Add two timestamps to measure PSCI cache flush overhead
Testing showed that the time spent in a cluster power down
operation is dominated by cache flushes. Add two more timestamps
in runtime instrumentation to keep track of the time spent
flushing the L1/L2 caches.
Change-Id: I4c5a04e7663543225a85d3c6b271d7b706deffc4
Signed-off-by: dp-arm <dimitris.papastamos@arm.com>
diff --git a/lib/psci/psci_off.c b/lib/psci/psci_off.c
index 1cc6ede..897bf31 100644
--- a/lib/psci/psci_off.c
+++ b/lib/psci/psci_off.c
@@ -107,12 +107,29 @@
psci_stats_update_pwr_down(end_pwrlvl, &state_info);
#endif
+#if ENABLE_RUNTIME_INSTRUMENTATION
+
+ /*
+ * Flush cache line so that even if CPU power down happens
+ * the timestamp update is reflected in memory.
+ */
+ PMF_CAPTURE_TIMESTAMP(rt_instr_svc,
+ RT_INSTR_ENTER_CFLUSH,
+ PMF_CACHE_MAINT);
+#endif
+
/*
* Arch. management. Perform the necessary steps to flush all
* cpu caches.
*/
psci_do_pwrdown_cache_maintenance(psci_find_max_off_lvl(&state_info));
+#if ENABLE_RUNTIME_INSTRUMENTATION
+ PMF_CAPTURE_TIMESTAMP(rt_instr_svc,
+ RT_INSTR_EXIT_CFLUSH,
+ PMF_NO_CACHE_MAINT);
+#endif
+
/*
* Plat. management: Perform platform specific actions to turn this
* cpu off e.g. exit cpu coherency, program the power controller etc.