[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: 64-bit Kernel Question



On Mon, 2009-03-09 at 12:54 -0600, Kevin DeKorte wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Petrus de Calguarium wrote:
> > So, there has been a lot of discussion and information about this issue. I 
> > am presently burning the DVD, might buy another 2 GB of RAM, and am about 
> > to commence installing the x86_64 version of F11alpha to a spare partition.
> > 
> > Now, I have another question, having arisen out of this discussion (and not 
> > precluding others that could easily arise, once I have a running system):
> > 
> > As memory requirements for 64-bit are anywhere from 50-100% greater and 
> > the only appreciable difference is a "psychological" performance boost, what 
> > REAL benefit is there, actually?
> > 
> > 
> 
> 64bit processors generally have more registers to work with, so you can
> get 10-20% speed increases when the compiler takes advantage of them,
> and I believe gcc does.

That depends on the architecture family. This is true for x86 vs. x86_64
because the 32-bit x86 family is so register-starved. AMD chose to fix
this when they developed the x86_64 ISA and added all the extra
registers.

Otoh, on PPC/PPC64 this isn't true. The 32-bit CPUs/modes have exactly
the same number of GPRs as the 64-bit versions so this benefit does not
exist there.

That's why Fedora/RHEL on 64-bit ppc has a mostly 32-bit userland:
there's no benefit to using 64-bit processes unless they really do
require the larger address space possible with 64-bit addressing (the
64-bit -libs and -devel packages are included for this reason).

Regards,
Bryn.



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]