Closed GoogleCodeExporter closed 9 years ago
> I guess the pages are sparsely written to the device, directly in it indexed
position, instead of using a mapping. Am I right?
Yes, there is direct one-to-one ident mapping between index and disk block
number. As you guessed much better (but slightly more complicated) would be to
reduce disk seeks, in case of disks, or spread out writes, in case of SSDs, for
better performance. This would require maintaining another mapping in zram to
match index locations with disk block numbers.
> By the way, the backing device seems gone in zram... to make it simpler?
Yes and there were some opposition to using the current approach for backing
swap from folks at LKML. So, currently I'm looking for other implement this
feature -- essentially making xvmalloc memory swappable.
Thanks for your efforts with this patch. This seems to be much requested
feature. I will merge this in 0.5.x branch if there are still enough users of
that version or redirect to this issue.
Original comment by nitingupta910@gmail.com
on 27 Jan 2011 at 3:44
Issue 75 has been merged into this issue.
Original comment by nitingupta910@gmail.com
on 2 Oct 2012 at 10:25
I don't think I will be able to push this feature to mainline anytime soon and
I'm no longer maintaining separate repo on this site. So, closing this issues
for now. Thanks.
Original comment by nitingupta910@gmail.com
on 2 Oct 2012 at 10:47
Original issue reported on code.google.com by
maac...@gmail.com
on 7 Jan 2011 at 8:49Attachments: