Subject: IA64 speed - errata
To: None <firstname.lastname@example.org>
From: Wojciech Puchar <email@example.com>
Date: 01/02/2003 18:42:12
yes i really missed to check it JUST BEFORE running my tests, but did it
few minutes before...
---------- Forwarded message ----------
Date: Thu, 2 Jan 2003 12:13:07 -0500
From: James Chamberlain <firstname.lastname@example.org>
Subject: Re: IA64?
My name is James Chamberlain, and I am the system administrator of the HP T=
Drive Program (http://testdrive.hp.com/). I was trolling the net this
morning and noticed your post to netbsd.users. I thought I might be able t=
provide a little information that might be of use to you.
The system you were using is a single processor Itanium II prototype with 1=
of memory. There are several reasons why the system may not have performed
as well as expected.
First and foremost, it is a shared system. I happened to notice when I log=
onto it a moment ago that one of our regulars has a CPU-bound process using
up 96% of the CPU time on numerical calculations. Since he didn't renice i=
as he should for long-running jobs, it would be running at the same priorit=
as your job, and could thus impact your job's performance.
Second, I'm not sure how Red Hat compiled gzip for the Itanium. It makes a
big difference. GCC is not yet optimized for the Itanium or Itanium II,
while Intel's compilers are. If you'd like to see a particularly good
illustration of this, try compiling something on one of our Alpha Linux box=
with GCC, and then compare the run times of the resulting program with thos=
of the same program compiled with ccc or cxx (the Compaq C & C++ compilers)=
From=20what I've heard, programs compiled with ccc and cxx tend to show a
ten-fold performance improvement on the Alpha over those compiled with gcc.
While I have not yet installed Intel's compilers on the Red Hat Advanced
Server Itanium II system, I hope to do so next week. In the mean time, the=
are installed on the Red Hat 7.2 Itanium I system (spe190), so you may wish
to try them there. If you'd like to forward this message on to the
netbsd.users list, please feel free to do so. If there is any more
information you would like, any further trouble you run into, or any
suggestions you'd like to make, please feel free to let me know and I'll se=
what I can do about them.
HP Test Drive Program
------------ Output from gpg ------------
gpg: Podpis z=B3o=BFony Thu Jan 2 18:13:07 2003 CET za pomoc=B1 DSA,
z u=BFyciem klucza o identyfikatorze 6249187D
gpg: Nie mo=BFna sprawdzi=E6 podpisu: klucz publiczny nie odnaleziony