perf tools: Add front-end and back-end stalled cycles support
Update perf tooling to deal with front-end and back-end stalled cycles events. Add both the default 'perf stat' output. Cc: 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-7y40wib8n002io7hjpn1dsrm@git.kernel.org Signed-off-by: Ingo Molnar <mingo@elte.hu>
This commit is contained in:
@@ -66,7 +66,8 @@ static struct perf_event_attr default_attrs[] = {
|
|||||||
{ .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
|
{ .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
|
||||||
|
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
|
||||||
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES },
|
||||||
@@ -84,7 +85,8 @@ static struct perf_event_attr detailed_attrs[] = {
|
|||||||
{ .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
|
{ .type = PERF_TYPE_SOFTWARE, .config = PERF_COUNT_SW_PAGE_FAULTS },
|
||||||
|
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
|
||||||
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_INSTRUCTIONS },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
|
||||||
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES },
|
{ .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_BRANCH_MISSES },
|
||||||
@@ -249,7 +251,7 @@ static void update_shadow_stats(struct perf_evsel *counter, u64 *count)
|
|||||||
update_stats(&runtime_nsecs_stats[0], count[0]);
|
update_stats(&runtime_nsecs_stats[0], count[0]);
|
||||||
else 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]);
|
update_stats(&runtime_cycles_stats[0], count[0]);
|
||||||
else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES))
|
else if (perf_evsel__match(counter, HARDWARE, HW_STALLED_CYCLES_BACKEND))
|
||||||
update_stats(&runtime_stalled_cycles_stats[0], count[0]);
|
update_stats(&runtime_stalled_cycles_stats[0], count[0]);
|
||||||
else 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]);
|
update_stats(&runtime_branches_stats[0], count[0]);
|
||||||
@@ -607,7 +609,7 @@ static void abs_printout(int cpu, struct perf_evsel *evsel, double avg)
|
|||||||
|
|
||||||
fprintf(stderr, " # %8.3f %% of all cache refs ", ratio);
|
fprintf(stderr, " # %8.3f %% of all cache refs ", ratio);
|
||||||
|
|
||||||
} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES)) {
|
} else if (perf_evsel__match(evsel, HARDWARE, HW_STALLED_CYCLES_BACKEND)) {
|
||||||
print_stalled_cycles(cpu, evsel, avg);
|
print_stalled_cycles(cpu, evsel, avg);
|
||||||
} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
|
} else if (perf_evsel__match(evsel, HARDWARE, HW_CPU_CYCLES)) {
|
||||||
total = avg_stats(&runtime_nsecs_stats[cpu]);
|
total = avg_stats(&runtime_nsecs_stats[cpu]);
|
||||||
|
|||||||
@@ -32,7 +32,8 @@ char debugfs_path[MAXPATHLEN];
|
|||||||
|
|
||||||
static struct event_symbol event_symbols[] = {
|
static struct event_symbol event_symbols[] = {
|
||||||
{ CHW(CPU_CYCLES), "cpu-cycles", "cycles" },
|
{ CHW(CPU_CYCLES), "cpu-cycles", "cycles" },
|
||||||
{ CHW(STALLED_CYCLES), "stalled-cycles", "idle-cycles" },
|
{ CHW(STALLED_CYCLES_FRONTEND), "stalled-cycles-frontend", "idle-cycles-frontend" },
|
||||||
|
{ CHW(STALLED_CYCLES_BACKEND), "stalled-cycles-backend", "idle-cycles-backend" },
|
||||||
{ CHW(INSTRUCTIONS), "instructions", "" },
|
{ CHW(INSTRUCTIONS), "instructions", "" },
|
||||||
{ CHW(CACHE_REFERENCES), "cache-references", "" },
|
{ CHW(CACHE_REFERENCES), "cache-references", "" },
|
||||||
{ CHW(CACHE_MISSES), "cache-misses", "" },
|
{ CHW(CACHE_MISSES), "cache-misses", "" },
|
||||||
|
|||||||
@@ -798,7 +798,6 @@ static struct {
|
|||||||
{ "COUNT_HW_BRANCH_INSTRUCTIONS", PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
|
{ "COUNT_HW_BRANCH_INSTRUCTIONS", PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
|
||||||
{ "COUNT_HW_BRANCH_MISSES", PERF_COUNT_HW_BRANCH_MISSES },
|
{ "COUNT_HW_BRANCH_MISSES", PERF_COUNT_HW_BRANCH_MISSES },
|
||||||
{ "COUNT_HW_BUS_CYCLES", PERF_COUNT_HW_BUS_CYCLES },
|
{ "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_L1D", PERF_COUNT_HW_CACHE_L1D },
|
||||||
{ "COUNT_HW_CACHE_L1I", PERF_COUNT_HW_CACHE_L1I },
|
{ "COUNT_HW_CACHE_L1I", PERF_COUNT_HW_CACHE_L1I },
|
||||||
{ "COUNT_HW_CACHE_LL", PERF_COUNT_HW_CACHE_LL },
|
{ "COUNT_HW_CACHE_LL", PERF_COUNT_HW_CACHE_LL },
|
||||||
@@ -811,6 +810,9 @@ static struct {
|
|||||||
{ "COUNT_HW_CACHE_RESULT_ACCESS", PERF_COUNT_HW_CACHE_RESULT_ACCESS },
|
{ "COUNT_HW_CACHE_RESULT_ACCESS", PERF_COUNT_HW_CACHE_RESULT_ACCESS },
|
||||||
{ "COUNT_HW_CACHE_RESULT_MISS", PERF_COUNT_HW_CACHE_RESULT_MISS },
|
{ "COUNT_HW_CACHE_RESULT_MISS", PERF_COUNT_HW_CACHE_RESULT_MISS },
|
||||||
|
|
||||||
|
{ "COUNT_HW_STALLED_CYCLES_FRONTEND", PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
|
||||||
|
{ "COUNT_HW_STALLED_CYCLES_BACKEND", PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
|
||||||
|
|
||||||
{ "COUNT_SW_CPU_CLOCK", PERF_COUNT_SW_CPU_CLOCK },
|
{ "COUNT_SW_CPU_CLOCK", PERF_COUNT_SW_CPU_CLOCK },
|
||||||
{ "COUNT_SW_TASK_CLOCK", PERF_COUNT_SW_TASK_CLOCK },
|
{ "COUNT_SW_TASK_CLOCK", PERF_COUNT_SW_TASK_CLOCK },
|
||||||
{ "COUNT_SW_PAGE_FAULTS", PERF_COUNT_SW_PAGE_FAULTS },
|
{ "COUNT_SW_PAGE_FAULTS", PERF_COUNT_SW_PAGE_FAULTS },
|
||||||
|
|||||||
Reference in New Issue
Block a user