Prime95 version 25.11

[Prime95]

Version 25.11 is ready for testing. For most GIMPS users, this version only contains a few very minor bug fixes.

Download links:
Windows: http://mersenneforum.org/gimps/p95v2511.zip
Windows 64-bit: http://mersenneforum.org/gimps/p64v2511.zip
Windows service: http://mersenneforum.org/gimps/winnt2511.zip
Windows 64-bit service: http://mersenneforum.org/gimps/win64nt2511.zip
Linux: http://mersenneforum.org/gimps/mprime2511.tar.gz
Linux 64-bit: http://mersenneforum.org/gimps/mprim...linux64.tar.gz
Mac OS X: http://mersenneforum.org/gimps/Prime95-MacOSX-2511.zip
FreeBSD: http://mersenneforum.org/gimps/mprim...FreeBSD.tar.gz
Mac OS X 32-bit command line: http://mersenneforum.org/gimps/mprime2511-MacOSX.tar.gz
Mac OS X 64-bit command line: http://mersenneforum.org/gimps/mprim...acOSX64.tar.gz
Source: http://mersenneforum.org/gimps/source2511.zip

This release contains both performance improvements and minor bug fixes.

From the included whatsnew.txt file:

1) PRP of k*b^n+c where b > 2 will be faster in many cases.
2) Zero-padded FFTs are about 3-4% faster. These FFTs are used in projects like SeventeenOrBust.
3) Mprime automatically sets the program to nice priority. Technically, this is not necessary since worker threads are set to the lowest possible priority. However, sysadmins might be alarmed to see a CPU intensive program not running at nice priority when executing a ps command (it takes a more complicated ps command to see that the CPU intensive threads are running at a lower priority).

Bug fixes are described here: http://www.mersenneforum.org/showpos...07&postcount=2

[Prime95]

Known bugs and fixes:

1) Yes, there is no 25.10. Prime95's version numbers and the underlying gwnum FFT library's version numbers are kept in sync. The 25.10 version number was used only for some gwnum library changes made for the PFGW project.
2) Yes, there is no 25.11 build 1 -- my mistake. I discovered it half way through making all the executables and elected not to spend time fixing it.
3) Some non-base-2 numbers get roundoff errors because of poor FFT length selection. Examples include: 211*210^2047-1, 211*210^2687-1, 211*210^7679-1. 896-bit generic modular reduction. Fixed in 26.2.
4) Roundoff error PRPing 2*3^599983-1. Fixed in 26.2.
5) The option to set maximum memory to use in each individual worker section of prime.txt was not working. Fixed in 26.2.
6) In the Mac OS X Gui, if you set the computer name or user id and then click on the Apply button, the update does not take place. Fixed in 26.2. In 25.11 you must tab out of the field, hit return, or click another field before clicking on the Apply button.
7) If using SequentialWorkTodo=0 and a worker has multiple priority work items, then the worker will needlessly stop and restart every 6 hours. Fixed in 26.2.
8) Under certain conditions if WELL_BEHAVED_WORK is set the worktodo.txt file is not written on exit. Fixed in 26.2.
9) If the GCD step in P-1 reports a factor but that factor turns out to be incorrect, then P-1 is supposed to restart from scratch. It did not. Fixed in 26.2.
10) Linux and Mac OS X 64-bit users reported occasional crashes after P-1 finds a factor. I found a potential memory allocation problem. Fixed in 26.2.

[opyrt]

Originally Posted by Prime95

Source: http://mersenneforum.org/gimps/source2511.tar.gz

Link seems broken?

[Mini-Geek]

Originally Posted by opyrt

Link seems broken?

http://mersenneforum.org/gimps/source2511.zip works. (difference is .zip/.tar.gz ending)

[umccullough]

Cool, I'll grab the sources and compile the two Haiku versions soon... (I see the code has been merged in!)

Any chance of getting an official haiku version hosted including the "security" code for use with GIMPS?

[ettl.martin]

Hi friends,

the bugs found in source259 are still in source2511

http://www.mersenneforum.org/showthr...t=11602&page=2

Best regards

Ettl Martin

[ATH]

Originally Posted by ettl.martin

Hi friends,

the bugs found in source259 are still in source2511

http://www.mersenneforum.org/showthr...t=11602&page=2

Best regards

Ettl Martin

More specific post #120 and #122:
http://www.mersenneforum.org/showpos...&postcount=120
http://www.mersenneforum.org/showpos...&postcount=122

[CADavis]

windows 64bit service link is pointing to 25.9 in the op btw

[Primeinator]

Is this version at all faster on 32bit Windows for LL testing over version 25.9 (even a small amount)?

[Prime95]

Originally Posted by Primeinator

Is this version at all faster on 32bit Windows for LL testing over version 25.9 (even a small amount)?

It is not one iota faster.

[Primeinator]

Originally Posted by Prime95

It is not one iota faster.

Thanks, just checking to see if I wanted update 25.9 or not. On a slightly different note, how much faster is the 64 bit version? I know that LL tests are not really subject to speed increases like trial factoring.