Problems with sr2sieve on x86-64 Windows

geoff · 2007-12-09, 00:49

There seems to be a problem with sieving above 2^51 (2250T) on Windows with all x86-64 sr2sieve executables. This will start to affect SoB/PSP soon.

I am looking into the problem but if it turns out to be related to bugs in the Linux->Windows cross compiler then there might not be a solution by the time SoB/PSP sieving reaches 2^51.

The only `fix' at the moment is to use the 32-bit (*-windows-x86.zip) executable on Windows.

Joe O · 2007-12-09, 11:28

What are the symptoms? How did you discover this?

In the meantime, I urge all 64-bit Windows users to immediately change to the 32-bit Windows executable.

geoff · 2007-12-10, 01:07

Quote:

Originally Posted by Joe O

What are the symptoms? How did you discover this?

In the meantime, I urge all 64-bit Windows users to immediately change to the 32-bit Windows executable.

The main symptom is a crash (access violation) soon after starting. It was first reported by Bryan at RieselSieve who was trying to get a 64-bit Windows BOINC client working. It has since been confirmed by others.

The problem seems to have been present in all the x86-64 Windows sr2sieve executables since I first started making them at version 1.5.21, but has gone unnoticed because it seems only to affect the code used for sieving p-ranges above 2^51, only RieselSieve is currently sieving that deep and they only use the 32-bit Windows executable.

geoff · 2007-12-13, 01:39

The problem with the Windows x86-64 executable seem to be fixed in sr2sieve version 1.6.16.

If anyone running 64-bit Windows can help with testing, what would be really useful is to reserve a current range of about 100G or so, then add the switchs `-D duplicates.txt --no-sse2' to the normal command line for manual sieving. Send me a copy of both the factors*.txt and duplicates.txt files, and I will check the range using the Linux executable. My email address is in README.txt included with sr2sieve.

geoff · 2007-12-16, 00:25

sr2sieve 1.6.16 on Windows x86-64 is producing correct results. Thanks to those who helped with testing.

2007-12-09, 00:49	#1
geoff Mar 2003 New Zealand 13·89 Posts	Problems with sr2sieve on x86-64 Windows There seems to be a problem with sieving above 2^51 (2250T) on Windows with all x86-64 sr2sieve executables. This will start to affect SoB/PSP soon. I am looking into the problem but if it turns out to be related to bugs in the Linux->Windows cross compiler then there might not be a solution by the time SoB/PSP sieving reaches 2^51. The only `fix' at the moment is to use the 32-bit (*-windows-x86.zip) executable on Windows.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
sr2sieve 1.9.3 "problems"	pepi37	Software	8	2013-03-15 00:58
GGNFS problems on Windows	Jeff Gilchrist	Factoring	19	2011-01-16 22:38
Windows 7 Problems	spaz	Information & Answers	6	2009-12-22 04:58
How to use sr2sieve	nuggetprime	Riesel Prime Search	40	2007-12-03 06:01
Sr2sieve on PPC/Linux	BlisteringSheep	Sierpinski/Riesel Base 5	172	2007-07-06 21:29

2007-12-09, 11:28	#2
Joe O Aug 2002 525₁₀ Posts	What are the symptoms? How did you discover this? In the meantime, I urge all 64-bit Windows users to immediately change to the 32-bit Windows executable.

2007-12-13, 01:39	#4
geoff Mar 2003 New Zealand 13×89 Posts	The problem with the Windows x86-64 executable seem to be fixed in sr2sieve version 1.6.16. If anyone running 64-bit Windows can help with testing, what would be really useful is to reserve a current range of about 100G or so, then add the switchs `-D duplicates.txt --no-sse2' to the normal command line for manual sieving. Send me a copy of both the factors*.txt and duplicates.txt files, and I will check the range using the Linux executable. My email address is in README.txt included with sr2sieve.

2007-12-16, 00:25	#5
geoff Mar 2003 New Zealand 2205₈ Posts	sr2sieve 1.6.16 on Windows x86-64 is producing correct results. Thanks to those who helped with testing.