You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Merged the CPU lanczos code from the msieve-lacuda branch, which includes long vector support; compiling with long vectors on a modern machine with a modern gcc has yielded a 10-15% speedup
Merged fixes by Greg Childers that enable handling singleton and clique removal for datasets larger then 16GB; after years of frustration this finally fixes the 'large dataset bug'
Fixed errors in MPI block Lanczos (thanks Greg Childers)
NFS Polynomial selection updates from many people:
improved estimate of final e-value
improved handling of the factors of leading alg. coefficient
reduced verbosity
search bounds informed by years of mersenneforum work
Removed exit() call from error path in NFS filtering (thanks jrcrombie)
Add support for more modern Visual Studio and CUDA versions (thanks Brian Gladman)
Made the maximum merge weight configurable (thanks Lionel Debroux)
Updated the makefile to build GPU code for more recent GPUs
Updated CUB version
Added a fix that avoids a crash when linking to GMP 6.2.0 (thanks bsquared)
Changed the default NFS filtering target density to 90
Converted readme file to markdown, fixed links (thanks Jupiter)