srsieve/sr2sieve enhancements - Page 2

rogue · 2012-03-20, 16:03

I got annoyed at the bad "factors per second" computation in sr1sieve. That is now fixed. In fact sr1sieve now outputs a status line that is the same as sr2sieve. The program now starts execution at low/idle priority. The computation of p/sec is still wrong, so that will be fixed, hopefully in the next release.

I've posted s1rsieve 1.4.2 here.

Note that sr5sieve.c/sr5sieve.h will be renamed to sr2sieve.c/sr2sieve.h in the next release of that software.

Puzzle-Peter · 2012-03-20, 20:10

Are you planning to put checkpointing into sr1sieve?

rogue · 2012-03-20, 22:23

Quote:

Originally Posted by Puzzle-Peter

Are you planning to put checkpointing into sr1sieve?

It isn't necessary. Using -i and -o with the same file name is all you need. When you ^C, the file will be updated with the last prime tested.

Puzzle-Peter · 2012-03-21, 15:44

Quote:

Originally Posted by rogue

It isn't necessary. Using -i and -o with the same file name is all you need. When you ^C, the file will be updated with the last prime tested.

Great, didn't know that. Thanks!

rogue · 2012-04-09, 15:52

I've posted srsieve 1.0.4 here.

This addresses a console output issue. The remaining candidates file is correct. An example is for k=1600 = (2^3*5)^2. srsieve output that the algebraic factorization was 30^2, not 40^2 as the code computed 2*3, not 2^3.

rogue · 2012-04-28, 16:38

I want to clean up the code base for srsieve/sr1sieve/sr2sieve because I know that there are various issues with compiling/linking on Win64, Mac, and Linux. They are easy to fix, but a pain if you run into them when you want to upgrade. I only have access to Win64 and Mac. If some of you could try to build on Win32 and Linux and give me some feedback, I would appreciate it.

BTW, I've already made a few changes to the code to address some of the issues. You can d/l the current sources (all in one zip file) from http://home.roadrunner.com/~mrodenkirch/sieves.zip

Mathew · 2012-04-28, 19:13

rogue,

Linux-64bit Ubuntu 11.10 just typed make in each of the folders and all binaries were created.

Did a simple test case everything looks good.
Algebraic factors were found, srfile can remove factors,

Linux-32bit Ubuntu 10.04
srsieve- changed Makefile to ARCH=i686 then make worked fine
Same test with 64-bit, same results.

sr1sieve- changed Makefile to ARCH=i686 errors
sr2sieve- changed Makefile to ARCH=i686 errors

See attached output for the errors.

rogue · 2012-04-30, 13:30

I've updated the zip file. Let me know if that solves the compile/link problems on 32-bit Linux.

Mathew · 2012-05-08, 21:29

rogue,

Using ARCH=x86 for sr1sieve and sr2sieve
I did the same tests as with 64-bit versions, everything looks good.

Sorry for the slow response.

KEP · 2012-05-27, 13:29

Could srsieve/sr2sieve be developed such as it supports n<=10,000,000,000? I'm asking because currently mersenne@home is appearing to do things the wrong way in their mersenne-prime searching. (At best they are only tripple cheking work that has already been done, but most likely they are waisting several thousand CPU years on doing work that will lead to no new discoveries and eliminate nothing new from the quoue of work). Before I continue to debate with the Polish guy at mersenne@home, I needs a version of srsieve/sr2sieve wich supports sieving of n<=10G for at least base 2. So can it be done?

If possible, could the package this time be with both a 64bit aswell as a 32bit (ready to use) version of srsieve/sr2sieve?

Regards

Kenneth

rogue · 2012-05-27, 13:40

Quote:

Originally Posted by KEP

Could srsieve/sr2sieve be developed such as it supports n<=10,000,000,000? I'm asking because currently mersenne@home is appearing to do things the wrong way in their mersenne-prime searching. (At best they are only tripple cheking work that has already been done, but most likely they are waisting several thousand CPU years on doing work that will lead to no new discoveries and eliminate nothing new from the quoue of work). Before I continue to debate with the Polish guy at mersenne@home, I needs a version of srsieve/sr2sieve wich supports sieving of n<=10G for at least base 2. So can it be done?

If possible, could the package this time be with both a 64bit aswell as a 32bit (ready to use) version of srsieve/sr2sieve?

Factors of Mersenne numbers have special form. There is no way that srsieve/sr2sieve could compete with current factoring techniques.

I cannot build a 32-bit Windows exe. I can only build a 64-bit Windows exe.

2012-03-20, 16:03	#12
rogue "Mark" Apr 2003 Between here and the 1100001100100₂ Posts	sr1sieve I got annoyed at the bad "factors per second" computation in sr1sieve. That is now fixed. In fact sr1sieve now outputs a status line that is the same as sr2sieve. The program now starts execution at low/idle priority. The computation of p/sec is still wrong, so that will be fixed, hopefully in the next release. I've posted s1rsieve 1.4.2 here. Note that sr5sieve.c/sr5sieve.h will be renamed to sr2sieve.c/sr2sieve.h in the next release of that software.

2012-04-28, 16:38	#17
rogue "Mark" Apr 2003 Between here and the 2²×7×223 Posts	I want to clean up the code base for srsieve/sr1sieve/sr2sieve because I know that there are various issues with compiling/linking on Win64, Mac, and Linux. They are easy to fix, but a pain if you run into them when you want to upgrade. I only have access to Win64 and Mac. If some of you could try to build on Win32 and Linux and give me some feedback, I would appreciate it. BTW, I've already made a few changes to the code to address some of the issues. You can d/l the current sources (all in one zip file) from http://home.roadrunner.com/~mrodenkirch/sieves.zip Last fiddled with by rogue on 2012-04-28 at 16:42

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Sieving twins with srsieve	henryzz	Twin Prime Search	0	2014-03-18 12:44
Intel announces multi-core enhancements for Haswell chips	ixfd64	Hardware	8	2012-02-10 20:32
LLRnet enhancements	kar_bon	No Prime Left Behind	10	2008-03-28 11:21
TODO list and suggestions/comments/enhancements	Greenbank	Octoproth Search	2	2006-12-03 17:28
Suggestions for future enhancements	Reboot It	Software	16	2003-10-17 01:31

2012-03-20, 20:10	#13
Puzzle-Peter Jun 2009 683 Posts	Are you planning to put checkpointing into sr1sieve?

2012-04-09, 15:52	#16
rogue "Mark" Apr 2003 Between here and the 2²·7·223 Posts	I've posted srsieve 1.0.4 here. This addresses a console output issue. The remaining candidates file is correct. An example is for k=1600 = (2^35)^2. srsieve output that the algebraic factorization was 30^2, not 40^2 as the code computed 23, not 2^3.

2012-04-30, 13:30	#19
rogue "Mark" Apr 2003 Between here and the 14144₈ Posts	I've updated the zip file. Let me know if that solves the compile/link problems on 32-bit Linux.

2012-05-08, 21:29	#20
Mathew Nov 2009 2·5²·7 Posts	rogue, Using ARCH=x86 for sr1sieve and sr2sieve I did the same tests as with 64-bit versions, everything looks good. Sorry for the slow response.

2012-05-27, 13:29	#21
KEP Quasi Admin Thing May 2005 3C1₁₆ Posts	Could srsieve/sr2sieve be developed such as it supports n<=10,000,000,000? I'm asking because currently mersenne@home is appearing to do things the wrong way in their mersenne-prime searching. (At best they are only tripple cheking work that has already been done, but most likely they are waisting several thousand CPU years on doing work that will lead to no new discoveries and eliminate nothing new from the quoue of work). Before I continue to debate with the Polish guy at mersenne@home, I needs a version of srsieve/sr2sieve wich supports sieving of n<=10G for at least base 2. So can it be done? If possible, could the package this time be with both a 64bit aswell as a 32bit (ready to use) version of srsieve/sr2sieve? Regards Kenneth