Factoring non-form composite
 

I've been factoring (on an extremely outdated computer) a C98, the factor of a c~103. I have been using Alpertron's ECM so far, and have done well over 1600 curves. I also have Msieve 1.07, and will use this once I get a C complier if it is suggested as practical.

My question is : is there another program that is decently fast for non-form composites?

Thanks in advance,

Roger

The consensus around here seems to be that Msieve is the fastest program for factoring numbers of that size.

As for what is practical, my ancient dual celeron 533Mhz box using msieve can factor a C103 in about a week using both processors.

The crossover point from MPQS to GNFS is somewhere in the vicinity of 100 digits, so GNFS might possibly be faster. It's a lot more involved, though, so if you want to factor only this one composite I wouldn't bother and just let Msieve do the job. If you plan on factoring such numbers on a regular basis, it might be worthwhile to install both Msieve and GGNFS and see which is faster for which number size.

Alex

Thanks to bhebden and akruppa,

To bhebden : When you say that a C103 is factored, do you mean with the factors being the largest possible (~51, 52 digits?) or having multiple factors of smaller sizes?

Thanks,

Roger

If you expect the general-purpose factoring method to take one week and you don't know the size of the factors in advance, it is better to spend several hours with ECM first as you did, because a random 98-digit number has a high probability to be at least partially factored with ECM.

The C103 I worked on was left over after a smaller factor was found in a C144 by ecm. It split into a P49*P55.

I have definately done hours with ECM : over 100, curve 2050+, still no result. The problem is I don't know if it's possible to use a faster program that can jump off from the point I've sieved to. As well, I am going to be doing many more of these composites, and the largest prime factor has a tends to grow in size along with the size of the original composite, so that ECM will take forever on an outdated computer.

Thanks,

Roger

[QUOTE=roger;92168]I have definately done hours with ECM : over 100, curve 2050+, still no result. The problem is I don't know if it's possible to use a faster program that can jump off from the point I've sieved to. As well, I am going to be doing many more of these composites, and the largest prime factor has a tends to grow in size along with the size of the original composite, so that ECM will take forever on an outdated computer.
[/QUOTE]
Maybe you should convince someone else to run msieve for you; a C98 takes maybe 10 hours on a fast modern machine.

I agree that if you have many of these to do that you should look into GGNFS.

jasonp

Whats the number?

I finished that number, at curve 2231 on Alpertrons ECM applet, and am now working on a C105, having done to approximately curve 700. I have Msieve 1.07, which everyone seems to agree is fastest for this type (non-form) of number, but have no way to use it. Apparently it needs compiling (there are only .c and .h files), and I have no programming background whatsoever. If someone could direct me to a download that gives an .exe file (something that doesn't require compiling or etc) that would be greatly appreciated.

Thanks,

Roger

If you are on windows, the .exe is available at the site.

See [url]http://www.boo.net/~jasonp/qs.html[/url]