[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <6bdd805d-6a99-a1d5-f376-3d2d0c915d38@collabora.com>
Date: Mon, 20 Feb 2023 12:15:31 +0500
From: Muhammad Usama Anjum <usama.anjum@...labora.com>
To: Peter Xu <peterx@...hat.com>
Cc: Muhammad Usama Anjum <usama.anjum@...labora.com>,
David Hildenbrand <david@...hat.com>,
linux-kernel@...r.kernel.org, linux-mm@...ck.org,
Axel Rasmussen <axelrasmussen@...gle.com>,
Mike Rapoport <rppt@...ux.vnet.ibm.com>,
Andrew Morton <akpm@...ux-foundation.org>,
Andrea Arcangeli <aarcange@...hat.com>,
Nadav Amit <nadav.amit@...il.com>
Subject: Re: [PATCH] mm/uffd: UFFD_FEATURE_WP_ZEROPAGE
Hi Peter,
Thank you so much for working on this.
On 2/18/23 4:10 AM, Peter Xu wrote:
> Hi, Muhammad,
>
> On Fri, Feb 17, 2023 at 05:31:19PM +0500, Muhammad Usama Anjum wrote:
>> I've just ran my single threaded selftest [1] over an over again to get
>> some numbers.
>>
>> Without zeropage
>> qemu has 6 cores: 26.0355
>
> Did you count in the time of read prefault? Or did you not prefault at
> all?
No, pre-faulting is not being done in both of the runs.
Without zeropage, I'm checking pte_none() to decide if page is dirty.
With zeropage, I'm just checking if WP flag isn't set to decide if page is
dirty.
>
>> With zeropage
>> qemu has 6 cores: 39.203
>>
>> 33% worse performance with zero pages
>>
>> Definitely, there can be better benchmark application. Please let me know
>> if I should write better benchmarks on my end.
>>
>> [1]
>> https://lore.kernel.org/all/20230202112915.867409-7-usama.anjum@collabora.com
>
> I'll have a closer look too next week.
>
> Thanks,
>
--
BR,
Muhammad Usama Anjum
Powered by blists - more mailing lists