kernel: sched: Align 'tdq_lock' on cache line size

The lock should be aligned on a cache line boundary so it isn't in a
cache line with other data. This is important for 'tdq_lock' as it is
used by every processor which constantly acquires and releases it. This
alignment can reduce how violently cache lines bounce between processor
local caches.

Signed-off-by: Ian Moffett <ian@osmora.org>

1 files changed, 6 insertions, 0 deletions

diff --git a/sys/kern/kern_sched.c b/sys/kern/kern_sched.c
index 0aebf5b..b7e6af5 100644
--- a/sys/kern/kern_sched.c
+++ b/sys/kern/kern_sched.c
@@ -58,7 +58,13 @@ static size_t nthread = 0;
 /*
  * Thread queue lock - all operations to `td_queue'
  * must be done with this lock acquired.
+ *
+ * This lock is aligned on a cache line boundary to ensure
+ * it has its own cacheline to reduce contention. This is
+ * because it is constantly acquired and released on every
+ * processor.
  */
+__cacheline_aligned
 static struct spinlock tdq_lock = {0};
 
 /*