Wheel factorization: Efficient?

[kar_bon]

Originally Posted by 3.14159

Good luck, that'll take you about 2-3 decades.

No!

[3.14159]

Originally Posted by kar_bon

No!

Did you bother presieving before your search?

[kar_bon]

Originally Posted by 3.14159

Did you bother presieving before your search?

What do you mean? Testing k*2^n-1 with n>1M without sieving?

Did you ever read the other threads or about other projects here?
Nobody would test such high numbers without sieving first!

[3.14159]

Originally Posted by kar_bon

What do you mean? Testing k*2^n-1 with n>1M without sieving?

When n> 1M, the number should be 330k digits in length.

Originally Posted by kar_bon

Did you ever read the other threads or about other projects here?
Nobody would test such high numbers without sieving first!

Nope, I'm always in the kook section.

[mdettweiler]

Originally Posted by 3.14159

Did you bother presieving before your search?

It's through the No Prime Left Behind project--see here for more information on the specific subproject.

We sieved the range to p=140T using sr2sieve in a collaborative effort. It took about a year, and for more than half of that time it was running as a "low-priority" effort alongside our other subprojects.

In fact, we've already found 5 megabit primes from the effort. We have a few select k's from the range loaded into an LLRnet server for automatic contributions--if you'd like to give it a try yourself, you can download the client here (if you're on Windows, you'll want the V0.73 client with LLR 3.8.1) and see our list of available servers here. In addition to the megabit server (listed there as #1), we also have servers covering over parts of the top-5000 spectrum: #2 is around n=844K right now (and is our current project priority); #3 and #4 are around n=755K; #5 is at n=670K; #6 is at n=578K; and #7 is doing sub-top-5000 fill-in work.

If you're interested, I encourage you to give it a try--I don't know how much computer resources you have access to, but even with only one or two cores, a couple months of steady work can find you a megabit prime.

[3.14159]

Originally Posted by mdettweiler

It's through the No Prime Left Behind project--see here for more information on the specific subproject.

Ah, co-op searches. I work as an individual when it comes to prime searching.

Originally Posted by mdettweiler

We sieved the range to p=140T using sr2sieve in a collaborative effort. It took about a year, and for more than half of that time it was running as a "low-priority" effort alongside our other subprojects.

10^14? That will take something of about a few days or weeks for b = 2.

Originally Posted by mdettweiler

#2 is around n=844K right now (and is our current project priority); #3 and #4 are around n=755K; #5 is at n=670K; #6 is at n=578K; and #7 is doing sub-top-5000 fill-in work.

Ah. I'm sieving for k * 2⁸⁵⁶⁷⁸⁰ + 1. (257920 digits). I think I'm going to have to continue sieving for a month if I want to find anything.

Originally Posted by mdettweiler

If you're interested, I encourage you to give it a try--I don't know how much computer resources you have access to, but even with only one or two cores, a couple months of steady work can find you a megabit prime.

330000 digits? I'll settle for that one later. Right now, I'm looking for 257920 digits. (So far, I've sieved up to about 3.72 * 10¹².)

[kar_bon]

Originally Posted by 3.14159

10^14? That will take something of about a few days for b = 2.

You have to think first before answering!

10^14 for 50 k-values in the range 1M<n<2M, and not for a constant n-search like yours!

[3.14159]

Originally Posted by kar_bon

You have to think first before answering!
10^14 for 50 k-values in the range 1M<n<2M, and not for a constant n-search like yours!

I could just increment n by 1 every time I sieve k to a certain amount (Preferably in the 10¹² to 10¹³ range) , and set PFGW to automatically test the numbers.

A screen-shot of my progress, here

[kar_bon]

Originally Posted by 3.14159

I could just increment n by 1 every time I sieve k to a certain amount (Preferably in the 10¹² to 10¹³ range) , and set PFGW to automatically test the numbers.

Do it with NewPGen and tell me the timings, please! I'm interested in, perhaps we've overseen a possibility to sieve more efficiently.

[3.14159]

Originally Posted by kar_bon

Do it with NewPGen and tell me the timings, please! I'm interested in, perhaps we've overseen a possibility to sieve more efficiently.

No need for sarcasm, it was merely a suggestion. I don't quite see the appeal to meaningless sarcasm.

[3.14159]

Also: Some other random factor work by me here.
(Largest prime factor)^2 + 1 sequence, began using 4943, struck 2*p three times in a row.

Factoring the cofactor 32174438273843079981142915951736520108605272417142076547338510339651236088951473095688773110437 (95 digits)

(Running GMP-ECM at 500 curves for a 30-digit factor, B1 = 250K, B2 = 30.7M)

Failed. Running SIQS.