Modify the way we account for CPU time spent (step 1)

Keep track of time spent by the cpu in various contexts in units of
"cputicks" and scale to real-world microsec^H^H^H^H^H^H^H^Hclock_t
only when somebody wants to inspect the numbers.

For now "cputicks" are still derived from the current timecounter
and therefore things should by definition remain sensible also on
SMP machines.  (The main reason for this first milestone commit is
to verify that hypothesis.)

On slower machines, the avoided multiplications to normalize timestams
at every context switch, comes out as a 5-7% better score on the
unixbench/context1 microbenchmark.  On more modern hardware no change
in performance is seen.

1 files changed, 21 insertions, 0 deletions

diff --git a/sys/kern/kern_tc.c b/sys/kern/kern_tc.c
index 929e570..397fd2f 100644
--- a/sys/kern/kern_tc.c
+++ b/sys/kern/kern_tc.c
@@ -131,6 +131,7 @@ sysctl_kern_boottime(SYSCTL_HANDLER_ARGS)
 #endif
 		return SYSCTL_OUT(req, &boottime, sizeof(boottime));
 }
+
 /*
  * Return the difference between the timehands' counter value now and what
  * was when we copied it to the timehands' offset_count.
@@ -782,3 +783,23 @@ inittimecounter(void *dummy)
 }
 
 SYSINIT(timecounter, SI_SUB_CLOCKS, SI_ORDER_SECOND, inittimecounter, NULL)
+
+static
+uint64_t
+tc_cpu_ticks(void)
+{
+	static uint64_t base;
+	static unsigned last;
+	uint64_t u;
+	struct timecounter *tc;
+
+	tc = timehands->th_counter;
+	u = tc->tc_get_timecount(tc) & tc->tc_counter_mask;
+	if (u < last)
+		base += tc->tc_counter_mask + 1;
+	last = u;
+	return (u + base);
+}
+
+uint64_t (*cpu_ticks)(void) = tc_cpu_ticks;
+uint64_t (*cpu_tickrate)(void) = tc_getfrequency;