Re: [PATCH]: 1/4 batch mark_page_accessed()

[Andrew Morton]

Marcelo Tosatti <marcelo.tosatti@xxxxxxxxxxxx> wrote:
>
> > I don't think that atomic_inc will be particularly
>  > costly. generic_file_{write,read}() call find_get_page() just before
>  > calling mark_page_accessed(), so cache-line with page reference counter
>  > is most likely still exclusive owned by this CPU. 
> 
>  Assuming that is true - what could cause the slowdown? 

It isn't true.  Atomic ops have a considerable overhead, and this is
unrelated to cache misses.   This is especially true of p4's.

Now, that overhead may be justified.  Needs more study.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/