I've folded in the following incremental which causes us to only do the core
calculation once since it looks so expensive. I think this is fine for now, and
I intended to merge it, but I think we should revisit the assembler instruction.
It's probably cleaner, and would support platforms like Windows and BSD which
are becoming more important.
Acked-by: Ethan Jackson <et...@nicira.com>
---
lib/ovs-thread.c | 12 ++++++++----
1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/lib/ovs-thread.c b/lib/ovs-thread.c
index f800d7d..1a633cf 100644
--- a/lib/ovs-thread.c
+++ b/lib/ovs-thread.c
@@ -374,11 +374,15 @@ parse_cpuinfo(long int *n_cores)
unsigned int
count_cpu_cores(void)
{
- long int n_cores;
+ static struct ovsthread_once once = OVSTHREAD_ONCE_INITIALIZER;
+ static long int n_cores;
- parse_cpuinfo(&n_cores);
- if (!n_cores) {
- n_cores = sysconf(_SC_NPROCESSORS_ONLN);
+ if (ovsthread_once_start(&once)) {
+ parse_cpuinfo(&n_cores);
+ if (!n_cores) {
+ n_cores = sysconf(_SC_NPROCESSORS_ONLN);
+ }
+ ovsthread_once_done(&once);
}
return n_cores > 0 ? n_cores : 0;
--
1.8.1.2
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
