Hello
Le mercredi 15 août 2007, Rafael J. Wysocki a écrit :
> Hi,
>
> On Saturday, 11 August 2007 00:32, Couriousous wrote:
> > Hello
> >
> > Here is an easy way to speed up the image writing. It simply tell the
> > kernel to immediatly start writing the data to disk asynchronously and
> > not to do the traditional buffering.
> >
> > Mesurement:
> > System: pentium D820, Harddisk ~58Mo/s, image size 500Mo
> > Time of suspend to disk, average of 3 run.
> >
> > Compression enabled:
> > Without patch: 35 sec.
> > With patch: 27 sec.
> > Speedup: 22%
> >
> > Compression disabled:
> > Without patch: 17 sec.
> > With patch: 14 sec.
> > Speedup: 18%
>
> OK, thanks.
>
> Still, I think we should only sync if we have called write(), no?
>
Yes, you're right, even if a lseek failing at this stage is pretty uncommon.
Regards,
Philippe
> ---
> suspend.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> Index: suspend/suspend.c
> ===================================================================
> --- suspend.orig/suspend.c
> +++ suspend/suspend.c
> @@ -236,8 +236,12 @@ static int write_area(int fd, void *buf,
> ssize_t cnt = 0;
>
> if (offset) {
> - if (lseek64(fd, offset, SEEK_SET) == offset)
> + if (lseek64(fd, offset, SEEK_SET) == offset) {
> cnt = write(fd, buf, size);
> + /* Ask kernel to immediatly start writing the data */
> + sync_file_range(fd, offset, size,
> + SYNC_FILE_RANGE_WRITE);
> + }
> if (cnt == size)
> res = 0;
> else
-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Suspend-devel mailing list
Suspend-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/suspend-devel
