mersenneforum.org > Data CEMPLLA: An alternative to GIMPS ?
 User Name Remember Me? Password
 Register FAQ Search Today's Posts Mark Forums Read

2015-09-16, 21:00   #45
Gordon

Nov 2008

7×71 Posts

Quote:
 Originally Posted by Prime95 I'm sure it would run on fewer, just slower.
I simply quoted his own introduction, it is quite plain and can't really be misinterpreted.

"However, to install the main CEMPLLA component, which is the one that performs the LLA (the "Lucas-Lehmer Assessment"), your Windows 64-bit system must have at least five (5) installed CUDA-capable NVidia GPU cards."

When he provides satisfactory answers to my 5 questions then he'll get taken more seriously, or takes up your offer to run a few exponents we give him and print the residues for the first 10,20 and 30 thousand iterations.

Until then watch out for

a squadron of stealth bomber pigs flying past....

2015-09-16, 21:08   #46
Gordon

Nov 2008

7·71 Posts

Quote:
 Originally Posted by Madpoo If I knew the download link I'd be tempted to try it out on a sandboxed PC I keep for just such occasions. It doesn't have GPU's but I'd be curious to see what else it does, and whatever that other non-GPU thing was that he mentioned (MPCG). I keep my sandbox handy because sometimes I need to run something and monitor the network traffic in/out, and it's isolated enough where I'm not concerned about worms necessarily, and it's easily restorable to a known state. So if Mr. Mod wants to PM me the original link (or if someone else has it) I can try it out and see if it seems dangerous on the surface.
PM sent

2015-09-16, 21:33   #47
CEMPLLA Author

Sep 2015

7 Posts

Quote:
 Originally Posted by Prime95 Here is the output from prime95: M34163669 is not prime. Res64: 80B623C5884701E0 And here from CUDALucas: M( 59754283 )C, 0x6586d1e8247885ac The format isn't important. The server stores this in its database as 32-bit exponent and 16 character string. The residue standard was started long before GIMPS, dating back to at least the 1980s. The exponent, the residue. That's the critical info. You might add total runtime or time per iteration. Whatever you think the user might find useful. This is what prime95 displays every 100000 iterations as it calculates: [Sep 16 09:53] Iteration: 17600000 / 37621183 [46.78%], roundoff: 0.090, ms/iter: 16.531, ETA: 3d 19:56 But now were talking about polish issues rather than critical functionality.
Thanks again Prime95. Your help has been nothing short of inspirational.

I'm going to add a "Details" button to the "proof of accuracy" part of CEMPLLA that will invoke a modal dialog box detailing the final iteration's Res64, a 32-bit CRC of the last iteration (mostly because there's no reason not to), and probably the Res64 of the last, say, 16 iterations. It will also have a "Save Details" button that will enable saving the info as a text file.

The same details will be saved for the CEMPLLA database's candidates, but in binary format on the cloud, for release when the primes are found.

For what it's worth, I also need to "code sign" CEMPLLA so Windows will stop referring to it as a "program from an unknown publisher". Coincidentally, I just read that the EFF and the ISRG have started a free root certificate program, so (hopefully) I may not have to shell out $bucks for a root certificate "subscription".. But further research is needed there.. As for the profiling stuff, I'm still on the fence about that.. Quote:  Originally Posted by Prime95 I'm sure there will be plenty more comments. Ignore the tone of the replies, but pay attention to the underlying feedback. I know you are excited and eager to get started. This task takes great patience and attention to the details is what keeps users in it for the long run. Advice well worth heeding, both now and in the future. Quote:  Originally Posted by Prime95 There are other uses for residues. For example, prime95 can print out the residue after say 1000 LL iterations. If your program can do the same, then we simply run 1000 iterations using both programs on a variety of big exponents. Your users will be reassured once they see this happen. No one can accuse you of bitcoin mining because they can see proof that your program works as advertised. When dealing with the general public, the higher your provable credibility the more GPUs you will attract to your cause. The above is quoted from one of your previous posts, which I cut and pasted here. Hope I didn't violate any forum rules doing that.. Anyway, that's another great idea (not a big surprise) - Again, in the "proof of accuracy" part of CEMPLLA, I'll have to add a capability that enables the participant to limit the iteration count to only a certain number of iterations before the test begins. This will probably involve adding two new buttons, "Options", and "Begin Test" (just thinking out loud here). Then, when that number of iterations has been completed, the test will "pause" (functionality that already exists), and the aforementioned "Details" button will appear, allowing the participant to view and/or save the info. Then the "Continue" button can be used to resume the test. In fact, I see no reason not to implement all of the above into the already existing pause and resume functionality of the CEMPLLA component. That way, it will be available for the database candidate tests as well.. Of course, all of the above will be entirely moot if no one even bothers to install CEMPLLA.. Which, of course, only proves your point that "This task takes great patience".. When I've implemented all of the above, I'll change the version to 1.1, and update the Readme.txt file.. Quote:  Originally Posted by Prime95 This is an example of where you can learn a lot from members here. If I'm not mistaken, all the 100M digit candidates are already factored to at least 2^64. Don't suppose you could point to a URL for that (por favor)? I googled "Mersenne Prime factors", and couldn't find it, but then, I didn't spend all day on it either.. Regardless though, the MPCG program allows the participant to directly input a "known K Factor" for any Mersenne Prime Candidate in the database, and can earn a substantial "division contribution" (explained in the docs) for achieving a modulus of zero. But it's hardly a foolproof system because, among other things, the divisor (=2*[K Factor]*P+1) is unfortunately limited to 64 bits. More grist for the vitriol mill, I'm afraid.. Quote:  Originally Posted by Prime95 Yes, I'm the author. If you pocket the entire award for yourself, I expect you'll be in the red by the time you find a prime. This is a labor of love. Well, yes, for the first two to three years of the CEMPLLA project, the only motivation I had was, if not "love", at the very least, curiosity, and a strong desire to see how far I could take it. Then it took on a life of its own, because every new challenge spawned a new insight, which spawned a new challenge, and on and on. As for being in the red, well, already there.. 2015-09-16, 21:40 #48 ewmayer 2ω=0 Sep 2002 República de California 22×7×389 Posts Quote:  Last but not least, CEMPLLA is not affiliated in any way, shape, or form with "GIMPS", which is an organization that is involved in, among other things, achieving the same goals as CEMPLLA. They are using software that is twenty (20) years old, and their software, unlike CEMPLLA, does not employ GPU technology, and is not hardware-extensible - i.e. adding more GPUs will not make their software run any faster. CEMPLLA, on the other hand, does employ GPU technology (albeit, NVidia only), and is hardware-extensible (i.e. the more NVidia GPU devices you add, the faster it will run). So. George, is it true that you've barely touched your code in 20 years? You lazy bum! [Pretty impressive, though, that you anticipated the entire Intel SSE2 and AVX instruction sets, down to the minutest detail, a decade or more before they were announced - where can I get that kind of advance insider info?] You should also start looking into this whole 'parallel programming' thing - sounds like it's an nVidia innovation, but maybe we can bug the major CPU vendors to also start supporting it. In a similar vein, it seems this whole dated, rickety 'Internet' contraption is over 30 years old - I can't believe people are still using it! 2015-09-16, 21:46 #49 Gordon Nov 2008 1111100012 Posts Quote:  Originally Posted by CEMPLLA Author Wrote some drivel that's been snipped. 1. Explain why it must be run as administator 2. Explain why virus checking must be disabled 3. Explain why wddm whatever must be turned off. - especially when Nvidia recommend you leave it turned on. Not airy-fairy crap, detailed technical explanations 4. What web sites the software will try to connect to, where it is hosted and who the registrant is 5. SOURCE code - because at the moment you simply come across as trying to get us to install your bitcoin mining software. ANSWERS??? Seems like over at the Nvidia developers forum they are just as sceptical as we are. Link not posted as it contains your download link. 2015-09-16, 21:53 #50 Prime95 P90 years forever! Aug 2002 Yeehaw, FL 6,737 Posts Quote:  Originally Posted by CEMPLLA Author Don't suppose you could point to a URL for that (por favor)? Try http://www.mersenne.info/trial_facto...a/1/300000000/ and http://www.mersenne.org/report_expon...p_hi=332300000 2015-09-16, 22:01 #51 chalsall If I May "Chris Halsall" Sep 2002 Barbados 22×32×5×72 Posts Quote:  Originally Posted by Gordon "However, to install the main CEMPLLA component, which is the one that performs the LLA (the "Lucas-Lehmer Assessment"), your Windows 64-bit system must have at least five (5) installed CUDA-capable NVidia GPU cards." It is interesting to note you can't even rent such a machine from Amazon -- the closest available is a g2.8xlarge which has four. A pity... I'd be willing to spend the ~$0.50 an hour for a spot instance (and another few cents an hour to "jail" it through another instance) just to sniff what this thing actually does traffic wise.

2015-09-16, 22:13   #52
Gordon

Nov 2008

1111100012 Posts

Quote:
 Originally Posted by chalsall It is interesting to note you can't even rent such a machine from Amazon -- the closest available is a g2.8xlarge which has four. A pity... I'd be willing to spend the ~\$0.50 an hour for a spot instance (and another few cents an hour to "jail" it through another instance) just to sniff what this thing actually does traffic wise.
Hmmm, over at the Nvidia forum he explains that if you don't have 5 then it installs something else instead. To quote the author

"Your system sounds like a good starting point for beta-testing. Fact is though, I didn't make it quite clear (my bad) that anyone running Windows 64-bit can participate, regardless of whether or not they have any NVidia GPUs at all. If the CEMPLLA Install program detects that you have five or more NVidia cards in your system, it will install "CEMPLLA", the program that executes the primality test on a selected prime candidate. Otherwise, it will install "MPCG", which is another program that ensures that the CEMPLLA System's 'prime candidate list' is as viable as possible.."

Or he could just download a list of all the factored exponents.

Perhaps the MPCG is the trojan...

1. cemplla - does what it says, tests for primality.
2. extremely few people will actually get that - the 5 gpu requirement
3. EVERYONE else runs this other program...

2015-09-16, 22:14   #53
wombatman
I moo ablest echo power!

May 2013

23·7·31 Posts

Quote:
 Originally Posted by Gordon 3. Explain why wddm whatever must be turned off. - especially when Nvidia recommend you leave it turned on.
I can help answer this one. Here's a description of what TDR does: https://msdn.microsoft.com/en-us/lib...(v=vs.85).aspx

I have increased the timeout value from the default of 2 seconds to 20 seconds to run CudaLucas without as many hangs. My understanding is that if the video card doesn't respond for the set value, Windows will automatically try and kill the display driver and then restart it, thus preventing a "freeze". This can be annoying, and it would still be good to get an explanation for turning it off completely, but it's not totally unheard of to mess with this particular parameter.

2015-09-16, 22:19   #54
tha

Dec 2002

2·389 Posts

Quote:
 Originally Posted by Madpoo Shh... don't you know if you say his name (or even his initials) 3 times, he'll show up...
RDS is by far the most knowledge mathematician around here as far as I can tell. Maybe he has learned that his time is spent best for the cause if he limits his time to those who show to be willing and able to learn. He can safely ignore this guy and leave the bashing to lesser gods pawns like us.

2015-09-16, 22:30   #55
CEMPLLA Author

Sep 2015

7 Posts

Quote:
 Originally Posted by science_man_88 @CEMPLLA Author I'm guessing you've thought about these: 1) can any mersenne numbers with co-prime exponents share a factor ? 2) can you reduce the number of factor candidates for any particular edit:(prime) exponent p ?
The MPCG program, which is installed by the CEMPLLA Install program if the machine does not have five or more GPU boards, is the "factoring" part of the CEMPLLA System. It's not nearly as advanced as the CEMPLLA program that's installed on machines with five or more GPU boards.

That said, it's operating principle is very simple. It attempts to find a modulus of zero for the selected prime candidate (2^P-1) by dividing it by a divisor (D) that is calculated from a "K Factor" (K), according to the formula: D = 2 * K * P + 1.

Each of these "K Factors" are tried, one by one, in numerical order by the MPCG program. It starts with a K Factor of one, and increments the "Current K Factor" by one for each "iteration". And, of course, it keeps track of the "Current K Factor" for each candidate.

If it finds a modulus of zero, it removes the prime candidate from it's database of available candidates, and no further factoring is done.

The MPCG program only exists to ensure that the CEMPLLA component (the one that performs the LLT) won't select factored "candidates" (which, by definition, are no longer "prime candidates" at all), or if it already has, will cause the CEMPLLA component to abandon the LLT as soon as the candidate's absence in the candidate database is detected.

So no attempt is ever made by the CEMPLLA System to find more than the numerically lowest factor of a Mersenne "candidate" (aka: number). So call me ignorant (and I'm sure there'll be at least 10 posters here who will), but that would seem to obviate the whole issue of co-primes and shared factors. As for reducing the number of factor candidates, the aforementioned equation is the only mechanism currently employed to do that.

However, there is, within the MPCG program, an opportunity for the participant to input a "Known K-Factor", the details of which I've actually explained in another post. It's not perfect, but it does give the mathematically-minded an opportunity to voluntarily contribute to the shared (and visible) CEMPLLA database of Mersenne numbers on a level beyond that of voluntary hardware utilization.

 Similar Threads Thread Thread Starter Forum Replies Last Post a1call Programming 19 2019-11-08 22:31 paulunderwood Miscellaneous Math 36 2019-08-26 08:09 a1call Miscellaneous Math 41 2019-07-21 14:19 CRGreathouse Number Theory Discussion Group 51 2018-12-16 21:55 ixfd64 Software 1 2008-04-26 21:28

All times are UTC. The time now is 22:28.

Sun Mar 29 22:28:55 UTC 2020 up 4 days, 20:01, 3 users, load averages: 1.42, 1.48, 1.43