On Oct 26, 2009, at 5:53 PM, Peter Kasting wrote:

> On Mon, Oct 26, 2009 at 2:48 PM, D. Richard Hipp <d...@hwaci.com>  
> wrote:
>
>> On Oct 26, 2009, at 5:00 PM, Peter Kasting wrote:
>>> as
>>> far as
>>> I can tell the current code shouldn't cause any performance hit by
>>> enabling
>>> this flag.  There's no additional monitoring/tracking it causes; it
>>> just
>>> exposes a few functions to free memory.  Has the memory management
>>> implementation changed since the previous thread, or is this feature
>>> now
>>> broken?
>>
>> We run the "fulltest" test suite on a build of SQLite that include
>> SQLITE_ENABLE_MEMORY_MANAGEMENT prior to every release.  The
>> sqlite3_release_memory() interface works as documented, as far as we
>> know.  Are you seeing something that would suggest otherwise?
>>
>
> I think you're misunderstanding the thrust of my question.  I'm  
> saying that
> based on a glance at the current code, it doesn't look like past  
> comments
> about the perf impact of this flag would be accurate anymore.  I  
> want to
> know if my supposition is true.  Presumably, if it is true, it's  
> either
> because memory management has been rearchitected since that old  
> thread, or
> else because the feature has become broken.  I'm not saying I think  
> it's
> broken, simply that that would be one possible explanation for the
> discrepancy between what it looks like the code does and what it  
> sounds like
> it used to do.
>
> The other question in my email, regarding the reason for the #if 0  
> inside
> sqlite3_release_memory(), is eprhaps more interesting to me.

The memory management has undergone multiple rewrites in the previous  
two years, but as far as we know, sqlite3_release_memory() still works  
as advertised.  Please let us know if you find otherwise.

Though not related to your specific question, additional background  
information on the memory allocation subsystem in SQLite can be found  
at http://www.sqlite.org/malloc.html

The code within the #if 0 was vestigial and has been removed.

D. Richard Hipp
d...@hwaci.com



_______________________________________________
sqlite-users mailing list
sqlite-users@sqlite.org
http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users

Reply via email to