cd compilescripts
setenv TYPE_SYSTEM .small.profile
gmake linuxall
------------------------------------------------------------------------------
On desktop:
	route add 192.168.2.0 192.168.1.4
------------------------------------------------------------------------------
On portable:
	ifconfig eth0 192.168.2.105
	route add default gw 192.168.2.4
	dd if=/dev/zero of=A.big bs=1024k count=500
------------------------------------------------------------------------------
On evm console:

root
brecis
gprof_kernel
mv kernel.gmon.out kernel.gmon.out.pre
******************************************************************************
  Move to non-evm machine for thru-put testing.

  ftp 192.168.1.100
  put A.big /dev/null
  put A.big /dev/null
  quit

  Move back to evm console.
******************************************************************************
gprof_kernel
ftp 192.168.1.100
m4
Enter password
prompt
mput kernel.gmon.out*
quit
shutdown -h now
******************************************************************************
  On i386 (192.168.1.100):
  cd usr-src/gmondecode
  gmake test
  ./COUNT
------------------------------------------------------------------------------

==============================================================================
I created a perl script (called COUNT) to try and count the instructions and
COP_0_COUNT register and difference between when system is up and running,
to when the two 500mb ftp tests were finished.  I tried not to count the
wait instruction when idle.  It appears pretty good.

The perf/cop*2 column for Diff is the lines that are of most importance.
0.30 verses 0.35 verses 0.3685 verses 0.3696.  To get the most instructions
executed, the value should be as close to 1.0 as possible.  Of source, this
doesn't necessarily get the fastest thru-put!
==============================================================================
