tracing/filters: Further optimise scalar vs cpumask comparison
authorValentin Schneider <vschneid@redhat.com>
Fri, 7 Jul 2023 17:21:54 +0000 (18:21 +0100)
committerSteven Rostedt (Google) <rostedt@goodmis.org>
Tue, 22 Aug 2023 09:13:29 +0000 (05:13 -0400)
Per the previous commits, we now only enter do_filter_scalar_cpumask() with
a mask of weight greater than one. Optimise the equality checks.

Link: https://lkml.kernel.org/r/20230707172155.70873-9-vschneid@redhat.com
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Juri Lelli <juri.lelli@redhat.com>
Cc: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Leonardo Bras <leobras@redhat.com>
Cc: Frederic Weisbecker <frederic@kernel.org>
Suggested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Valentin Schneider <vschneid@redhat.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
kernel/trace/trace_events_filter.c

index fd72dacc5d1b81c15a9f4aa464f505e533a80e4a..3a529214a21b7cf79c8133e9defffe6bfec330e6 100644 (file)
@@ -667,6 +667,25 @@ do_filter_cpumask(int op, const struct cpumask *mask, const struct cpumask *cmp)
 /* Optimisation of do_filter_cpumask() for scalar fields */
 static inline int
 do_filter_scalar_cpumask(int op, unsigned int cpu, const struct cpumask *mask)
+{
+       /*
+        * Per the weight-of-one cpumask optimisations, the mask passed in this
+        * function has a weight >= 2, so it is never equal to a single scalar.
+        */
+       switch (op) {
+       case OP_EQ:
+               return false;
+       case OP_NE:
+               return true;
+       case OP_BAND:
+               return cpumask_test_cpu(cpu, mask);
+       default:
+               return 0;
+       }
+}
+
+static inline int
+do_filter_cpumask_scalar(int op, const struct cpumask *mask, unsigned int cpu)
 {
        switch (op) {
        case OP_EQ:
@@ -966,12 +985,7 @@ static int filter_pred_cpumask_cpu(struct filter_pred *pred, void *event)
        const struct cpumask *mask = (event + loc);
        unsigned int cpu = pred->val;
 
-       /*
-        * This inverts the usual usage of the function (field is first element,
-        * user parameter is second), but that's fine because the (scalar, mask)
-        * operations used are symmetric.
-        */
-       return do_filter_scalar_cpumask(pred->op, cpu, mask);
+       return do_filter_cpumask_scalar(pred->op, mask, cpu);
 }
 
 /* Filter predicate for COMM. */