Okay, due to my wonky printk's the previous patch was pretty lousy at
its stated purpose of reducing confusion.  Attached is a better
version.  I tried to keep output mostly consistant, so:


raid5: measuring checksumming speed
   8regs     :  1900.000 MB/sec
   8regs_prefetch:  2396.000 MB/sec
   32regs    :  1652.000 MB/sec
   32regs_prefetch:  1700.000 MB/sec
   pIII_sse  :  1904.000 MB/sec
   pII_mmx   :  4696.000 MB/sec
   p5_mmx    :  6252.000 MB/sec
raid5: using function: p5_mmx (6252.000 MB/sec)

and

raid5: automatically using best checksumming function: pIII_sse
   pIII_sse  :  2280.000 MB/sec
raid5: using function: pIII_sse (2280.000 MB/sec)

-- 
Josh Litherland <josh@emperorlinux.com>
Emperor Linux