[QUOTE=a1call;475485]I hope you are aware that their crawls have been terminated probably by a robots.txt or something on this site.[/QUOTE]

Incorrect. Archive.org ignores the robots.txt file.

[QUOTE=chalsall;475488]Incorrect. Archive.org ignores the robots.txt file.[/QUOTE]

Doesn't seem to be the case at the moment. But I recall failing to find archived pages in the past. It must have been a glitch on their side or mine.:smile:

[QUOTE=a1call;475493]But I recall failing to find archived pages in the past. It must have been a glitch on their side or mine.:smile:[/QUOTE]

Unlike Google, Archive.org takes a few months to expose their spidering to public querering.

As someone who has scripts watching the logs, they ignore robots.txt (and they've also said this publiclially).

[QUOTE=pacionet;475470]So ... how many other verifications left before the announcement ?[/QUOTE]

My slowpoke Mlucas-avx2 run @4096K - started more or less as a backup to ATH's AWS run, in case he hit any instance-related issues - is 72 hours in, ETA = 12h, but no reason to delay the announcement for that.

Hmm.

Then there is no need to calculate the residue with start value s[SUB]0[/SUB]=10, neither. Ever.
See page 63 of the PDF.

The theorem 9.1 from Jansen (a.k.a. Woltman conjecture) is often misquoted as that [FONT=&quot]ε[/FONT][SUB]4[/SUB] and [FONT=&quot]ε[/FONT][SUB]10[/SUB] are equal when p={5 or 7} mod 8, but it says instead
[FONT=&quot]ε[/FONT][SUB]4[/SUB] [FONT=&quot]▪[/FONT] [FONT=&quot]ε[/FONT][SUB]10[/SUB] = 1 [FONT=&quot]↔[/FONT] p={5 or 7} mod 8 (except p==5), that is:
[FONT=&quot]ε[/FONT][SUB]10[/SUB] = [FONT=&quot]ε[/FONT][SUB]4[/SUB] [FONT=&quot]▪[/FONT] ((p%8>=5) ? 1 : -1) (except p==5)

[QUOTE=Madpoo;475345]Oh, and it was mentioned that AirSquirrels is doing a run with gpuOwL. For the astute observer, you'll recall that it only does 2M and 4M FFT sizes. Although, we were biting our nails and wondering whether it would complete successfully with a 4M FFT size. I think it's really pushing the bleeding edge there, but then again I really don't know at what point it just starts spitting out round off errors, but it sure seems like it'd be close?

Here's another clue... in the decimal expansion, there are 1 or more instances of the sequence 3141592 :smile:

Let's see, what other odd tidbits can be gleaned from the decimal expansion that won't give it away to anyone crazy enough to do the expansion for thousands of exponents and looking in them to see if they spot the same things...

Well, here's an interesting one. There's a digit that's repeated 10 times in a row. I won't say which digit just to keep it interesting. There's also a sequence in there of 8 increasing digits in a row (could be 0-7, 1-8, or 2-9 ... I won't say). :smile: Well heck, there are just so many digits in there, you can find all kinds of funny patterns if you look.[/QUOTE]

I just got through 1852 exponents that had not had an LL test on Dec 23 (when I downloaded all LL results (below 101M) for making an error rate plot) between 70M and 77.75M, and I found there are twelve Mersenne numbers that fulfill those clues.

It was quite an educational exercise and it gave me some "hands on" experience of counting Mersenne numbers: I already realized that I must only take the prime exponents, then that I could save time by excluding the exponents that had had a test on Dec 23, and finally that I had missed that there are a lot of numbers with small factors, so that I could save even more time by downloading the factored numbers (to be able to skip them). Also, it gave me an opportunity to use the gmp library again.

[QUOTE=Batalov;475483]
P.S. Archive.org kept it! - [URL]https://web.archive.org/web/20160131070826/http://www.math.leidenuniv.nl/scripties/PhDJansen.pdf[/URL][/QUOTE]

[url]https://openaccess.leidenuniv.nl/handle/1887/20310[/url]

FYI, Andreas (ATH) kindly gave me access to the AWS c5.9xlarge instance on which he completed his Mlucas run @4608K - he started on a mix of C4 and C5.18xlarge (and started with the slower avx2 build due to compile problems with the Mlucas 17.1 release in avx512 mode), and only late in the verify found the c5.9xlarge gave him slightly better speeds than the c5.18xlarge at half the price. I wanted to test his avx-512 build @4096K, as well - long story short, I got a best time of 2.4 msec/iter at that smaller FFT length on the c5.9xlarge, i.e. his verify would have needed just a little over 48 hours had it used that setup for the whole run. Looks like it's time for me to set up a user account for AWS. :)

[QUOTE=ewmayer;475504]FYI, Andreas (ATH) kindly gave me access to the AWS c5.9xlarge instance on which he completed his Mlucas run @4608K - he started on a mix of C4 and C5.18xlarge (and started with the slower avx2 build due to compile problems with the Mlucas 17.1 release in avx512 mode), and only late in the verify found the c5.9xlarge gave him slightly better speeds than the c5.18xlarge at half the price. I wanted to test his avx-512 build @4096K, as well - long story short, I got a best time of 2.4 msec/iter at that smaller FFT length on the c5.9xlarge, i.e. his verify would have needed just a little over 48 hours had it used that setup for the whole run. Looks like it's time for me to set up a user account for AWS. :)[/QUOTE]


So are you telling me it is time to upgrade the kit?

[QUOTE=airsquirrels;475505]So are you telling me it is time to upgrade the kit?[/QUOTE]

Not at all, though with all the bubble-priced *coins you've been mining using your GPU farm, you should be flush. :)

Your Xeon is doing just fine for my F30 run, where I want the speeds of my 2 runs (one on 32-core Xeon, one on 64-core KNL) to be reasonably close to each other.

But for short-time occasional work, like, say, the next M-prime verify, AWS rent-a-beast is a great way to go.

[QUOTE=ewmayer;475507]But for short-time occasional work, like, say, the next M-prime verify, AWS rent-a-beast is a great way to go.[/QUOTE]

I'd looked at doing something similar with an Azure GPU instance (Tesla graphics cards) but they didn't show up in my current subscription. The NCv3 instances with a Tesla v100 are apparently only available on request.

The NC6s_v3 is currently showing up at $1.53/hr and if it took something like 30 hours (am I even close? No idea...) then we're talking $46 USD for a rapid verification run.

I sure thought that'd be a cool use of my $200 credit on my personal account but oh well. Maybe next time.