Mersenne Prime Hunting Software (With Emphasis on Utility for Current & Future Wavefronts)
Total Page:16
File Type:pdf, Size:1020Kb
Mersenne prime hunting software (with emphasis on utility for current & future wavefronts) Part 1 Software possibilities vs. approach and device type Approach Device Type Intel, similar CPUs other CPUs NVIDIA GPU AMD GPU Intel iGP Trial Factor Prime95, mprime, etc ? MfaktC Mfakto ? P-1 factor Prime95, mprime, etc ? CUDAPm1 ? ? PRP test Prime95, mprime, Mlucas Mlucas? GPUOwl ? LL test Prime95, mprime, Mlucas, etc MlucasCUDALucas clLucas (gpuOwL) ?* * GpuOwL ran on an iGP equipped i7-7500U test system; throughput was ~25% of the coinciding drop in Prime95 throughput; possible pilot error (me)? Part 2 Data regarding software that may be suitable or appears to be in common current use. Listed alphabetically Software Notes # URL for download or link to it, & discussion forumVersion (win/lin) Approx Date Compute capability >=1.3; fft lengths 1K to 65536K, exponents 7500 to 1,143,276,383 (~1.065*230) (User error can result in https://sourceforge.net/p/cudalucas/wiki/Home/ And 2.05.1 Feb 2015 CUDALucas false positives) http://www.mersenneforum.org/showthread.php?t=12576 (2.06beta) (May 5 2017) Compute capability >=1.3; fft lengths 1K to 65536K (32760K, exponents up to 580M on 1.5GB gpu) min exponent 86243; https://sourceforge.net/projects/cudapm1/ And CUDAPm1 B2<10 9 http://www.mersenneforum.org/showthread.php?t=17835 0.20 Jan 2016 New; 8M, 4M or 2M fft length versions, 2M:~25-39M exp, 4M: 50- 78M exponent; 8M to 155M exponent ; logging; Jacobi check; V0.60 is LL, V0.7 & up is PRP3 with Gerbicz check. User build https://github.com/preda/gpuowl & see GpuOwL from source. http://www.mersenneforum.org/showthread.php?t=22204 1.9 Nov 2017 Compute capability >=1.1; factor <2 95 ; http://www.mersennewiki.org/index.php/Mfaktc And Oct '14/ Feb Mfaktc 1,000,000 <exponent<=2 32 -1 Http://mersenneforum.org/showthread.php?t=12827 0.21 '15 OpenCL>=1.1 various AMD GPUs; factor<292; exponent < http://www.mersennewiki.org/index.php/Mfakto And 0.14 Nov 2014 Mfakto 4294967291 http://mersenneforum.org/showthread.php?t=15646 (0.15pre6) (Feb 2016) (Intel & non-Intel), supports exponents up to 4,294,967,291 (232- Mlucas 5) User build from source required http://mersenneforum.org/mayer/README.html 17.1 patch Dec 2017 Freebsd, linux. V24.12 & up support LLtest exponents up to 596M, V29+ to ~900M; V24.11+ factoring up to 2 billion. Jacobi check Varies/ varies/ Jan Mprime added for LL V29.3; PRP w/Gerbicz check added V29.4 https://www.mersenne.org/download/ 29.4b7 2018 Win95-10/ MacOS etc. V24.12 & up support LLtest exponents up to 596M, V29+ to ~900M; V24.11+ factoring up to 2 billion. Jacobi 29.4b7 / Prime95 check added for LL V29.3; PRP w/Gerbicz check added V29.4 https://www.mersenne.org/download/ varies Jan 2018 /? # For some, over 1/10 the supported maximum exponent LL test can be very long run times to completion (weeks, months, years, decades …) # Amount of available system memory or gpu memory may impose lower limits. Versions as listed exclude beta and prerelease versions known to exist in some cases. Mirror site http://download.mersenne.ca/index.html contains many, not all mersenne wiki reference information http://www.mersennewiki.org/index.php/Main_Page reference info on exponents: http://www.mersenne.ca/exponent/290001377 example; change exponent # changes since last update: (suggestions welcome) added PRP row in part 1 Mfaktc & mfakto factor size limits, CUDAPm1 B2 limit added; demoted Mfakto from IGP support; promoted GpuOwl & removed from NVIDIA expanded, formatted separate criteria for inclusion updated prime95 & mprime clarified LL vs PRP3 on GpuOwl versions The following are NOT part of the table above. Listing here is NOT endorsement, approval, etc. Listing here is primarily for my reference and to save people the time of suggesting items already identified. Possible candidates for future inclusion in the table above (after considering numerous criteria) Listed alphabetically clLucas Documentation, requirements, limits? http://mersenneforum.org/cllucas/ 1.04 / 1.044 Jan 2016 closed source executable package & doc installer; https://devtalk.nvidia.com/default/topic/1016010/cemplla-version-1-5-has-been-released/?offset=15 CEMPLLA install failures reported; speed? Accuracy? Etc. http://mersenneforum.org/showthread.php?t=20489&page=9 1.5 June 2017 multiple CPU types & operating systems, OpenMP. http://glucas.sourceforge.net/glucas/Installing-Glucas.html Glucas Old. https://sourceforge.net/projects/glucas/files/ 2.9.0 July 2002 https://github.com/Almajester/gpuLucas and fork https://github.com/ah42/gpuLucas/tags and discussion at Feb 2012 / gpuLucas CUDA level 3.2 or higher; source only, need to build Http://www.mersenneforum.org/showthread.php?t=14310 0.9.4 March 2012 LLR (May add in future, a list of software of which I concluded a sufficient number of criteria are not sufficiently met to include in the table. Such listing should not be considered disparaging, libelous, etc. only a judgment that it is less suitable by a sufficient margin to exclude. Consider it encouragement to adopt additional or different features, or recognition the listed software serves a different purpose. Purpose of list would be to reduce repeat suggestions. May suggest areas of change or addition that would raise chance of user acceptance. K. Kriesel Page 1 of 1 01/18/2018 18:28:42.