Re: [sqlite] poor insert performance with 3.7/Windows

[Shane Harrelson]

I've not seen the performance degradation you've reported here in my
testing.
If you could provide any more details that would help in reproducing this,
it would be appreciated.

And yes, you can safely modify the SQLITE_FCNTL_SIZE_HINT in os_win.c to be
a "no-op" in the same way as os_unix.c.

-Shane

On Fri, Jul 23, 2010 at 6:26 PM, GB  wrote:

>  Hi all,
>
> I had to recognise that our data-transforming (read: heavily inserting)
> procedures had a massive drop in performance with 3.7. Some
> investigation showed that this is related to "pager_write_pagelist"
> where a hint to an estimated target filesize is given to the filesystem,
> which on Windows translates to a winTruncate. This in turn causes a
> low-level-flush of file metadata and performance drops to nil. Is the
> setting of the file size really needed here or can we safely drop it?
> Dropping it seems to work for ourselves,
>
> Regards,
> Gerd
> ___
> sqlite-users mailing list
> sqlite-users@sqlite.org
> http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
>
___
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users

[sqlite] poor insert performance with 3.7/Windows

[GB]

  Hi all,

I had to recognise that our data-transforming (read: heavily inserting) 
procedures had a massive drop in performance with 3.7. Some 
investigation showed that this is related to "pager_write_pagelist" 
where a hint to an estimated target filesize is given to the filesystem, 
which on Windows translates to a winTruncate. This in turn causes a 
low-level-flush of file metadata and performance drops to nil. Is the 
setting of the file size really needed here or can we safely drop it? 
Dropping it seems to work for ourselves,

Regards,
Gerd
___
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users