Re: Minor optimizations in lazy_scan_heap

On Tue, Dec 4, 2012 at 11:02 PM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

> On Mon, Dec 3, 2012 at 1:23 AM, Pavan Deolasee <pavan(dot)deolasee(at)gmail(dot)com>
> >
> > Comments ? Anyone thinks any/all of above is useful ?
>
> I doubt that any of these things make enough difference to be worth
> bothering with,

You're right. These are not big ticket optimisations, still I felt they are
worth doing because tiny bits add up over a time and also because the code
may become little simpler. The benchmarks don't show anything interesting
though. The time taken to scan 100K+ bits is sub-second. So even when I
tried with the attached patch, the numbers did not show any noticeable
difference. It might be worth trying with a table with 1M or 10M data
blocks, but I don't have such a hardware to test.

The patch itself can be improved further, especially we can possibly
optimise the loop and test 32-bits at a time, instead of 8 I am doing
currently. Not sure its worth though.

Thanks,
Pavan

--
Pavan Deolasee
http://www.linkedin.com/in/pavandeolasee