YAFU (with assistance from other free software) uses the most powerful modern algorithms (and implementations of them) to factor input integers in a completely automated way. The automation within YAFU is state-of-the-art, combining factorization algorithms in an intelligent and adaptive methodology that minimizes the time to find the factors of arbitrary input integers.
YAFU has been used several times in the academic literature. If you have academic work that requires integer factorization, YAFU might be able to help.
YAFU is most optimized for general inputs up to 160 digits in size, although there is support for inputs much larger than that, if they have a special form. There are also specialized functions for handling lists of inputs and ranges of contiguous smaller inputs.
YAFU is primarily a command-line driven tool. You provide the number to factor
and, via screen output and log files, YAFU will provide you the factors.
But that's not all! You also get an interactive environment similar to MATLAB
or PARI/GP, where you can type commands and store results. The functionality
of this interface is limited, but perhaps useful, and I have plans to make it
better. YAFU also provides a vast amount of flexibility, through many many
options and a very capable expression interpreter. If you know what you are
doing, or if you read the documentation enough, you can customize the operation
of YAFU a great deal. You should have received a copy of docfile.txt, which
explains in some detail all of the available functions, how to use them,
and how to influence their behavior. A community of factorization enthusiasts
can be found at mersenneforum.org, many of whom are willing and able to
answer questions.
No installation necessary, just put the binary in whatever location you like and run it. Files that YAFU generates are placed alonside the binary, and directories/files that YAFU needs are with respect to the binary location (such as the ggnfs sievers, see below)
Which precompiled version should I use?
As of version 2.12, the following precompiled versions are available: yafu-x64-sse41.exe yafu-x64-avx2.exe yafu-x64-avx512.exe yafu-x64-avx512-ifma.exe
SSE41, AVX2, AVX512, and AVX512-IFMA refer to instruction set extensions that
the versions respectively support. SIQS, ECM, and yafu's prime sieve
all get progressively faster with more modern extensions.
See here for more information:
https://en.wikipedia.org/wiki/Advanced_Vector_Extensions
https://en.wikipedia.org/wiki/AVX-512#CPUs_with_AVX-512
Pick the best one that your cpu supports.
Roughly speaking: If you have an Intel Haswell CPU or newer (newer than 2013), use avx2. If you have an Intel SkylakeX or certain Cannon Lake or Cascade Lake cpus, use avx512. If you have an Ice Lake, Tiger Lake, Rocket Lake, or Sapphire Rapids cpu, use avx512-ifma. If you have a Zen4 or Zen5 AMD cpu, use avx512-ifma. If you have another AMD cpu newer than 2015, use avx2. Otherwise use sse41.
If you have something other than Intel or AMD, (e.g., powerPC, Arm), then yafu may not work for you.
Support and an active user community can be found here: https://www.mersenneforum.org/node/58
To run the general or special number field sieve using YAFU, you will need to provide ggnfs executable files and make the location of these files known to YAFU at runtime. The easiest way to obtain the files is through Jeff Gilchrist's website: http://gilchrist.ca/jeff/factoring/index.html
Download the .zip file for your OS and unzip it to a directory on your
computer. The easiest way to let yafu know where these are is to create/modify
a yafu.ini file (which should appear in the same directory as your YAFU
executable) to contain a line specifying ggnfs_dir in the following way:
ggnfs_dir=
For instance, if your ggnfs executables were located adjacent to the yafu folder in the directory tree, in a folder called ggnfs-bin, then the line would be: ggnfs_dir=..\ggnfs-bin\
The ggnfs executables accomplish the sieving portion of NFS. The other phases are all handled internally to YAFU via a supporting library, msieve. Here are some quick references on for more information about the number field sieve or msieve: http://mathworld.wolfram.com/NumberFieldSieve.html http://en.wikipedia.org/wiki/General_number_field_sieve http://sourceforge.net/projects/msieve/
Pre-compiled binaries are provided for windows. However, anyone is welcome to compile from source for your own system. If you can make a binary which beats the performance of one of the pre-compiled versions for a particular architecture, I'd love to hear about it!
make yafu [NFS=1] [USE_SSE41=1] [USE_AVX2=1] [USE_BMI2=1] [SKYLAKEX=1] [ICELAKE=1]
I've built and tested with gcc and icc. Enable either by using COMPILER=gcc or COMPILER=icc on the make line.
Optionally enable GNFS factorizations by setting NFS=1 during make. This will require the makefile to be edited to point to the libmsieve.a library. By default, it is assumed to be located at ../msieve/ relative to the Makefile.
YAFU now requires several other projects to be built before yafu. These are: gmp (https://gmplib.org/) gmp-ecm (http://ecm.gforge.inria.fr/) ytools (https://github.com/bbuhrow/ytools) ysieve (https://github.com/bbuhrow/ysieve)
Some support can be had here: http://www.mersenneforum.org/forumdisplay.php?f=96
If your computer supports them, yafu can make use of several modern instruction set extensions, including SSE41, AVX2, and AVX512 (various extensions). SKYLAKEX will add support for AVX512F and AVX512BW. ICELAKE will add support for AVX512F, AVX512BW, and AVX512IFMA. Primarily these are used in SIQS. If you have AVX512 on your cpu, yafu will also use AVX-ECM as the default ECM implementation. For a standalone version of AVX-ECM, please refer to: https://github.com/bbuhrow/avx-ecm
Build files are available for several versions of MSVC, but really only build.vc22 should be used.
You will need several other visual studio projects installed and built before attempting to build yafu.
As of yafu version 2.12 (Nov. 2024) I get these from these locations: mpir: https://github.com/BrianGladman/mpir ecm: https://gitlab.inria.fr/zimmerma/ecm/-/releases or https://github.com/sethtroisi/gmp-ecm pthreads: https://github.com/BrianGladman/pthreads (needed for msieve) msieve: https://sourceforge.net/projects/msieve/ ytools: https://github.com/bbuhrow/ytools ysieve: https://github.com/bbuhrow/ysieve
These all come with MSVC build files that should work.
To configure yafu once its dependencies are built, use the libs_and_extensions.props file in the build.vc22 directory. Here you can supply directories to the builds, relative to yafu's project directories (e.g., build.vc22\ecm)
A little farther down in the libs_and_extensions.props file, edit the PreprocessorDefinitions line with any instruction set extensions you want to enable for yafu's build.
I have also had success installing and using the free Intel compiler inside visual studio 2022.
YAFU might build in mingw-64 environment. I haven't tried in a long time, good luck. If you get this to work, please let me know.
If you build yafu on other platforms or using other IDE's or compilers, please let me know about it.
Detailed documentation is available in docfile.txt, which can be viewed during an interactive YAFU session by typing 'help'. The default yafu.ini also contains descriptions of available options.
If you want to see the docfile from within the program, it needs to be in the same directory as the executable.
Check back at https://github.com/bbuhrow/yafu for updates.
Here's a fun test case for factor(), which uses many of the algorithms in yafu factor(2056802480868100646375721251575555494408897387375737955882170045672576386016591560879707933101909539325829251496440620798637813)
neat example for ecm: 140870298550359924914704160737419905257747544866892632000062896476968602578482966342704