Re: Intel Atom Processor
On 2010-02-17 22:23:17 -0600, Stan Hoeppner wrote:
> Vincent Lefevre put forth on 2/17/2010 6:21 AM:
> > On 2010-02-16 09:52:06 -0600, Stan Hoeppner wrote:
> >> As a bonus, due to various architectural reasons I won't delve into,
> >> 32bit binaries will usually run slightly faster than the 64 bit
> >> cousins, and they'll take up a little bit less disk space.
> >
> > No, this depends on the application (and "usually" doesn't mean very
> > much because applications will depend on what the machine is used
> > for). And some people would completely disagree with you, e.g.:
>
> Vincent, you're quoting me out of context. Apparently you haven't
> read the entire thread. Or you're misquoting me intentionally. I
> clearly stated that my comments pertained to a certain application
> class on this platform.
AFAIK nothing has been said about such a certain application class.
> > http://www.phoronix.com/scan.php?page=article&item=ubuntu_32_pae&num=1
>
> These benchmarks are irrelevant to the current discussion. They are
> on a capable dual core Intel Core Processor,
I know that the processor is different, but what you said could
be interpreted as a general remark on the amd64 architecture.
Unfortunately you didn't provide any benchmarks yourself. So,
what you said is purely gratuitous.
> and the application mix is not the same as that of an average Atom
> user.
The OP isn't necessarily an "average Atom user".
> > There's another point is favor of amd64: floating-point arithmetic.
> > As SSE is used by default on amd64, FP arithmetic is much cleaner
> > there in practice.
>
> We all know this already (or at least should). However, again, it's
> irrelevant to _this_ thread.
Why?
> This thread, and my comments, deal with Atom based systems with less
> than 2GB of RAM. This thread is not about 32bit vs 64bit binary
> performance in general.
No, this thread is about answering the OP's question, which is:
Which architecture should I use for an Intel Atom Processor?
Also, the FP behavior is *not* binary performance.
> It's about 32bit vs 64bit binary performance on the Atom processor
> and the bulk of applications that users will run on such a platform
> daily, which basically includes only these two apps:
>
> 1. Web brower
> 2. Email client
The OP did *not* say that he would use only these two applications.
Anyway, even a web browser is affected by the x86 FP behavior (at
least under Debian), and any user can be affected by it since users
don't control scripts from remote web sites.
> x86-64 optimizations and performance enhancements will rarely be
> taken advantage of for this class of machine. FP (SSE/SSE2) isn't
> going to make a lick of difference.
Wrong! See
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=309797
for instance (this is for Mozilla, but Iceweasel on Debian/stable
is affected too -- I could test it).
--
Vincent Lefèvre <vincent@vinc17.net> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / Arénaire project (LIP, ENS-Lyon)
Reply to: