> Ingo> files are used. With one big file (or a few big files), the i_shared_lock
> Ingo> will always bounce between CPUs wildly in read() workloads, degrading
>
> Will there be difference between bounces of a rwlock in the radix tree
> variant and the cache misses in hashed locks variant for the case of
> concurrently accessed large file ?
definitely, because in the case of page buckets there are many locks
hashed in a mapping-neutral way. Ie. different parts of the same file will
likely map to different spinlocks. In the radix tree case all pages in the
inode will map to the same spinlock.
Ingo
-
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/