Hi,
On large machines (I have found this on 8 and 16 way) performance
suffers because ____cacheline_aligned_in_smp does not prevent other
data from sharing the cacheline. This has been brought up a number of
times in the past.
Also on UP machines we waste memory because we use __cacheline_aligned
all over the place. In all the places I found, it was an SMP
optimisation and there was no need for it on UP.
This patch introduces __cacheline_aligned_in_smp which has the same
behaviour as __cacheline_aligned. I also changed a number of
__cacheline_aligned -> __cacheline_aligned_in_smp.
Anton
diff -ru 2.4.15-pre4/fs/buffer.c 2.4.15-pre4_work/fs/buffer.c
--- 2.4.15-pre4/fs/buffer.c Thu Nov 15 13:38:03 2001
+++ 2.4.15-pre4_work/fs/buffer.c Thu Nov 15 15:55:34 2001
@@ -73,7 +73,7 @@
static rwlock_t hash_table_lock = RW_LOCK_UNLOCKED;
static struct buffer_head *lru_list[NR_LIST];
-static spinlock_t lru_list_lock = SPIN_LOCK_UNLOCKED;
+static spinlock_t lru_list_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
static int nr_buffers_type[NR_LIST];
static unsigned long size_buffers_type[NR_LIST];
diff -ru 2.4.15-pre4/fs/dcache.c 2.4.15-pre4_work/fs/dcache.c
--- 2.4.15-pre4/fs/dcache.c Thu Oct 25 11:29:42 2001
+++ 2.4.15-pre4_work/fs/dcache.c Thu Nov 15 15:50:43 2001
@@ -29,7 +29,7 @@
#define DCACHE_PARANOIA 1
/* #define DCACHE_DEBUG 1 */
-spinlock_t dcache_lock = SPIN_LOCK_UNLOCKED;
+spinlock_t dcache_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
/* Right now the dcache depends on the kernel lock */
#define check_lock() if (!kernel_locked()) BUG()
diff -ru 2.4.15-pre4/include/linux/cache.h 2.4.15-pre4_work/include/linux/cache.h
--- 2.4.15-pre4/include/linux/cache.h Thu Oct 25 11:29:44 2001
+++ 2.4.15-pre4_work/include/linux/cache.h Thu Nov 15 15:45:34 2001
@@ -34,4 +34,12 @@
#endif
#endif /* __cacheline_aligned */
+#ifndef __cacheline_aligned_in_smp
+#ifdef CONFIG_SMP
+#define __cacheline_aligned_in_smp __cacheline_aligned
+#else
+#define __cacheline_aligned_in_smp
+#endif /* CONFIG_SMP */
+#endif
+
#endif /* __LINUX_CACHE_H */
diff -ru 2.4.15-pre4/kernel/fork.c 2.4.15-pre4_work/kernel/fork.c
--- 2.4.15-pre4/kernel/fork.c Thu Oct 25 11:29:44 2001
+++ 2.4.15-pre4_work/kernel/fork.c Thu Nov 15 16:17:30 2001
@@ -206,7 +206,7 @@
return retval;
}
-spinlock_t mmlist_lock __cacheline_aligned = SPIN_LOCK_UNLOCKED;
+spinlock_t mmlist_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
int mmlist_nr;
#define allocate_mm() (kmem_cache_alloc(mm_cachep, SLAB_KERNEL))
diff -ru 2.4.15-pre4/kernel/sched.c 2.4.15-pre4_work/kernel/sched.c
--- 2.4.15-pre4/kernel/sched.c Thu Nov 15 13:38:04 2001
+++ 2.4.15-pre4_work/kernel/sched.c Thu Nov 15 16:16:31 2001
@@ -88,8 +88,8 @@
*
* task->alloc_lock nests inside tasklist_lock.
*/
-spinlock_t runqueue_lock __cacheline_aligned = SPIN_LOCK_UNLOCKED; /* inner */
-rwlock_t tasklist_lock __cacheline_aligned = RW_LOCK_UNLOCKED; /* outer */
+spinlock_t runqueue_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED; /* inner */
+rwlock_t tasklist_lock __cacheline_aligned_in_smp = RW_LOCK_UNLOCKED; /* outer */
static LIST_HEAD(runqueue_head);
diff -ru 2.4.15-pre4/kernel/softirq.c 2.4.15-pre4_work/kernel/softirq.c
--- 2.4.15-pre4/kernel/softirq.c Tue Nov 6 11:13:56 2001
+++ 2.4.15-pre4_work/kernel/softirq.c Thu Nov 15 16:16:49 2001
@@ -42,7 +42,7 @@
irq_cpustat_t irq_stat[NR_CPUS];
-static struct softirq_action softirq_vec[32] __cacheline_aligned;
+static struct softirq_action softirq_vec[32] __cacheline_aligned_in_smp;
/*
* we cannot loop indefinitely here to avoid userspace starvation,
@@ -146,8 +146,8 @@
/* Tasklets */
-struct tasklet_head tasklet_vec[NR_CPUS] __cacheline_aligned;
-struct tasklet_head tasklet_hi_vec[NR_CPUS] __cacheline_aligned;
+struct tasklet_head tasklet_vec[NR_CPUS] __cacheline_aligned_in_smp;
+struct tasklet_head tasklet_hi_vec[NR_CPUS] __cacheline_aligned_in_smp;
void __tasklet_schedule(struct tasklet_struct *t)
{
diff -ru 2.4.15-pre4/mm/filemap.c 2.4.15-pre4_work/mm/filemap.c
--- 2.4.15-pre4/mm/filemap.c Thu Nov 15 13:38:04 2001
+++ 2.4.15-pre4_work/mm/filemap.c Thu Nov 15 15:49:34 2001
@@ -47,7 +47,7 @@
unsigned int page_hash_bits;
struct page **page_hash_table;
-spinlock_t pagecache_lock ____cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
+spinlock_t pagecache_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
/*
* NOTE: to avoid deadlocking you must never acquire the pagemap_lru_lock
* with the pagecache_lock held.
@@ -57,7 +57,7 @@
* pagemap_lru_lock ->
* pagecache_lock
*/
-spinlock_t pagemap_lru_lock ____cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
+spinlock_t pagemap_lru_lock __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
#define CLUSTER_PAGES (1 << page_cluster)
#define CLUSTER_OFFSET(x) (((x) >> page_cluster) << page_cluster)
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Nov 15 2001 - 21:00:41 EST