feature-tests.mak
index b253db6..b041ca6 100644
--- a/tools/perf/feature-tests.mak
+++ b/tools/perf/feature-tests.mak
@@ -9,8 +9,8 @@ endef
 ifndef NO_DWARF
 define SOURCE_DWARF
 #include <dwarf.h>
-#include <libdw.h>
-#include <version.h>
+#include <elfutils/libdw.h>
+#include <elfutils/version.h>
 #ifndef _ELFUTILS_PREREQ
 #error
 #endif
diff --git a/tools/perf/util/probe-finder.h b/tools/perf/util/probe-finder.h
index bba69d4..beaefc3 100644
--- a/tools/perf/util/probe-finder.h
+++ b/tools/perf/util/probe-finder.h
@@ -34,9 +34,9 @@ extern int find_available_vars_at(int fd, struct perf_probe_event *pev,
 				  bool externs);
 
 #include <dwarf.h>
-#include <libdw.h>
-#include <libdwfl.h>
-#include <version.h>
+#include <elfutils/libdw.h>
+#include <elfutils/libdwfl.h>
+#include <elfutils/version.h>
 
 struct probe_finder {
 	struct perf_probe_event	*pev;		/* Target probe event */
-- 
1.6.2.5

[PATCH 3/4] perf python scripting: Fixup cut'n'paste error in sctop script

[Arnaldo Carvalho de Melo]

From: Arnaldo Carvalho de Melo <acme@redhat.com>

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 tools/perf/scripts/python/sctop.py |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/tools/perf/scripts/python/sctop.py b/tools/perf/scripts/python/sctop.py
index 547cbe9..7a6ec2c 100644
--- a/tools/perf/scripts/python/sctop.py
+++ b/tools/perf/scripts/python/sctop.py
@@ -17,7 +17,7 @@ from perf_trace_context import *
 from Core import *
 from Util import *
 
-usage = "perf trace -s syscall-counts.py [comm] [interval]\n";
+usage = "perf trace -s sctop.py [comm] [interval]\n";
 
 for_comm = None
 default_interval = 3
-- 
1.6.2.5

[PATCH 3/3] perf tools: Change my maintainer address

[Arnaldo Carvalho de Melo]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1175 bytes --]

From: Arnaldo Carvalho de Melo <acme@redhat.com>

Also remove old snail mail address from CREDITS, moved years ago.

LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 CREDITS     |    2 --
 MAINTAINERS |    2 +-
 2 files changed, 1 insertions(+), 3 deletions(-)

diff --git a/CREDITS b/CREDITS
index 41d8e63..494b6e4 100644
--- a/CREDITS
+++ b/CREDITS
@@ -2365,8 +2365,6 @@ E: acme@redhat.com
 W: http://oops.ghostprotocols.net:81/blog/
 P: 1024D/9224DF01 D5DF E3BB E3C8 BCBB F8AD  841A B6AB 4681 9224 DF01
 D: IPX, LLC, DCCP, cyc2x, wl3501_cs, net/ hacks
-S: R. Brasílio Itiberê, 4270/1010 - Água Verde
-S: 80240-060 - Curitiba - Paraná
 S: Brazil
 
 N: Karsten Merker
diff --git a/MAINTAINERS b/MAINTAINERS
index 8e6548d..f94cd37 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -4602,7 +4602,7 @@ PERFORMANCE EVENTS SUBSYSTEM
 M:	Peter Zijlstra <a.p.zijlstra@chello.nl>
 M:	Paul Mackerras <paulus@samba.org>
 M:	Ingo Molnar <mingo@elte.hu>
-M:	Arnaldo Carvalho de Melo <acme@redhat.com>
+M:	Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
 S:	Supported
 F:	kernel/perf_event*.c
 F:	include/linux/perf_event.h
-- 
1.6.2.5

[PATCH 4/4] perf python scripting: Add futex-contention script

[Arnaldo Carvalho de Melo]

From: Arnaldo Carvalho de Melo <acme@redhat.com>

The equivalent to this SystemTAP script:

http://sourceware.org/systemtap/wiki/WSFutexContention

[root@doppio ~]# perf trace futex-contention
Press control+C to stop and show the summary
^Cnpviewer.bin[15242] lock 7f0a8be19104 contended 29 times, 72806 avg ns
npviewer.bin[15242] lock 7f0a8be19130 contended 2 times, 1355 avg ns
synergyc[17245] lock f127f4 contended 1 times, 1830569 avg ns
firefox[15116] lock 7f2b7238af0c contended 168 times, 1230390 avg ns
synergyc[17245] lock f2fc20 contended 1 times, 33149 avg ns
npviewer.bin[15255] lock 7f0a8be19074 contended 155 times, 73047 avg ns
npviewer.bin[15255] lock 7f0a8be190a0 contended 127 times, 7088 avg ns
synergyc[17247] lock f12854 contended 1 times, 46741 avg ns
synergyc[17245] lock f12610 contended 1 times, 7358 avg ns
[root@doppio ~]#

Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
Cc: Tom Zanussi <tzanussi@gmail.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
 .../python/Perf-Trace-Util/lib/Perf/Trace/Util.py  |   18 +++++++
 .../scripts/python/bin/futex-contention-record     |    2 +
 .../scripts/python/bin/futex-contention-report     |    4 ++
 tools/perf/scripts/python/futex-contention.py      |   50 ++++++++++++++++++++
 4 files changed, 74 insertions(+), 0 deletions(-)
 create mode 100644 tools/perf/scripts/python/bin/futex-contention-record
 create mode 100644 tools/perf/scripts/python/bin/futex-contention-report
 create mode 100644 tools/perf/scripts/python/futex-contention.py

diff --git a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py
index 99ff1b7..13cc02b 100644
--- a/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py
+++ b/tools/perf/scripts/python/Perf-Trace-Util/lib/Perf/Trace/Util.py
@@ -8,6 +8,12 @@
 
 import errno, os
 
+FUTEX_WAIT = 0
+FUTEX_WAKE = 1
+FUTEX_PRIVATE_FLAG = 128
+FUTEX_CLOCK_REALTIME = 256
+FUTEX_CMD_MASK = ~(FUTEX_PRIVATE_FLAG | FUTEX_CLOCK_REALTIME)
+
 NSECS_PER_SEC    = 1000000000
 
 def avg(total, n):
@@ -26,6 +32,18 @@ def nsecs_str(nsecs):
     str = "%5u.%09u" % (nsecs_secs(nsecs), nsecs_nsecs(nsecs)),
     return str
 
+def add_stats(dict, key, value):
+	if not dict.has_key(key):
+		dict[key] = (value, value, value, 1)
+	else:
+		min, max, avg, count = dict[key]
+		if value < min:
+			min = value
+		if value > max:
+			max = value
+		avg = (avg + value) / 2
+		dict[key] = (min, max, avg, count + 1)
+
 def clear_term():
     print("\x1b[H\x1b[2J")
 
diff --git a/tools/perf/scripts/python/bin/futex-contention-record b/tools/perf/scripts/python/bin/futex-contention-record
new file mode 100644
index 0000000..5ecbb43
--- /dev/null
+++ b/tools/perf/scripts/python/bin/futex-contention-record
@@ -0,0 +1,2 @@
+#!/bin/bash
+perf record -a -e syscalls:sys_enter_futex -e syscalls:sys_exit_futex $@
diff --git a/tools/perf/scripts/python/bin/futex-contention-report b/tools/perf/scripts/python/bin/futex-contention-report
new file mode 100644
index 0000000..c826813
--- /dev/null
+++ b/tools/perf/scripts/python/bin/futex-contention-report
@@ -0,0 +1,4 @@
+#!/bin/bash
+# description: futext contention measurement
+
+perf trace $@ -s "$PERF_EXEC_PATH"/scripts/python/futex-contention.py
diff --git a/tools/perf/scripts/python/futex-contention.py b/tools/perf/scripts/python/futex-contention.py
new file mode 100644
index 0000000..11e70a3
--- /dev/null
+++ b/tools/perf/scripts/python/futex-contention.py
@@ -0,0 +1,50 @@
+# futex contention
+# (c) 2010, Arnaldo Carvalho de Melo <acme@redhat.com>
+# Licensed under the terms of the GNU GPL License version 2
+#
+# Translation of:
+#
+# http://sourceware.org/systemtap/wiki/WSFutexContention
+#
+# to perf python scripting.
+#
+# Measures futex contention
+
+import os, sys
+sys.path.append(os.environ['PERF_EXEC_PATH'] + '/scripts/python/Perf-Trace-Util/lib/Perf/Trace')
+from Util import *
+
+process_names = {}
+thread_thislock = {}
+thread_blocktime = {}
+
+lock_waits = {} # long-lived stats on (tid,lock) blockage elapsed time
+process_names = {} # long-lived pid-to-execname mapping
+
+def syscalls__sys_enter_futex(event, ctxt, cpu, s, ns, tid, comm,
+			      nr, uaddr, op, val, utime, uaddr2, val3):
+	cmd = op & FUTEX_CMD_MASK
+	if cmd != FUTEX_WAIT:
+		return # we don't care about originators of WAKE events
+
+	process_names[tid] = comm
+	thread_thislock[tid] = uaddr
+	thread_blocktime[tid] = nsecs(s, ns)
+
+def syscalls__sys_exit_futex(event, ctxt, cpu, s, ns, tid, comm,
+			     nr, ret):
+	if thread_blocktime.has_key(tid):
+		elapsed = nsecs(s, ns) - thread_blocktime[tid]
+		add_stats(lock_waits, (tid, thread_thislock[tid]), elapsed)
+		del thread_blocktime[tid]
+		del thread_thislock[tid]
+
+def trace_begin():
+	print "Press control+C to stop and show the summary"
+
+def trace_end():
+	for (tid, lock) in lock_waits:
+		min, max, avg, count = lock_waits[tid, lock]
+		print "%s[%d] lock %x contended %d times, %d avg ns" % \
+		      (process_names[tid], tid, lock, count, avg)
+
-- 
1.6.2.5

Re: [GIT PULL 0/3] perf tools improvements

[Ingo Molnar]

* Arnaldo Carvalho de Melo <acme@infradead.org> wrote:

> From: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
> 
> Hi Ingo,
> 
>         Please pull from:
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 perf/core
> 
> Regards,
> 
> - Arnaldo
> 
> Arnaldo Carvalho de Melo (1):
>   perf tools: Change my maintainer address
> 
> Robert Morell (1):
>   perf tools: Remove hardcoded include paths for elfutils
> 
> Stephane Eranian (1):
>   perf stat: Add no-aggregation mode to -a
> 
>  CREDITS                                |    2 -
>  MAINTAINERS                            |    2 +-
>  tools/perf/Documentation/perf-stat.txt |    5 +
>  tools/perf/Makefile                    |    4 +-
>  tools/perf/builtin-stat.c              |  169 +++++++++++++++++++++++++++-----
>  tools/perf/feature-tests.mak           |    4 +-
>  tools/perf/util/probe-finder.h         |    6 +-
>  7 files changed, 157 insertions(+), 35 deletions(-)

Pulled, thanks a lot Arnaldo!

	Ingo