/proc/interrupts and reproducibility of FP caculations

[Sanzo Miyazawa <miyazawa@smlab.sci.gunma-u.ac.jp>]

Does anyone undesrstand why the following things occur on my opteron?

1. I am running a 64 bit kernel on opteron, which I compiled
by "gcc-3.3 -m64" from a linux 2.4.21 obtained from x86-64.org.

	smk8m> cat /proc/version
	Linux version 2.4.21-030702-k8-smp-64gb-devfs-raid-ext3 (root@smk8m) (gcc version 3.3 (Debian)) #1 SMP Sat Jul 5 07:53:12 JST 2003
	smk8m> arch
	x86_64

The following is the output of /proc/interrupts.

	smk8m> cat /proc/inter*
	           CPU0       CPU1       
	  0:   85028306          0    IO-APIC-edge  timer
	  1:          5          0    IO-APIC-edge  keyboard
	  2:          0          0          XT-PIC  cascade
	  4:          6          0    IO-APIC-edge  serial
	 12:          6          0    IO-APIC-edge  PS/2 Mouse
	 24:   15249253          0   IO-APIC-level  ioc0, eth0
	 25:         44          0   IO-APIC-level  ioc1
	NMI:     667118     677988 
	LOC:   85025444   85025445 
	ERR:          0
	MIS:          0

Does anyone understand why only cpu0 is used for timer and ..?

2. I tried to run a program which use a function called dsyev in LAPACK on
this opteron; dsyev is a program of calculation eigen values and vectors
of a symmetric matrix.   A strange thing is that eigen vectors differ
every run.  Of course, they are always same on xeon systems.
(LAPACK library is the 32 bit version.)

ii  lapack         3.0.20000531a- a library of linear algebra routines - share
ii  lapack-dev     3.0.20000531a- a library of linear algebra routines - stati