Sergey Senozhatsky <sergey.senozhatsky.work@xxxxxxxxx> writes:
> On (07/04/18 10:20), Huang, Ying wrote:
>> > On (06/27/18 21:51), Andrew Morton wrote:
>> >> On Fri, 22 Jun 2018 11:51:30 +0800 "Huang, Ying" <ying.huang@xxxxxxxxx> wrote:
>> >>
>> >> > This is the final step of THP (Transparent Huge Page) swap
>> >> > optimization. After the first and second step, the splitting huge
>> >> > page is delayed from almost the first step of swapout to after swapout
>> >> > has been finished. In this step, we avoid splitting THP for swapout
>> >> > and swapout/swapin the THP in one piece.
>> >>
>> >> It's a tremendously good performance improvement. It's also a
>> >> tremendously large patchset :(
>> >
>> > Will zswap gain a THP swap out/in support at some point?
>> >
>> >
>> > mm/zswap.c: static int zswap_frontswap_store(...)
>> > ...
>> > /* THP isn't supported */
>> > if (PageTransHuge(page)) {
>> > ret = -EINVAL;
>> > goto reject;
>> > }
>>
>> That's not on my TODO list. Do you have interest to work on this?
>
> I'd say I'm interested. Can't promise that I'll have enough spare time
> any time soon, tho.
Thanks!
> The numbers you posted do look fantastic indeed, embedded devices
> [which normally use zswap/zram quite heavily] _probably_ should see
> some performance improvement as well once zswap [and may be zram] can
> handle THP.
Yes. I think so too.
Best Regards,
Huang, Ying
