[GIT PULL -tip][PATCH 0/3 -tip] perf stat patches
From: Jaswinder Singh Rajput
Date: Fri Jun 26 2009 - 17:42:41 EST
Ingo,
These patch set :
fixes perf stat output to handle big names like 'L1-d$-prefetch-misses'
Add support for HARDWARE, SOFTWARE and cache events :
perf stat -e all-sw-events
perf stat -e sw-events
perf stat -e all-hw-events
perf stat -e hw-events
perf stat -e all-cache-events
perf stat -e cache-events
The following changes since commit 18fee47ce1ce511b088ddf65f4e5eb700a9f297a:
Ingo Molnar (1):
Merge branch 'perfcounters/urgent'
are available in the git repository at:
git://git.kernel.org/pub/scm/linux/kernel/git/jaswinder/linux-2.6-tip.git master
Jaswinder Singh Rajput (3):
perf stat: fix stat output
perf_counter tools: Add support to set of multiple events in one shot
perf_counter tools: Add support for all CACHE events
tools/perf/builtin-stat.c | 11 ++--
tools/perf/util/parse-events.c | 129 +++++++++++++++++++++++++++++++++++++++-
2 files changed, 132 insertions(+), 8 deletions(-)
Complete diff :
diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index 8420ec5..9ff9dd5 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -32,6 +32,7 @@
* Wu Fengguang <fengguang.wu@xxxxxxxxx>
* Mike Galbraith <efault@xxxxxx>
* Paul Mackerras <paulus@xxxxxxxxx>
+ * Jaswinder Singh Rajput <jaswinder@xxxxxxxxxx>
*
* Released under the GPL v2. (and only v2, not any later version)
*/
@@ -250,7 +251,7 @@ static void nsec_printout(int counter, u64 *count, u64 *noise)
{
double msecs = (double)count[0] / 1000000;
- fprintf(stderr, " %14.6f %-20s", msecs, event_name(counter));
+ fprintf(stderr, " %14.6f %-24s", msecs, event_name(counter));
if (attrs[counter].type == PERF_TYPE_SOFTWARE &&
attrs[counter].config == PERF_COUNT_SW_TASK_CLOCK) {
@@ -264,7 +265,7 @@ static void nsec_printout(int counter, u64 *count, u64 *noise)
static void abs_printout(int counter, u64 *count, u64 *noise)
{
- fprintf(stderr, " %14Ld %-20s", count[0], event_name(counter));
+ fprintf(stderr, " %14Ld %-24s", count[0], event_name(counter));
if (runtime_cycles_avg &&
attrs[counter].type == PERF_TYPE_HARDWARE &&
@@ -294,7 +295,7 @@ static void print_counter(int counter)
scaled = event_scaled_avg[counter];
if (scaled == -1) {
- fprintf(stderr, " %14s %-20s\n",
+ fprintf(stderr, " %14s %-24s\n",
"<not counted>", event_name(counter));
return;
}
@@ -305,8 +306,7 @@ static void print_counter(int counter)
abs_printout(counter, count, noise);
if (scaled)
- fprintf(stderr, " (scaled from %.2f%%)",
- (double) count[2] / count[1] * 100);
+ fprintf(stderr, " (%7.2fx scaled)", (double)count[1]/count[2]);
fprintf(stderr, "\n");
}
@@ -417,7 +417,6 @@ static void print_stat(int argc, const char **argv)
for (counter = 0; counter < nr_counters; counter++)
print_counter(counter);
-
fprintf(stderr, "\n");
fprintf(stderr, " %14.9f seconds time elapsed.\n",
(double)walltime_nsecs_avg/1e9);
diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 4d042f1..331b296 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -40,6 +40,68 @@ static struct event_symbol event_symbols[] = {
{ CSW(CPU_MIGRATIONS), "cpu-migrations", "migrations" },
};
+struct event_type_symbol {
+ char *symbol;
+ char *alias;
+};
+
+static struct event_type_symbol event_type_symbols[] = {
+ [PERF_TYPE_HARDWARE] = { "hw-events", "all-hw-events", },
+ [PERF_TYPE_SOFTWARE] = { "sw-events", "all-sw-events", },
+ [PERF_TYPE_TRACEPOINT] = { "", "", },
+ [PERF_TYPE_HW_CACHE] = { "cache-events", "all-cache-events", },
+ [PERF_TYPE_RAW] = { "", "", },
+};
+
+struct event_cache_symbol {
+ u8 type;
+ u64 config;
+};
+
+#define CHCACHE(x, y, z) \
+.type = PERF_TYPE_HW_CACHE, \
+.config = (PERF_COUNT_HW_CACHE_##x | (PERF_COUNT_HW_CACHE_OP_##y << 8) |\
+ (PERF_COUNT_HW_CACHE_RESULT_##z << 16))
+
+/*
+ * Generalized Hardware cache counters events
+ * L1I is READ and PREFETCH only
+ * ITLB and BPU is READ only
+ */
+static struct event_cache_symbol event_cache_symbols[] = {
+ { CHCACHE(L1D, READ, ACCESS) },
+ { CHCACHE(L1D, READ, MISS) },
+ { CHCACHE(L1D, WRITE, ACCESS) },
+ { CHCACHE(L1D, WRITE, MISS) },
+ { CHCACHE(L1D, PREFETCH, ACCESS) },
+ { CHCACHE(L1D, PREFETCH, MISS) },
+
+ { CHCACHE(L1I, READ, ACCESS) },
+ { CHCACHE(L1I, READ, MISS) },
+ { CHCACHE(L1D, PREFETCH, ACCESS) },
+ { CHCACHE(L1D, PREFETCH, MISS) },
+
+ { CHCACHE(LL, READ, ACCESS) },
+ { CHCACHE(LL, READ, MISS) },
+ { CHCACHE(LL, WRITE, ACCESS) },
+ { CHCACHE(LL, WRITE, MISS) },
+ { CHCACHE(LL, PREFETCH, ACCESS) },
+ { CHCACHE(LL, PREFETCH, MISS) },
+
+ { CHCACHE(DTLB, READ, ACCESS) },
+ { CHCACHE(DTLB, READ, MISS) },
+ { CHCACHE(DTLB, WRITE, ACCESS) },
+ { CHCACHE(DTLB, WRITE, MISS) },
+ { CHCACHE(DTLB, PREFETCH, ACCESS) },
+ { CHCACHE(DTLB, PREFETCH, MISS) },
+
+ { CHCACHE(ITLB, READ, ACCESS) },
+ { CHCACHE(ITLB, READ, MISS) },
+
+ { CHCACHE(BPU, READ, ACCESS) },
+ { CHCACHE(BPU, READ, MISS) },
+};
+
#define __PERF_COUNTER_FIELD(config, name) \
((config & PERF_COUNTER_##name##_MASK) >> PERF_COUNTER_##name##_SHIFT)
@@ -237,6 +299,60 @@ parse_generic_hw_symbols(const char *str, struct perf_counter_attr *attr)
return 0;
}
+static int set_multiple_events(unsigned int type)
+{
+ struct perf_counter_attr attr;
+ int i;
+
+ switch (type) {
+ case PERF_TYPE_HARDWARE:
+ case PERF_TYPE_SOFTWARE:
+ for (i = 0; i < ARRAY_SIZE(event_symbols); i++) {
+ if (event_symbols[i].type == type) {
+ memset(&attr, 0, sizeof(attr));
+ attr.type = event_symbols[i].type;
+ attr.config = event_symbols[i].config;
+ attrs[nr_counters] = attr;
+ nr_counters++;
+ }
+ }
+ break;
+
+ case PERF_TYPE_HW_CACHE:
+ for (i = 0; i < ARRAY_SIZE(event_cache_symbols); i++) {
+ memset(&attr, 0, sizeof(attr));
+ attr.type = event_cache_symbols[i].type;
+ attr.config = event_cache_symbols[i].config;
+ attrs[nr_counters] = attr;
+ nr_counters++;
+ }
+ break;
+
+ default:
+ return -1;
+ }
+
+ /*
+ * parse_events() is assuming that only single event will be set,
+ * but we are setting multiple events so we need to return magical 1
+ */
+ return 1;
+}
+
+static int check_type_events(const char *str, unsigned int i)
+{
+ if (strlen(event_type_symbols[i].symbol))
+ if (!strncmp(str, event_type_symbols[i].symbol,
+ strlen(event_type_symbols[i].symbol)))
+ return 1;
+
+ if (strlen(event_type_symbols[i].alias))
+ if (!strncmp(str, event_type_symbols[i].alias,
+ strlen(event_type_symbols[i].alias)))
+ return 1;
+ return 0;
+}
+
static int check_events(const char *str, unsigned int i)
{
if (!strncmp(str, event_symbols[i].symbol,
@@ -288,6 +404,12 @@ static int parse_event_symbols(const char *str, struct perf_counter_attr *attr)
return 0;
}
+ for (i = 0; i < ARRAY_SIZE(event_type_symbols); i++) {
+ if (check_type_events(str, i)) {
+ return set_multiple_events(i);
+ }
+ }
+
for (i = 0; i < ARRAY_SIZE(event_symbols); i++) {
if (check_events(str, i)) {
attr->type = event_symbols[i].type;
@@ -314,8 +436,11 @@ again:
if (ret < 0)
return ret;
- attrs[nr_counters] = attr;
- nr_counters++;
+ /* No need to set attrs and increment counter when already set */
+ if (ret == 0) {
+ attrs[nr_counters] = attr;
+ nr_counters++;
+ }
str = strstr(str, ",");
if (str) {
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/