The P-1 factoring CUDA program

[Chuck]

Please make it use a worktodo file to stage work.

[ATH]

Originally Posted by owftheevil

Cudapm1 output:

M61076737 has a factor: 432634830991289176546683053423

Run with B1 = 65000, B2 = 12035000, n = 3360k, d = 2310, e =2, 8 rp per pass. It used about 600Mb of device memory. Stage 2 took ~53 minutes.

Edit: Looks like about 15 minutes longer to make e = 4.

To compare with CPU speed running the same curve in Prime95.
Laptop with Corei7 2720QM sandy bridge:
using 1 core: stage1 43min, stage2 ~ 8h (3 Gb RAM)
using 4 cores: stage1: 19 min, stage2 ~ 3.8h (3 Gb RAM)

I only completed ~20% of stage2 and extrapolated the runtime.

[owftheevil]

I've been thinking about the numbers and it seems that 53m for stage 2 is faster than possible. I have a strong suspicion that there is an extra or missing factor of 2 in the code causing only half the rps to get processed. Off to look for it.

[owftheevil]

Found it. Sorry about the false expectations.

[kladner]

It is interesting to watch the process of things getting worked out.

[NBtarheel_33]

Originally Posted by owftheevil

Found it. Sorry about the false expectations.

53 minutes, 106 minutes, still a huge improvement over CPU P-1!

[NBtarheel_33]

Originally Posted by Aramis Wyler

Though it's hard to trade 25x p-1 work for 125x factoring work.

Factoring to 7x bits (assuming an increase of one bit level) gives you (roughly) a 1/7x = 1.27-1.43% chance of finding a factor.

P-1 with decent bounds will typically give you a 5-8% chance of finding a factor.

So, given 125 TF attempts, we'd expect roughly 1.6-1.8 factors found. On the other hand, 25 TF attempts should yield roughly 1.25-2.0 factors found.

If GPU P-1 allows us to increase bounds or make more frequent use of the Brent-Suyama extension, the expected number of successes will be at or above the higher end of this range. In that case, it would make complete sense to trade 125x TF for 25x P-1.

Note also that GPU P-1 will make use of the *GPU* RAM, rather than the system RAM. This could bring in P-1'ers who were previously unable to dedicate large quantities of RAM to Stage 2.

[garo]

P-1 with 2GB memory in the 61M range gives a probability of success of 3.3-3.6% depending on the TF level. Dunno where you got 5-8%.

[NBtarheel_33]

Originally Posted by garo

P-1 with 2GB memory in the 61M range gives a probability of success of 3.3-3.6% depending on the TF level. Dunno where you got 5-8%.

OK, probably won't see 8% unless you're a fan of strong P-1 with high B1 and B2, but one doesn't have to stretch too much to see 5%...

From James' site (http://mersenne.ca):

M61000000, factored to 70 bits, assuming 2 L-L tests saved, with B1=670,000 and B2=16,750,000, using K*B^N+C = 1*2^61000000-1
Probability = 5.664070%

M65000000, factored to 70 bits, with B1=800,000 and B2=24,000,000, using K*B^N+C = 1*2^65000000-1
Probability = 6.224824%

[Uncwilly]

Originally Posted by garo

P-1 with 2GB memory in the 61M range gives a probability of success of 3.3-3.6% depending on the TF level. Dunno where you got 5-8%.

I have seeing Prime95 giving around 3.75% for 60M exponents that have been taken to 73.

[c10ck3r]

Any luck getting Winbloze compiled? Once it's compiled and available, I'll reinstall my 460 to play with it :)