On Fri, Nov 30, 2007, James Cloos <[EMAIL PROTECTED]> said:

>>>>>> "Aaron" == Aaron Stone <[EMAIL PROTECTED]> writes:
> 
> Aaron> I would be absolutely shocked if someone could find a real-world
> Aaron> performance difference for DBMail running in 32 or 64 mode on the
> Aaron> same chip. We just don't do enough integer calculation to matter
> Aaron> a hoot, I believe.
> 
> Given the use of INT8 (aka bigint) in the pg schema, my earlier reply in
> this thread assumed that dbmail used long long or int64_t to hold those
> values.  If that is true, there should be an advantage to -m64 compiles
> on amd64 systems.  Gcc just generates better code.  And the better ABI
> should also make a difference -- especially for function calls.

Ok, that's a fair point - the 64 bit ABI's are definitely better. But I
still don't believe that that alone will make a major difference. It might
even be that the improved ABI efficiency is a wash with the increased
length of data. Lots of hand-waving going on...

> I'd only expect to notice the difference under load, though.

Benchmarks would be awesome :-)

Aaron



_______________________________________________
DBmail mailing list
DBmail@dbmail.org
https://mailman.fastxs.nl/mailman/listinfo/dbmail

Reply via email to