Operation: Billion Digits

[Uncwilly]

Originally Posted by ET_

Do you know who worked the first two from 68 to 69?

I used the search function on these forums (note that if you search "3321931099" you won't find "M3321931099").

Aitsen
M3321930371 no factor from 2^60 to 2^69.
http://www.mersenneforum.org/showthr...ht=M3321930371
Joshua2
M3321931099 no factor from 2^67 to 2^69.
http://www.mersenneforum.org/showthr...ht=M3321931099

[VBCurtis]

Luigi--
Would it be difficult to allow tenths of a bit to be entered for factoring limits? It appears we're all intimidated by the time requirements to factor from 73 to 74 bits, but if (for example) I could divide the work into 2 (or more) pieces, I'd tackle 73 to 73.6 on one machine, 73.6 to 74 on a second machine, and make getting numbers to 74 bits (and thus level 8!) feasible.

Can the program be altered easily to accept such entries? Is this more bookkeeping trouble than it's worth?
-Curtis

[ET_]

Originally Posted by VBCurtis

Luigi--
Would it be difficult to allow tenths of a bit to be entered for factoring limits? It appears we're all intimidated by the time requirements to factor from 73 to 74 bits, but if (for example) I could divide the work into 2 (or more) pieces, I'd tackle 73 to 73.6 on one machine, 73.6 to 74 on a second machine, and make getting numbers to 74 bits (and thus level 8!) feasible.

Can the program be altered easily to accept such entries? Is this more bookkeeping trouble than it's worth?
-Curtis

It's feasible.
I'm just looking for some free time to implement also a couple of improvements, as the time needed for each search and a new sieve to speed up the search.

If you have requests, this is the right time to write them down

Luigi

[wblipp]

Originally Posted by ET_

If you have requests, this is the right time to write them down

Luigi,

I've been thinking about trial factoring in other bases, but I don't know if it makes sense to adapt your program. The parameters of interest are pretty far from Operation Billion Digits, so the design choices may make adapation infeasible.

Factors of p^q-1 are either q or 2kq+1. But the exponent q I'm interested in is tiny compared to OBD - usually less than 100 and nearly always less than 1000. On the other hand, the base b of interest can be large compared to "2" - often 10^7 and occasionally 10^150.

William

[ET_]

Originally Posted by wblipp

Luigi,

I've been thinking about trial factoring in other bases, but I don't know if it makes sense to adapt your program. The parameters of interest are pretty far from Operation Billion Digits, so the design choices may make adapation infeasible.

Factors of p^q-1 are either q or 2kq+1. But the exponent q I'm interested in is tiny compared to OBD - usually less than 100 and nearly always less than 1000. On the other hand, the base b of interest can be large compared to "2" - often 10^7 and occasionally 10^150.

William

It sounds interesting.

Three questions:

1 - May you point me to links with heuristics (like mod reductions to speed up the process)?

2 - Are you sure that programs like LLR, PRP, PFGW, NewPgen don't offer that feature? They are far better optimized and tested.

3 - Are you sure that trial factoring on your numbers would be more efficient than, say, ECM or QS?

If the answer is "yes" for all questions, I will gladly start a new project to help your search

Luigi

[Joshua2]

I only have two requests.
Main request is to have it search faster, while still being sure it finds all factors. :)
Although, if it was way way faster, it might be worth it to miss some, cause so many more numbers could be searched.

Also, I second the request to be able to not have to do whole bits at a time, ie 73.0 to 73.5.

[wblipp]

Originally Posted by ET_

Three questions:

Thinking about it some more, this is probably not the way to do this. Trial factoring would not be the method of choice for substantial factors, it would be a fast check for small factors before invoking other methods. And the "p" in the 2kp+1 would be the same set of small p's over and over, so resieving from scratch every time may not be efficient, either.

[Phinne]

Hello,
I'm totally new to this, and I hope I'll do this right. I'm gonna try the M3321928241 exponent.

[ET_]

Originally Posted by Phinne

Hello,
I'm totally new to this, and I hope I'll do this right. I'm gonna try the M3321928241 exponent.

Hi Phinne, welcome aboard.

M3321928241 has been factored up to 73 bits, so if you want to try it you should test it from 73 to 74 bits: it's a long way, maybe more than a week... Are you still interested in it or you want to change to an easier one?

Luigi

[gribozavr]

I'm going to try new Linux 64-bit binary. I reserve M3321928241.

[ET_]

Originally Posted by gribozavr

I'm going to try new Linux 64-bit binary. I reserve M3321928241.

You can reserve it on the checkout thread.

And check your mail as well... it seems that I resolved the segmentation fault problem on 32-bit pentium4 version... at least on my computer.

Luigi