On 11/8/05 9:38 AM, "Stephan Szabo" <[EMAIL PROTECTED]> wrote:
>Cool – can you provide your test case please? I’ll ask our folks to do the same, but as I recall we did some pretty thorough testing and found that it doesn’t help. Moreover, the conclusion was that the current algorithm isn’t designed to use memory effectively.
> Just as we find with a similar comparison (with a "popular commercial,
> proprietary database" :-) Though some might suggest you increase
> work_mem or other tuning suggestions to speed sorting, none work. In
> fact, we find that increasing work_mem actually slows sorting slightly.
I wish you'd qualify your statements, because I can demonstrably show that
I can make sorts go faster on my machine at least by increasing work_mem
under some conditions.
Recognize also that we’re looking for a factor of 10 or more improvement here – this is not a small increase that’s needed.