Sieve question

themerrill · 2009-01-29, 00:23

How much faster are the newer versions (1.8.7) of the sieve program than the last known 'stable' one (1.7.15)? Would it be worth it for me to switch to the experimental program once i finish my range, or should I stick with what we know works?
Also, is there any way to help raise up what our last known stable version is closer to the recent experimental versions

masser · 2009-01-29, 03:43

Try the latest version on your machine. If the speed increase is significant and you appear to find the expected number of factors over a large range (plus or minus 10%), I would recommend you just use the latest version.

geoff · 2009-01-30, 02:24

From memory I think my C2D 2.66GHz in 64-bit mode does about 575K p/sec with version 1.7.15, and about 670K p/sec with version 1.8.7. (For the 221 k sr5data.txt).

The experimental version is updated every time I try out a new piece of code or add a new feature. The stable version is only updated to fix bugs.

It is usually OK to use the latest experimental version, especially if it was posted more than a few days ago and you are running Linux. I am always using the Linux version myself to sieve for one project or another, but I rely on others to notice problems in the Windows version.

themerrill · 2009-01-31, 07:33

I am running Windows, but I'm going to take the risk and see what happens.
Currently running 1.8.8, and it is crunching about 50k more p each second than 1.7.15 (its about a 15% gain in speed). I'm working with the 211k data set. Thats a pretty nice gain.

2009-01-29, 00:23	#1
themerrill Jun 2008 12₈ Posts	Sieve question How much faster are the newer versions (1.8.7) of the sieve program than the last known 'stable' one (1.7.15)? Would it be worth it for me to switch to the experimental program once i finish my range, or should I stick with what we know works? Also, is there any way to help raise up what our last known stable version is closer to the recent experimental versions

2009-01-29, 03:43	#2
masser Jul 2003 wear a mask 2²·419 Posts	Try the latest version on your machine. If the speed increase is significant and you appear to find the expected number of factors over a large range (plus or minus 10%), I would recommend you just use the latest version. Last fiddled with by masser on 2009-01-29 at 03:43

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Advantage of lattice sieve over line sieve	binu	Factoring	3	2013-04-13 16:32
46k sieve, max. n = 5M	Cruelty	Riesel Prime Search	11	2010-03-10 22:15
Sieve Vs PRP	Chino112	Prime Sierpinski Project	6	2007-03-28 19:15
Help with PSP Sieve	cswchan	Prime Sierpinski Project	7	2007-02-03 19:24
Which sieve to use for n^n-1?	Siemelink	Factoring	11	2006-11-08 18:08

2009-01-30, 02:24	#3
geoff Mar 2003 New Zealand 13·89 Posts	From memory I think my C2D 2.66GHz in 64-bit mode does about 575K p/sec with version 1.7.15, and about 670K p/sec with version 1.8.7. (For the 221 k sr5data.txt). The experimental version is updated every time I try out a new piece of code or add a new feature. The stable version is only updated to fix bugs. It is usually OK to use the latest experimental version, especially if it was posted more than a few days ago and you are running Linux. I am always using the Linux version myself to sieve for one project or another, but I rely on others to notice problems in the Windows version.

2009-01-31, 07:33	#4
themerrill Jun 2008 A₁₆ Posts	I am running Windows, but I'm going to take the risk and see what happens. Currently running 1.8.8, and it is crunching about 50k more p each second than 1.7.15 (its about a 15% gain in speed). I'm working with the 211k data set. Thats a pretty nice gain.