[TAG fastest working version with unaligned loads but aligned stores
peter@cordes.ca**20080319052545] 
<
[adjust lea offsets to simply addressing modes
peter@cordes.ca**20080319044041] 
[new version of rshift using unaligned loads and not overshooting.  2.0c/l Conroe, but 2.5c/l Harpertown
peter@cordes.ca**20080319040110] 
[use little-endian notation for register contents
peter@cordes.ca**20080318230601] 
[fix comments on alternatives to shufpd
peter@cordes.ca**20080318201949] 
[update benchmark comments
peter@cordes.ca**20080318190152] 
[update benchmark comments
peter@cordes.ca**20080318181416] 
[remove stale comment
peter@cordes.ca**20080317044629] 
[TAG good working version, 16byte alignment req. 2cycle core2
peter@cordes.ca**20080316233700] 
> {
}