View Single Post
Old 2007-05-27, 05:08   #9
axn
 
axn's Avatar
 
Jun 2003

112668 Posts
Default

Quote:
Originally Posted by cheesehead View Post
My private usage: One of my private definitions of "efficiency" in this context is:

(total L-L time saved by finding a factor)
/
(time spent in P-1 effort to find that factor)
As far as efficiency goes, this does /seem/ like a useful metric. And as you noted, with this metric, you can discount the LL cost as a constant multiplier.

However, as per the expression I have given, the correct metric is: (total L-L time saved by finding a factor) - (time spent in P-1 effort to find that factor) i.e. the net savings in the execution of an LL test. This is what we need to maximize. Here, one cannot discount the LL cost.

To illustrate: (contd in next post)
axn is offline   Reply With Quote