Skip to contentSkip to conversation

perf, x86: Don't reset the LBR as frequently

tip:perf/pebs tip-bot for Peter Zijlstra 1 participant 1 message Mar 10, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw
Commit-ID:  b83a46e7da4a948cc852ba7805dfb1a392dec861
Gitweb:     http://git.kernel.org/tip/b83a46e7da4a948cc852ba7805dfb1a392dec861
Author:     Peter Zijlstra <[email protected]>
AuthorDate: Mon, 8 Mar 2010 13:51:12 +0100
Committer:  Ingo Molnar <[email protected]>
CommitDate: Wed, 10 Mar 2010 13:23:38 +0100

perf, x86: Don't reset the LBR as frequently

If we reset the LBR on each first counter, simple counter rotation which first deschedules all counters and then reschedules the new ones will lead to LBR reset, even though we're still in the same task context.

Reduce this by not flushing on the first counter but only flushing on different task contexts.

LKML-Reference: <new-submission>

---
 arch/x86/kernel/cpu/perf_event_intel_lbr.c |    9 ++++-----
 1 files changed, 4 insertions(+), 5 deletions(-)
arch/x86/kernel/cpu/perf_event_intel_lbr.c +4 -5
index 0145f99..f278136 100644
@@ -72,12 +72,11 @@ static void intel_pmu_lbr_enable(struct perf_event *event)
 	WARN_ON_ONCE(cpuc->enabled);
 
 	/*
-	 * Reset the LBR stack if this is the first LBR user or
-	 * we changed task context so as to avoid data leaks.
+	 * Reset the LBR stack if we changed task context to
+	 * avoid data leaks.
 	 */
 
-	if (!cpuc->lbr_users ||
-	    (event->ctx->task && cpuc->lbr_context != event->ctx)) {
+	if (event->ctx->task && cpuc->lbr_context != event->ctx) {
 		intel_pmu_lbr_reset();
 		cpuc->lbr_context = event->ctx;
 	}
@@ -93,7 +92,7 @@ static void intel_pmu_lbr_disable(struct perf_event *event)
 		return;
 
 	cpuc->lbr_users--;
-	BUG_ON(cpuc->lbr_users < 0);
+	WARN_ON_ONCE(cpuc->lbr_users < 0);
 
 	if (cpuc->enabled && !cpuc->lbr_users)
 		__intel_pmu_lbr_disable();