View Single Post
2007-05-27, 10:33   #11

"Richard B. Woods"
Aug 2002
Wisconsin USA

22·3·599 Posts

Quote:
 Originally Posted by axn1 Here, one cannot discount the LL cost. < snip > Let http://mersenneforum.org/cgi-bin/mimetex.cgi?t be the LL cost < snip > The table gives the break even pts for each of the P-1 levels as specified. Now, the thing to note is that, given a sufficiently large $t$, a combo with a lower coefficient for $t$ /will/ be better. So let's calculate how large is sufficiently large.
No, no, that's not how Prime95 works.

It's not choosing an exponent (or choosing an L-L cost) to best utilize combination X, or Y, or Z. If one wants to analyze that way, one can, but that's not what the Prime95 software is doing.

When Prime95 is given the task of choosing a B1/B2 combination for doing P-1 on some exponent, it's selecting the combination that's most efficient to use on the given exponent (while also taking into consideration the specified "Available Memory", the CPU type, and the bit-level to which the exponent has been previously TFed). So the L-L cost is not a variable whose value can be chosen; it's a constant during any one search for an optimal B1/B2 combination.

In my example, combinations X, Y, and Z are just representing three out of hundreds or thousands of possible combinations that Prime95 examines each time it's choosing P-1 bounds for one single exponent.

Last fiddled with by cheesehead on 2007-05-27 at 11:17