Could something be wrong with Linux sr5sieve?

jasong · 2007-03-10, 18:54

1315 factors found 5700-5750, expected 1440.

That's a pretty extreme difference, in my opinion. What do you guys think?

masser · 2007-03-10, 19:00

No, that's not too bad. I wouldn't be too worried.

geoff · 2007-03-10, 23:53

Quote:

Originally Posted by jasong

1315 factors found 5700-5750, expected 1440.

That's a pretty extreme difference, in my opinion. What do you guys think?

Don't take the expected number of factors too seriously, the formula is not very accurate. It is calculated as N*(1-log(pmin)/log(pmax)) where N is the number of remaining terms, and pmin <= p <= pmax the range of factors tested.

Does anyone know a better estimate?

geoff · 2007-03-12, 03:13

I you are concerned that factors are being missed, there is a small test sieve sr5check.tar.gz at http://www.geocities.com/g_w_reynolds/sr5sieve/tests. There are instructions in the enclosed README, but briefly if you download sr5check.tar.gz and run `sr5sieve -i sr5check.txt -p 100e6 -P 150e6' it should find 10533 new factors, and the factors.txt file should match the checked-factors.txt file (which has been verified with NewPGen 2.82).

Another test in included in the sr5sieve source (see INSTALL): running `make realclean' then `make ARCH=k8 check' will test the modular arithmetic functions against GMP (for the -k8 build in this case). You need to have libgmp installed for this test.

Of course there are many corner cases that are not covered, but if both of these tests pass then that should give a reasonable assurance that most factors are being found in normal use.

2007-03-10, 18:54	#1
jasong "Jason Goatcher" Mar 2005 6663₈ Posts	Could something be wrong with Linux sr5sieve? 1315 factors found 5700-5750, expected 1440. That's a pretty extreme difference, in my opinion. What do you guys think?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Is Moore's Law wrong, or is it wrong-headed (6th time around)	jasong	jasong	12	2016-05-27 11:01
what I do wrong	pepi37	Linux	4	2015-07-12 09:13
sr5sieve Benchmark thread	axn	Sierpinski/Riesel Base 5	25	2010-05-28 23:57
what wrong with this?	graeme	Lounge	22	2003-08-18 03:05
what wrong with this?		Puzzles	0	1970-01-01 00:00

2007-03-10, 19:00	#2
masser Jul 2003 wear a mask 2·829 Posts	No, that's not too bad. I wouldn't be too worried.

2007-03-12, 03:13	#4
geoff Mar 2003 New Zealand 13×89 Posts	I you are concerned that factors are being missed, there is a small test sieve sr5check.tar.gz at http://www.geocities.com/g_w_reynolds/sr5sieve/tests. There are instructions in the enclosed README, but briefly if you download sr5check.tar.gz and run `sr5sieve -i sr5check.txt -p 100e6 -P 150e6' it should find 10533 new factors, and the factors.txt file should match the checked-factors.txt file (which has been verified with NewPGen 2.82). Another test in included in the sr5sieve source (see INSTALL): running `make realclean' then `make ARCH=k8 check' will test the modular arithmetic functions against GMP (for the -k8 build in this case). You need to have libgmp installed for this test. Of course there are many corner cases that are not covered, but if both of these tests pass then that should give a reasonable assurance that most factors are being found in normal use.