20201118, 06:01  #1 
Nov 2020
3_{10} Posts 
hi there, and a question
Hi,
I recently found this site with a lot of oldschool enthusiasts. Glad to great you all! My primary goal was to factor a 512bit number, but during the process I revealed a whole and interesting mathematical universe :) With different small programs that like optimization, authors that publish their findings and knowledge, constantly improved algorithms etc. Probably a portion of my enthusiasm will spill on the tools too. I'm a system programmer on Windows and used to program in x64 bit asm with all modern extensions. So, yafu worked well on a 256bit key. With a 384bit key it starts to food with numbers like : 4752 13239466028743 882405501769571690163767 4524 20577102057791 893319195611906354590837 save 7.244624e014 4.8911 4724113.50 9.246223e009 rroots 2 9522397161563954881537604020 8087755128263 920091041457194254691175 7788 6938120111755 9516 35989004298241 741777587812416893280037 What could it be ? Can it be suppressed, I think the output slows down the calculation. "silent" option in .ini doesn't help 
20201118, 06:28  #2 
"Curtis"
Feb 2005
Riverside, CA
17·271 Posts 
Welcome!
256bit inputs are small, and YAFU will run the quadratic sieve to factor them. This method is faster for inputs below about 100 decimal digits, but the time to factor scales more steeply than the Number Field Sieve (NFS). 384 bits is well above that 100 digits, so YAFU is running NFS. What you see on screen is the output of the first phase of NFS algorithm polynomial selection. That phase will run for roughly 5% of the total factorization time. Sieving takes 8085% of the total time, and the linear algebra takes 1015% at the end. A 384bit input on a desktop quadcore should take a couple hours to factor. Difficulty doubles every 5 decimal digits or so, so a 512bit number will take 510 days (depending on the speed of your hardware). 
20201118, 06:48  #3 
Romulan Interpreter
Jun 2011
Thailand
3×11×277 Posts 
Welcome to the fry.
You are on track, don't worry about the numbers you see on screen. As I remember, they can be disabled, yes, but it is a bit tricky. Please have a look in yafu readme file. Yafu uses different algorithms to factor numbers, according with the number size. It starts "small", with trial factoring and Pollard algorithms, and if your number "survives", it moves "higher", to ecm, siqs, (s/g)nfs. For "higher" part, it used third party tools, external or incorporated inside, and messages you see may come from these "external" tools (in case, poly select, msieve). The reason you don't see them for a 256bit number (about 70 something digits) is the fact that this number is always factored by ecm or siqs, it never reaches nfs stage. The situation changes for composites over (about) 106 digits. Run a "tune" to your system to see where exactly is the cutting point (it is systemdependent). Sysprogrammer here too, btw. 
20201118, 12:53  #4  
Feb 2017
Nowhere
23·181 Posts 
Quote:
The performance info above should be pasted into any FAQ or README file that's supposed to tell wouldbe users what to expect. 

20201118, 14:27  #5  
"Ben"
Feb 2007
6441_{8} Posts 
Quote:
As LaurV mentioned, the endless lines of numbers come from 3rd party software (msieve) and have annoyed me before too... mostly when I want to see something output to the screen from earlier in the day but which has long ago been shoved out the back of the terminal's screen buffer. However from a performance standpoint I believe they are Mostly Harmless. 
