db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Rick Hillegas <rick.hille...@oracle.com>
Subject Re: could those doing cleanup work add some extra comments about performance implications?
Date Fri, 26 Apr 2013 16:02:28 GMT
On 4/26/13 8:46 AM, Mike Matrigali wrote:
> Could you add a comment to future "Cleanup" work you do, about 
> expected performance issues.  Especially with changes that are changing
> pass by reference to array copy.
>
> I have been reviewing so far, but falling behind.  So far the code paths
> have not seemed important to me: ie. run time statistics, error paths,
> ..., so likely not an issue.  We don't have very good performance tests
> to catch individual changes, so best if developer making change lends
> his insight to the change.  Just looking for a line saying that changes
> are unlikely to affect performance because .... (compile time only,
> error path, once per query execution, ...)
>
>
> The kind of change I would worry about is something that changes a 
> reference to a copy, that would be executed for every row looked at,
> or every row returned or even worse every column looked at.  ie. 
> something where the very small performance hit is linearly affected
> by number of rows in a table.
>
> /mikem
>
>
Hi Mike,

This is a good suggestion. My understanding of the cleanup patches so 
far is the following, which I hope that your reviews confirm:

o We are trying to avoid copying/cloning objects on the inner loops of 
the storage and SQL execution layers

o Copying by the compiler is generally ok

Thanks,
-Rick

Mime
View raw message