On Tue, 29 Dec 1998 06:39:41 -0500 (EST), "Mark H. Wood"
<mwood@IUPUI.Edu> said:
>> It struck me that reading for example 16kb at once is a lot faster than
>> reading 4kb 4 times. So why not read 4 or 8 pages at every segfault (and
>> when the program ?
> Not stupid at all, though its value might be affected by other details of
> Linux memory management.
I implemented this recently, and the 2.2-pre kernels page in a cluster
of 16 pages for all swap or mmap faults by default (fewer pages on
smaller-memory machines). The performance improvement on sequential
mmap(), swapin and program loading is remarkable.
> Actually your description is so much like VMS pagefault clustering
> that it's scary.
It's not accidental, either --- my last job was working on VMS kernel
internals. :)
> IIRC the default cluster size is a SYSGEN parameter (somewhat like
> /proc/sys/somethingorother)
The default cluster size is /proc/sys/vm/page-cluster.
> Since programs' memory use patterns differ, I think a good deal of
> performance measurement would be required to determine whether clustering
> actually helps enough to be worth the added complexity, and whether making
> the cluster size tunable is worthwhile. If we're reading 4kB/fault now
> then there may not be much to be gained from clustering.
Probably not worth worrying about unless there is an madvise() hint, I'd
expect.
--Stephen
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/