Patches contributed by Eötvös Lorand University
commit dcd9936a5a6d89512b5323c1145647f2dbe0236f
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 04:36:37 2011 +0200
perf stat: Factor our shadow stats
Create update_shadow_stats() which is then used in both read_counter_aggr()
and read_counter().
This not only simplifies the code but also fixes a bug: HW_CACHE_REFERENCES
was not updated in read_counter().
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-9uc55z3g88r47exde7zxjm6p@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 0de3a2002f49..e5e82f62c784 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -193,6 +193,23 @@ static inline int nsec_counter(struct perf_evsel *evsel)
return 0;
}
+/*
+ * Update various tracking values we maintain to print
+ * more semantic information such as miss/hit ratios,
+ * instruction rates, etc:
+ */
+static void update_shadow_stats(struct perf_evsel *counter, u64 *count)
+{
+ if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
+ update_stats(&runtime_nsecs_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
+ update_stats(&runtime_cycles_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
+ update_stats(&runtime_branches_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
+ update_stats(&runtime_cacherefs_stats[0], count[0]);
+}
+
/*
* Read out the results of a single counter:
* aggregate counts across CPUs in system-wide mode
@@ -218,14 +235,7 @@ static int read_counter_aggr(struct perf_evsel *counter)
/*
* Save the full runtime - to allow normalization during printout:
*/
- if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
- update_stats(&runtime_nsecs_stats[0], count[0]);
- else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
- update_stats(&runtime_cycles_stats[0], count[0]);
- else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
- update_stats(&runtime_branches_stats[0], count[0]);
- else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
- update_stats(&runtime_cacherefs_stats[0], count[0]);
+ update_shadow_stats(counter, count);
return 0;
}
@@ -245,12 +255,7 @@ static int read_counter(struct perf_evsel *counter)
count = counter->counts->cpu[cpu].values;
- if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
- update_stats(&runtime_nsecs_stats[cpu], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
- update_stats(&runtime_cycles_stats[cpu], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
- update_stats(&runtime_branches_stats[cpu], count[0]);
+ update_shadow_stats(counter, count);
}
return 0;
commit 749141d926faf23ef811686a8050e7cf13dc223f
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 04:24:57 2011 +0200
perf stat: Make all displayed event names parseable as well
Right now we display this by default:
0.202204 task-clock-msecs # 0.282 CPUs
0 context-switches # 0.000 M/sec
0 CPU-migrations # 0.000 M/sec
85 page-faults # 0.420 M/sec
The task-clock-msecs event cannot actually be passed back as an
event name, the event name we recognize is 'task-clock'.
So change the output of the cpu-clock and task-clock events
to be idempotent.
( Units should be printed out in the right-side column, if needed. )
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-lexrnbzy09asscgd4f7oac4i@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index b686269427ea..b5bfef12f399 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -70,8 +70,8 @@ static const char *hw_event_names[] = {
};
static const char *sw_event_names[] = {
- "cpu-clock-msecs",
- "task-clock-msecs",
+ "cpu-clock",
+ "task-clock",
"page-faults",
"context-switches",
"CPU-migrations",
commit ceb53fbf6dbb1df26d38379a262c6981fe73dd36
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 04:06:33 2011 +0200
perf stat: Fail more clearly when an invalid modifier is specified
Currently we fail without printing any error message on "perf stat -e task-clock-msecs".
The reason is that the task-clock event is matched and the "-msecs" postfix is assumed
to be an event modifier - but is not recognized.
This patch changes the code to be more informative:
$ perf stat -e task-clock-msecs true
invalid event modifier: '-msecs'
Run 'perf list' for a list of valid events and modifiers
And restructures the return value of parse_event_modifier() to allow
the printing of all variants of invalid event modifiers.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-wlaw3dvz1ly6wple8l52cfca@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 8e54bdb553c3..b686269427ea 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -721,15 +721,19 @@ parse_numeric_event(const char **strp, struct perf_event_attr *attr)
return EVT_FAILED;
}
-static enum event_result
+static int
parse_event_modifier(const char **strp, struct perf_event_attr *attr)
{
const char *str = *strp;
int exclude = 0;
int eu = 0, ek = 0, eh = 0, precise = 0;
- if (*str++ != ':')
+ if (!*str)
return 0;
+
+ if (*str++ != ':')
+ return -1;
+
while (*str) {
if (*str == 'u') {
if (!exclude)
@@ -750,14 +754,16 @@ parse_event_modifier(const char **strp, struct perf_event_attr *attr)
++str;
}
- if (str >= *strp + 2) {
- *strp = str;
- attr->exclude_user = eu;
- attr->exclude_kernel = ek;
- attr->exclude_hv = eh;
- attr->precise_ip = precise;
- return 1;
- }
+ if (str < *strp + 2)
+ return -1;
+
+ *strp = str;
+
+ attr->exclude_user = eu;
+ attr->exclude_kernel = ek;
+ attr->exclude_hv = eh;
+ attr->precise_ip = precise;
+
return 0;
}
@@ -800,7 +806,12 @@ parse_event_symbols(const struct option *opt, const char **str,
return EVT_FAILED;
modifier:
- parse_event_modifier(str, attr);
+ if (parse_event_modifier(str, attr) < 0) {
+ fprintf(stderr, "invalid event modifier: '%s'\n", *str);
+ fprintf(stderr, "Run 'perf list' for a list of valid events and modifiers\n");
+
+ return EVT_FAILED;
+ }
return ret;
}
commit b908debd4eef91471016138569f7a9e292be682e
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 03:55:40 2011 +0200
perf tools: Accept case-insensitive symbolic event variants
We currently fail on something like '-e CPU-migrations', with:
invalid or unsupported event: 'CPU-migrations'
While 'CPU-migrations' is how we actually print out the event
in the default perf stat output:
Performance counter stats for 'true':
0.202204 task-clock-msecs # 0.282 CPUs
0 context-switches # 0.000 M/sec
0 CPU-migrations # 0.000 M/sec
So change the matching to be case-insensitive.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-omcm3edjjtx83a4kh2e244se@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 1869e4c646db..8e54bdb553c3 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -649,13 +649,15 @@ static int check_events(const char *str, unsigned int i)
int n;
n = strlen(event_symbols[i].symbol);
- if (!strncmp(str, event_symbols[i].symbol, n))
+ if (!strncasecmp(str, event_symbols[i].symbol, n))
return n;
n = strlen(event_symbols[i].alias);
- if (n)
- if (!strncmp(str, event_symbols[i].alias, n))
+ if (n) {
+ if (!strncasecmp(str, event_symbols[i].alias, n))
return n;
+ }
+
return 0;
}
commit d58f4c82fed69fdd4a79fa54fe17fd14d98c27aa
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 03:42:18 2011 +0200
perf stat: Print cache misses as percentage
Before:
113,393,041 cache-references # 83.636 M/sec
7,052,454 cache-misses # 5.202 M/sec
After:
112,589,441 cache-references # 87.925 M/sec
6,556,354 cache-misses # 5.823 %
misses/hits percentages are more expressive than absolute numbers
or rates.
(Also prettify the CPUs printout line to not have a trailing whitespace.)
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-axm28f43x439bl41zkvfzd63@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 3a29041f8575..0de3a2002f49 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -157,6 +157,7 @@ static double stddev_stats(struct stats *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 runtime_cacherefs_stats[MAX_NR_CPUS];
struct stats walltime_nsecs_stats;
static int create_perf_stat_counter(struct perf_evsel *evsel)
@@ -219,10 +220,12 @@ static int read_counter_aggr(struct perf_evsel *counter)
*/
if (perf_evsel__match(counter, SOFTWARE, SW_TASK_CLOCK))
update_stats(&runtime_nsecs_stats[0], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
+ else if (perf_evsel__match(counter, HARDWARE, HW_CPU_CYCLES))
update_stats(&runtime_cycles_stats[0], count[0]);
- if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
+ else if (perf_evsel__match(counter, HARDWARE, HW_BRANCH_INSTRUCTIONS))
update_stats(&runtime_branches_stats[0], count[0]);
+ else if (perf_evsel__match(counter, HARDWARE, HW_CACHE_REFERENCES))
+ update_stats(&runtime_cacherefs_stats[0], count[0]);
return 0;
}
@@ -404,7 +407,7 @@ static void nsec_printout(int cpu, struct perf_evsel *evsel, double avg)
return;
if (perf_evsel__match(evsel, SOFTWARE, SW_TASK_CLOCK))
- fprintf(stderr, " # %10.3f CPUs ",
+ fprintf(stderr, " # %10.3f CPUs",
avg / avg_stats(&walltime_nsecs_stats));
}
@@ -452,6 +455,15 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
fprintf(stderr, " # %10.3f %%", ratio);
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_CACHE_MISSES) &&
+ runtime_cacherefs_stats[cpu].n != 0) {
+ total = avg_stats(&runtime_cacherefs_stats[cpu]);
+
+ if (total)
+ ratio = avg * 100 / total;
+
+ fprintf(stderr, " # %10.3f %%", ratio);
+
} else if (runtime_nsecs_stats[cpu].n != 0) {
total = avg_stats(&runtime_nsecs_stats[cpu]);
commit 11ba2b85f506306c8dfc9fe144aa4ddc43242855
Author: Ingo Molnar <mingo@elte.hu>
Date: Sun Apr 24 15:05:10 2011 +0200
perf stat: Print stalled cycles percentage
Print:
611,527 cycles
400,553 instructions # ( 0.71 instructions per cycle )
77,809 stalled-cycles # ( 12.71% of all cycles )
0.000610987 seconds time elapsed
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Link: http://lkml.kernel.org/n/tip-fd6x8r1cpyb6zhlrc4ix8m45@git.kernel.org
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 03f0e45f1479..3a29041f8575 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -442,7 +442,7 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
if (total)
ratio = avg / total;
- fprintf(stderr, " # %10.3f IPC ", ratio);
+ fprintf(stderr, " # ( %4.2f instructions per cycle )", ratio);
} else if (perf_evsel__match(evsel, HARDWARE, HW_BRANCH_MISSES) &&
runtime_branches_stats[cpu].n != 0) {
total = avg_stats(&runtime_branches_stats[cpu]);
@@ -450,7 +450,7 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
if (total)
ratio = avg * 100 / total;
- fprintf(stderr, " # %10.3f %% ", ratio);
+ fprintf(stderr, " # %10.3f %%", ratio);
} else if (runtime_nsecs_stats[cpu].n != 0) {
total = avg_stats(&runtime_nsecs_stats[cpu]);
@@ -459,6 +459,13 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
ratio = 1000.0 * avg / total;
fprintf(stderr, " # %10.3f M/sec", ratio);
+ } else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES)) {
+ total = avg_stats(&runtime_cycles_stats[cpu]);
+
+ if (total)
+ ratio = avg / total * 100.0;
+
+ fprintf(stderr, " # (%5.2f%% of all cycles )", ratio);
}
}
commit 5c543e3c442d6382db127152c7096ca6a55283de
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 12:02:04 2011 +0200
perf events, x86: Mark constrant tables read mostly
Various constraint tables were not marked read-mostly.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-wpqwwvmhxucy5e718wnamjiv@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index efa2704c9dfd..067a48b13a76 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -36,7 +36,7 @@ static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly =
[PERF_COUNT_HW_BUS_CYCLES] = 0x013c,
};
-static struct event_constraint intel_core_event_constraints[] =
+static struct event_constraint intel_core_event_constraints[] __read_mostly =
{
INTEL_EVENT_CONSTRAINT(0x11, 0x2), /* FP_ASSIST */
INTEL_EVENT_CONSTRAINT(0x12, 0x2), /* MUL */
@@ -47,7 +47,7 @@ static struct event_constraint intel_core_event_constraints[] =
EVENT_CONSTRAINT_END
};
-static struct event_constraint intel_core2_event_constraints[] =
+static struct event_constraint intel_core2_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
@@ -70,7 +70,7 @@ static struct event_constraint intel_core2_event_constraints[] =
EVENT_CONSTRAINT_END
};
-static struct event_constraint intel_nehalem_event_constraints[] =
+static struct event_constraint intel_nehalem_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
@@ -86,19 +86,19 @@ static struct event_constraint intel_nehalem_event_constraints[] =
EVENT_CONSTRAINT_END
};
-static struct extra_reg intel_nehalem_extra_regs[] =
+static struct extra_reg intel_nehalem_extra_regs[] __read_mostly =
{
INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0xffff),
EVENT_EXTRA_END
};
-static struct event_constraint intel_nehalem_percore_constraints[] =
+static struct event_constraint intel_nehalem_percore_constraints[] __read_mostly =
{
INTEL_EVENT_CONSTRAINT(0xb7, 0),
EVENT_CONSTRAINT_END
};
-static struct event_constraint intel_westmere_event_constraints[] =
+static struct event_constraint intel_westmere_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
@@ -110,7 +110,7 @@ static struct event_constraint intel_westmere_event_constraints[] =
EVENT_CONSTRAINT_END
};
-static struct event_constraint intel_snb_event_constraints[] =
+static struct event_constraint intel_snb_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
@@ -123,21 +123,21 @@ static struct event_constraint intel_snb_event_constraints[] =
EVENT_CONSTRAINT_END
};
-static struct extra_reg intel_westmere_extra_regs[] =
+static struct extra_reg intel_westmere_extra_regs[] __read_mostly =
{
INTEL_EVENT_EXTRA_REG(0xb7, MSR_OFFCORE_RSP_0, 0xffff),
INTEL_EVENT_EXTRA_REG(0xbb, MSR_OFFCORE_RSP_1, 0xffff),
EVENT_EXTRA_END
};
-static struct event_constraint intel_westmere_percore_constraints[] =
+static struct event_constraint intel_westmere_percore_constraints[] __read_mostly =
{
INTEL_EVENT_CONSTRAINT(0xb7, 0),
INTEL_EVENT_CONSTRAINT(0xbb, 0),
EVENT_CONSTRAINT_END
};
-static struct event_constraint intel_gen_event_constraints[] =
+static struct event_constraint intel_gen_event_constraints[] __read_mostly =
{
FIXED_EVENT_CONSTRAINT(0x00c0, 0), /* INST_RETIRED.ANY */
FIXED_EVENT_CONSTRAINT(0x003c, 1), /* CPU_CLK_UNHALTED.CORE */
commit 94403f8863d0d1d2005291b2ef0719c2534aa303
Author: Ingo Molnar <mingo@elte.hu>
Date: Sun Apr 24 08:18:31 2011 +0200
perf events: Add stalled cycles generic event - PERF_COUNT_HW_STALLED_CYCLES
The new PERF_COUNT_HW_STALLED_CYCLES event tries to approximate
cycles the CPU does nothing useful, because it is stalled on a
cache-miss or some other condition.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-fue11vymwqsoo5to72jxxjyl@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 9ae4a2aa7398..efa2704c9dfd 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -1413,6 +1413,9 @@ static __init int intel_pmu_init(void)
x86_pmu.enable_all = intel_pmu_nhm_enable_all;
x86_pmu.extra_regs = intel_nehalem_extra_regs;
+ /* Install the stalled-cycles event: 0xff: All reasons, 0xa2: Resource stalls */
+ intel_perfmon_event_map[PERF_COUNT_HW_STALLED_CYCLES] = 0xffa2;
+
if (ebx & 0x40) {
/*
* Erratum AAJ80 detected, we work it around by using
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index ee9f1e782800..ac636dd20a0c 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -52,6 +52,7 @@ enum perf_hw_id {
PERF_COUNT_HW_BRANCH_INSTRUCTIONS = 4,
PERF_COUNT_HW_BRANCH_MISSES = 5,
PERF_COUNT_HW_BUS_CYCLES = 6,
+ PERF_COUNT_HW_STALLED_CYCLES = 7,
PERF_COUNT_HW_MAX, /* non-ABI */
};
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 952b4ae3d954..1869e4c646db 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -38,6 +38,7 @@ static struct event_symbol event_symbols[] = {
{ CHW(BRANCH_INSTRUCTIONS), "branch-instructions", "branches" },
{ CHW(BRANCH_MISSES), "branch-misses", "" },
{ CHW(BUS_CYCLES), "bus-cycles", "" },
+ { CHW(STALLED_CYCLES), "stalled-cycles", "" },
{ CSW(CPU_CLOCK), "cpu-clock", "" },
{ CSW(TASK_CLOCK), "task-clock", "" },
diff --git a/tools/perf/util/python.c b/tools/perf/util/python.c
index f5e38451fdc5..406f613ee619 100644
--- a/tools/perf/util/python.c
+++ b/tools/perf/util/python.c
@@ -798,6 +798,7 @@ static struct {
{ "COUNT_HW_BRANCH_INSTRUCTIONS", PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
{ "COUNT_HW_BRANCH_MISSES", PERF_COUNT_HW_BRANCH_MISSES },
{ "COUNT_HW_BUS_CYCLES", PERF_COUNT_HW_BUS_CYCLES },
+ { "COUNT_HW_STALLED_CYCLES", PERF_COUNT_HW_STALLED_CYCLES },
{ "COUNT_HW_CACHE_L1D", PERF_COUNT_HW_CACHE_L1D },
{ "COUNT_HW_CACHE_L1I", PERF_COUNT_HW_CACHE_L1I },
{ "COUNT_HW_CACHE_LL", PERF_COUNT_HW_CACHE_LL },
commit 7bd5fafeb414cf00deee32c82834f8bf1426b9ac
Merge: fa7b69475a6c ec75a71634da
Author: Ingo Molnar <mingo@elte.hu>
Date: Tue Apr 26 19:36:14 2011 +0200
Merge branch 'perf/urgent' into perf/stat
Merge reason: We want to queue up dependent changes.
Signed-off-by: Ingo Molnar <mingo@elte.hu>
commit ec75a71634dabe439db91c1ef51d5099f4493808
Author: Ingo Molnar <mingo@elte.hu>
Date: Wed Apr 27 11:51:41 2011 +0200
perf events, x86: Work around the Nehalem AAJ80 erratum
On Nehalem CPUs the retired branch-misses event can be completely bogus,
when there are no branch-misses occuring. When there are a lot of branch
misses then the count is pretty accurate. Still, this leaves us with an
event that over-counts a lot.
Detect this erratum and work it around by using BR_MISP_EXEC.ANY events.
These will also count speculated branches but still it's a lot more
precise in practice than the architectural event.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Link: http://lkml.kernel.org/n/tip-yyfg0bxo9jsqxd6a0ovfny27@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
diff --git a/arch/x86/kernel/cpu/perf_event_intel.c b/arch/x86/kernel/cpu/perf_event_intel.c
index 9194b0698d63..9ae4a2aa7398 100644
--- a/arch/x86/kernel/cpu/perf_event_intel.c
+++ b/arch/x86/kernel/cpu/perf_event_intel.c
@@ -25,7 +25,7 @@ struct intel_percore {
/*
* Intel PerfMon, used on Core and later.
*/
-static const u64 intel_perfmon_event_map[] =
+static u64 intel_perfmon_event_map[PERF_COUNT_HW_MAX] __read_mostly =
{
[PERF_COUNT_HW_CPU_CYCLES] = 0x003c,
[PERF_COUNT_HW_INSTRUCTIONS] = 0x00c0,
@@ -1308,7 +1308,7 @@ static void intel_clovertown_quirks(void)
* AJ106 could possibly be worked around by not allowing LBR
* usage from PEBS, including the fixup.
* AJ68 could possibly be worked around by always programming
- * a pebs_event_reset[0] value and coping with the lost events.
+ * a pebs_event_reset[0] value and coping with the lost events.
*
* But taken together it might just make sense to not enable PEBS on
* these chips.
@@ -1412,6 +1412,18 @@ static __init int intel_pmu_init(void)
x86_pmu.percore_constraints = intel_nehalem_percore_constraints;
x86_pmu.enable_all = intel_pmu_nhm_enable_all;
x86_pmu.extra_regs = intel_nehalem_extra_regs;
+
+ if (ebx & 0x40) {
+ /*
+ * Erratum AAJ80 detected, we work it around by using
+ * the BR_MISP_EXEC.ANY event. This will over-count
+ * branch-misses, but it's still much better than the
+ * architectural event which is often completely bogus:
+ */
+ intel_perfmon_event_map[PERF_COUNT_HW_BRANCH_MISSES] = 0x7f89;
+
+ pr_cont("erratum AAJ80 worked around, ");
+ }
pr_cont("Nehalem events, ");
break;