mfaktc: a CUDA program for Mersenne prefactoring

[ET_]

Thank you Oliver!


Gotta do some testing this weekend... for M412341233412341

Luigi

P.S. May I ask Kevin for a Windows executable soon?

[TheJudger]

Originally Posted by ET_

Gotta do some testing this weekend... for M412341233412341

But not with mfaktc. Exponents are still limited to 32 bit. Currently I don't plan to expand this, sorry!

Oliver

[ET_]

Originally Posted by TheJudger

But not with mfaktc. Exponents are still limited to 32 bit. Currently I don't plan to expand this, sorry!

Oliver

No problem about it... there is plenty of search space within 32 bits...

Luigi

[kjaget]

Originally Posted by ET_

Thank you Oliver!


Gotta do some testing this weekend... for M412341233412341

Luigi

P.S. May I ask Kevin for a Windows executable soon?

Yep, I'll work on it tonight EDT. Based on the past few (pre)releases it should work out of the box.

Kevin

[ixfd64]

This could greatly help in finding factors of small double Mersenne numbers.

[TheJudger]

A quicky on GTX 480 (Fermi)

Raw GPU (candidates per second (percentage of 71bit kernel)):

GPU kernel	M66362159	M3321932839
71bit		114.0M		90.5M
75bit		186.7M (164%)	151.4M (167%)
95bit		162.7M (143%)	131.1M (145%)

Definitely faster than I've expected!
Those mul24 hurts Fermi more than expected!
Of course these are preliminary numbers. The new kernel isn't checked well.

Oliver

[Uncwilly]

Originally Posted by fivemack

76 bits:

<checks bank account>

[kjaget]

OK, windows x64 version here. I did a few quick tests and it seems to be fine but let me know if anyone finds any problems.

The zip file was too large when I included cudart.dll, but it's still available on this thread so grab it from there if you need it (http://www.mersenneforum.org/showpos...&postcount=201).

mfaktc-0.07-win64.zip

[henryzz]

Originally Posted by kjaget

OK, windows x64 version here. I did a few quick tests and it seems to be fine but let me know if anyone finds any problems.

The zip file was too large when I included cudart.dll, but it's still available on this thread so grab it from there if you need it (http://www.mersenneforum.org/showpos...&postcount=201).

Attachment 5255

Would have fitted if you hadn't included the source. That would be my preference though I don't know about others.

[henryzz]

Originally Posted by kjaget

OK, windows x64 version here. I did a few quick tests and it seems to be fine but let me know if anyone finds any problems.

The zip file was too large when I included cudart.dll, but it's still available on this thread so grab it from there if you need it (http://www.mersenneforum.org/showpos...&postcount=201).

Attachment 5255

Whats the exponent limit on old cards(8600 GTS)? This binary didn't work with billion digit exponents although it did work with M165150761.

[ET_]

CUDA device info
name: GeForce 9500M GS
compute capabilities: 1.1
maximum threads per block: 512
number of multiprocessors: 4 (32 shader cores)
clock rate: 950MHz

got assignment: exp=3321928097 bit_min=69 bit_max=70

tf(3321928097, 69, 70);
k_min = 88848372480
k_max = 177696745119
class 0: tested 42270720 candidates in 7882ms (5362943/sec) (avg. wait: 163644usec)
avg. wait > 500usec, increasing SievePrimes to 27000

It works fine on Windows 7 64 bits...

Luigi