P-1 factoring anyone?

[delta_t]

It's been mentioned if enough RAM is dedicated to P-1 that the Brent-Suyama extension would be used at the cost of more time completing the assignment. Do we know what kind of time difference (savings) there would be if the Brent-Suyama extension were not used? It sounds like more P-1 may be able to get done but is that much time really saved?

[Dubslow]

Originally Posted by delta_t

It's been mentioned if enough RAM is dedicated to P-1 that the Brent-Suyama extension would be used at the cost of more time completing the assignment. Do we know what kind of time difference (savings) there would be if the Brent-Suyama extension were not used? It sounds like more P-1 may be able to get done but is that much time really saved?

Much of the discussion has been that we don't have a clue as to the answers for these questions. Axn was asking if anybody had data, but so far not.

[kladner]

Originally Posted by Dubslow

Much of the discussion has been that we don't have a clue as to the answers for these questions. Axn was asking if anybody had data, but so far not.

I think the results I posted above came about with just shy of 1GB per worker doing P-1. I had 2048MB allocated with 2 workers. But I could be mistaken. These might have happened in the wee hours when 4096MB were available, and I wasn't watching.

[delta_t]

Originally Posted by Dubslow

Much of the discussion has been that we don't have a clue as to the answers for these questions. Axn was asking if anybody had data, but so far not.

Yeah I saw that earlier. Unfortunately I don't have any numbers either. I'm thinking of trying a test of it though.

[Dubslow]

Originally Posted by kladner

I think the results I posted above came about with just shy of 1GB per worker doing P-1. I had 2048MB allocated with 2 workers. But I could be mistaken. These might have happened in the wee hours when 4096MB were available, and I wasn't watching.

I meant the data about the effectiveness of B-S finding factors.

[kladner]

Originally Posted by Dubslow

I meant the data about the effectiveness of B-S finding factors.

Oops. OK.

[Dubslow]

S'all good.
Sorry, that's the problem with the interblags, a.k.a. wobitubes, is that it seems so harsh














Weeeeeeeeeeeeeeeeeeeeeeeeeeeee!!!!!!!!!!!
</shit>
<thread>

[James Heinrich]

Originally Posted by KingKurly

Some day, I'll catch up to the LL wavefront. Maybe.

I'm also doing the same thing, and definite progress is being made. 6 months ago there was stragglers down to about 30M, we're now cleaned up to 47.4M. Of course, new assignments become available all the time as older first-time LLs with no P-1 come in, but we'll keep cleaning them up. I generally run my "near-current" P-1s with 2 tests saved, and re-done old (e.g. 10M range) with 3 tests saved.

[James Heinrich]

Originally Posted by Brain

256MB are not enough for current P-1 assignments. PrimeNet will assign other work then. Give 512MB per core a try. I'm happy with that.

Originally Posted by kladner

It certainly seems that P-1 will gulp down as much as it is allowed.

To do a proper P-1 on a 54M range exponent, the minimum RAM requirement is somewhere around 466MB (so 512MB is reasonable); a "generous" allocation would be about 1400MB; and Prime95 should be able to use up to around 11GB (if processing all stage2 relative primes in one pass).

[Mr. P-1]

Originally Posted by James Heinrich

To do a proper P-1 on a 54M range exponent, the minimum RAM requirement is somewhere around 466MB (so 512MB is reasonable); a "generous" allocation would be about 1400MB; and Prime95 should be able to use up to around 11GB (if processing all stage2 relative primes in one pass).

Originally Posted by http://mersenne-aries.sili.net/prob.php?exponent=54000000&b1=595000&b2=14875000&factorbits=&K=1&C=-1&submitbutton=Calculate

Recommended RAM allocation: min=466MB; good=1,387MB; max=11,340MB; insane=55,577MB;

That's the minimum recommended, not the minimum needed to do any stage 2 at all, right?

Given that half of all exponents never get any stage 2 at all, a machine with 300MB available (the minimum needed to get a P-1 assignment) will on average do a better P-1 than if this task were left to the LL-testing machine.

[James Heinrich]

Originally Posted by Mr. P-1

That's the minimum recommended, not the minimum needed to do any stage 2 at all, right?

Min/good/max/insane memory values are estimated amounts required to process 8/48/480/2400 relative primes respectively. I'm not sure what Prime95 would do with 300MB for a 54M P-1 assignment. Probably worth a quick experiment if you're curious. Just starting stage1 shows this:

Optimal P-1 factoring of M54952927 using up to 300MB of memory.
Assuming no factors below 2^71 and 2 primality tests saved if a factor is found.
Optimal bounds are B1=535000, B2=7222500
Chance of finding a factor is an estimated 4.19%
effort = 2.76GHz-days

Optimal P-1 factoring of M54952927 using up to 10000MB of memory.
Assuming no factors below 2^71 and 2 primality tests saved if a factor is found.
Optimal bounds are B1=580000, B2=13340000
Chance of finding a factor is an estimated 4.74%
effort = 3.92GHz-days

How it plays out in Stage 2 I'm not sure.