Performance in 64-bit?

[Greg A. Woods]

At Thu, 14 Dec 2006 15:09:37 -0800,
Andrew Laurence wrote:
> 
> I've seen occasional conversation here regarding running Cyrus in 
> 64-bit mode,

The use of the phrase "64-bit mode" implies restricting the
converstation to those architectures which have the ability to support
different modes of execution, e.g. the UltraSPARC and AMD64 CPUs.  :-)


> but is there an advantage aside from very large quotas? 

Unless your C compiler is in the dark ages there shouldn't be any need
to run on a 64-bit CPU just to be able to use 64-bit integers such as
those used for large quotas.


> Is Cyrus known to run faster in 64-bit, for instance?

Well that depends on what your definitino of "faster" is.

In one environment I support it would be impossible to manage the same
user load on a machine without more than 4GB of RAM and so without
running it on a true 64-bit platform with more than 32-bit direct
addressing (we use an AlphaServer ES40 with 16GB of RAM and four
processors) the speed would be "infinitely" slow, i.e. it wouldn't run
at all.  I.e. there's usually a lot more to a _system_ than just a
single instance of a program running on a given processor.

FYI the Alpha CPU, with the right associated machine architecture, runs
applications quite well -- i.e. with comparable and sometimes better
performance than, say, 32-bit Intel CPUs of the same generation, and to
the best of my knowledge the Alpha only supports 64-bit execution.
Often applications running on an Alpha will require more RAM, but
usually the memory and cache architecture are designed in such a way
that this doesn't adversely affect their performance.  I'd venture to
guess that the same can be said of an AMD64 (in 64-bit mode), as well as
the newer 64-bit SPARC machines (again, in 64-bit mode).

-- 
						Greg A. Woods

H:+1 416 218-0098 W:+1 416 489-5852 x122 VE3TCP RoboHack <woods at robohack.ca>
Planix, Inc. <woods at planix.com>       Secrets of the Weird <woods at weird.com>