> Just for the sake of discussion I've attached a performance
> graph for various C++ data structures plus the Unrolled LL.
> The tests where run on a dell vostro 1400 laptop. As you can
> see the graphs show the ULL to be quite efficient for
> insert/delete from the front/back of the list. I
into the list. So this would be very
problematic for page movement.
Ken
--- On Tue, 10/27/09, John Crenshaw <johncrens...@priacta.com> wrote:
> From: John Crenshaw <johncrens...@priacta.com>
> Subject: Re: [sqlite] Idea for improving page cache
> To: "General Dis
cpy are looked at).
John
-Original Message-
From: sqlite-users-boun...@sqlite.org
[mailto:sqlite-users-boun...@sqlite.org] On Behalf Of Nicolas Williams
Sent: Tuesday, October 27, 2009 6:29 PM
To: General Discussion of SQLite Database
Subject: Re: [sqlite] Idea for improving page cache
On Tue, Oct
On Tue, Oct 27, 2009 at 04:28:11PM -0400, John Crenshaw wrote:
> "advantage" kind of depends. ULL is more specialized. You gain some
> benefit, but also lose some as well. For example, consider what is
> involved in doing a sorted insert into an ULL. On the other hand, you
> can get all of the
users-boun...@sqlite.org [mailto:sqlite-users-boun...@sqlite.org]
On Behalf Of Kristoffer Danielsson
Sent: Tuesday, October 27, 2009 3:50 PM
To: sqlite-users@sqlite.org
Subject: Re: [sqlite] Idea for improving page cache
I really like the concept of ULL. Check this one out:
http://blogs.msdn.com/devd
> wrote:
>
>> From: Pavel Ivanov <paiva...@gmail.com>
>> Subject: Re: [sqlite] Idea for improving page cache
>> To: kennethinbox-sql...@yahoo.com, "General Discussion of SQLite Database"
>> <sqlite-users@sqlite.org>
>> Date: Tuesday, October 27
--- On Tue, 10/27/09, Pavel Ivanov <paiva...@gmail.com> wrote:
> From: Pavel Ivanov <paiva...@gmail.com>
> Subject: Re: [sqlite] Idea for improving page cache
> To: kennethinbox-sql...@yahoo.com, "General Discussion of SQLite Database"
> <sqlite-users@sq
; From: johncrens...@priacta.com
> To: sqlite-users@sqlite.org
> Subject: Re: [sqlite] Idea for improving page cache
>
> Supposing that the reduced cache misses are worth it, I think it would be
> better to simply allocate the nodes from a pool. Allocating from a pool
> maximiz
-users-boun...@sqlite.org [mailto:sqlite-users-boun...@sqlite.org]
On Behalf Of Pavel Ivanov
Sent: Tuesday, October 27, 2009 1:38 PM
To: kennethinbox-sql...@yahoo.com; General Discussion of SQLite Database
Subject: Re: [sqlite] Idea for improving page cache
Are you sure that there will be improvement
@live.se>
> wrote:
>
>> From: Kristoffer Danielsson <kristoffer.daniels...@live.se>
>> Subject: Re: [sqlite] Idea for improving page cache
>> To: sqlite-users@sqlite.org
>> Date: Tuesday, October 27, 2009, 1:03 PM
>>
>> In game development you se
--- On Tue, 10/27/09, Kristoffer Danielsson <kristoffer.daniels...@live.se>
wrote:
> From: Kristoffer Danielsson <kristoffer.daniels...@live.se>
> Subject: Re: [sqlite] Idea for improving page cache
> To: sqlite-users@sqlite.org
> Date: Tuesday, October 27, 20
qlite.org
> Subject: Re: [sqlite] Idea for improving page cache
>
> Are you sure that there will be improvement with ULL?
> If you're talking about improving due to CPU internal cache then first
> of all you have to store in the list pointers to pages, not pages
> themselves (you don't
Are you sure that there will be improvement with ULL?
If you're talking about improving due to CPU internal cache then first
of all you have to store in the list pointers to pages, not pages
themselves (you don't want to store several pages in one chunk of
memory, do you?). So you're getting one
Hi All,
I have an idea that could improve the page cache performance.
Instead of using a regular linked list to connect pages that are on the cache
use an "unrolled linked list". On some architectures due to the CPU caching
the ULL is about 40 times faster.
Still this is mostly
14 matches
Mail list logo
