View Single Post
Old 2006-07-25, 14:31   #8
VJS
 
VJS's Avatar
 
Dec 2004

13·23 Posts
Default

Quote:
Originally Posted by hhh
The new .dat for the Sob part is independent from the PSP-part. You can go ahead and make a new dat, Lars will notice and make a new one accordingly.
Unlike you, I don't believe we will get a big performance burst by this. Factor density has dropped, sieve speed too, and I expect a speed increase of at most 1%. At this level I would almost fear more factor misses related to human error in the changing of the dat than recource wasting.
Yours H.

I think there is a little bit of miscommunication here...

AFAIK removing n from the dat has no effect on speed, memory footprint yes but speed no. The only effect on speed is n-range and number of k.

Also the dats are created by a program so there is just as much chance that the current dat is flawed as much as the new dat would be.

THe major issue behind making a new dat is archieve ability. Yes it is more important for SOB since we are currently do P-1. But beyond this it is purely archieve and gap checking. Gap checking is important due to factor drop off as you had suggested and human error.

Often people will keep their ranges for a few weeks or months. If there is a gap it gives us the possibility to go back at these people and say hey were are these factors... often times they are true gaps sometimes they are simple errors which are fixed quickly by an e-mail exhanging some files.
VJS is offline   Reply With Quote