Primenet web design

[Prime95]

Originally Posted by cxc

Hi again James, I’d like to take the opportunity to thank you

Hear, hear. James is an absolute wizard at writing elegant PHP code. The functionality PrimeNet's web pages has grown immensely since James took over.

[cxc]

I’ve had a bit more of a poke around the exponent report page to see if there are any unwanted consequences of allowing the Fermat exponents to be reported – and I found one bug.

The main negative consequence is that reporting on any range of Mersenne exponents (e.g., 4000 to 4100) which includes a small Fermat exponent (e.g. F12, exponent 4096) suddenly gets a lot slower to return data, because of the quantity of ECM data for those ones. It’s not as drastic for the ones larger than F15, but I can appreciate why the display of the Fermats might have been deliberately suppressed in the past. If there are complaints about this, one solution might be to have another tick box (like the ‘Show full details’ box) which excludes the Fermats, set off by default so that you have to tick the box to enable them to display. My apologies in advance!

As for the bug, I’m sorry to report the code James has concocted to pick up the Fermat exponents and render them as F12, F13, etc. also picks up F0 = M2, the first Mersenne prime, seeing as the exponent is a power of 2, but incorrectly displays it as ‘Fermat F1’ because of the –1 / +1 difference between the form of Mersennes and Fermats :)

https://www.mersenne.org/report_expo...=1&swversion=1

[James Heinrich]

Originally Posted by cxc

... but I can appreciate why the display of the Fermats might have been deliberately suppressed in the past.

To be clear, there is no deliberate "suppression" of data. It's just that the whole system (mersenne.ca as well) is generally based on the assumption that k=1,b=2,n=<prime>,c=-1 and things like Fermat numbers which don't fit this pattern need special effort to be included.

Originally Posted by cxc

also picks up F0 = M2, the first Mersenne prime, seeing as the exponent is a power of 2, but incorrectly displays it as ‘Fermat F1’

Thanks, fixed.

[James Heinrich]

Originally Posted by cxc

Herewith

Thanks. We're still working on how we want to store Fermat factors in the database (some, but not all, of them are in the current data). For now I've added a hardcoded list for F12..F29 so they should display as expected on the exponent pages.

Originally Posted by cxc

The main negative consequence is that reporting on any range of Mersenne exponents (e.g., 4000 to 4100) which includes a small Fermat exponent (e.g. F12, exponent 4096) suddenly gets a lot slower to return data, because of the quantity of ECM data for those ones.

There seems to be a large amount of unnecessary duplication of F-ECM entries for some Fermat numbers (I quick-counted more than 10k duplicates in F12 alone), George will have to decide if those are true duplicate entries and do some appropriate pruning. I have added some display-suppression to make things a little faster in the interim, but it needs to be properly fixed at the database end sometime.

[cxc]

Originally Posted by James Heinrich

To be clear, there is no deliberate "suppression" of data. It's just that the whole system (mersenne.ca as well) is generally based on the assumption that k=1,b=2,n=<prime>,c=-1 and things like Fermat numbers which don't fit this pattern need special effort to be included.

All the same, I could imagine with the system being geared one way, there could have been little enthusiasm for making room for the exceptions, so I’m grateful this is not the case and you have taken the time to look into this.

F12 appears to be the nuisance case for F-ECM results, over a quarter of a million curves have been run and 10k of those involved a successful recovery of one or other known prime factors or composites. Some factors have turned up thousands of times in the data, others very rarely:

2017-05-02;Yxinity;F-ECM;Factor: 108197010425778846616814891389230502649564783828829791078210812084288831489 / (ECM curve 1, B1=800000000, B2=80000000000) [composite factor = p₁·p₂·p₃·p₆]
   43 curves: F-ECM;Factor: 45477879701734570611058964078361695337745924097 [composite factor = p₁·p₂·p₃·p₄·p₅]
   10 curves: F-ECM;Factor: 396532184444319600058061052745788134326273 [composite factor = p₂·p₃·p₄·p₅]
    5 curves: F-ECM;Factor: 1747953014374992168284948845521282121729 [composite factor = p₁·p₃·p₄·p₅]
  136 curves: F-ECM;Factor: 239012339910309306759513718485221377 [composite factor = p₁·p₂·p₃·p₅]
    2 curves: F-ECM;Factor: 6218500374143300163916104838414337 [composite factor = p₂·p₄·p₅]
 3166 curves: F-ECM;Factor: 36204694129087842739610650509313 [composite factor = p₁·p₂·p₃·p₄]
   18 curves: F-ECM;Factor: 27411763534674941587065980108801 [composite factor = p₁·p₄·p₅]
   13 curves: F-ECM;Factor: 2084004044941618697168113057793 [composite factor = p₂·p₃·p₅]
   29 curves: F-ECM;Factor: 315677127964214900640956417 [composite factor = p₂·p₃·p₄]
 1429 curves: F-ECM;Factor: 239009526063309834309009409 [composite factor = p₄·p₅]
   11 curves: F-ECM;Factor: 1391535943463300009328641 [composite factor = p₁·p₃·p₄]
    9 curves: F-ECM;Factor: 32681785846327290249217 [composite factor = p₂·p₅]
 2449 curves: F-ECM;Factor: 190276431449381650433 [composite factor = p₁·p₂·p₃]
   17 curves: F-ECM;Factor: 144064538330727383041 [composite factor = p₁·p₅]
    2 curves: F-ECM;Factor: 4950514524072886273 [composite factor = p₂·p₄]
   17 curves: F-ECM;Factor: 21822356733001729 [composite factor = p₁·p₄]
   13 curves: F-ECM;Factor: 1659064351850497 [composite factor = p₂·p₃]
 3086 curves: F-ECM;Factor: 1256132134125569 [prime factor p₅]
    2 curves: F-ECM;Factor: 7313319444481 [composite factor = p₁·p₃]
   46 curves: F-ECM;Factor: 190274191361 [prime factor p₄]

And the NF-ECM data would be even more voluminous and less useful. The option for “Include full ECM history (every "NF-ECM" result, slow)” in practice grinds to a complete halt for each of the small Fermats F12 to F15.

[Andrew Usher]

Well, that shows why Primenet must have the known factors - one doesn't want a human to have to look at each of those results 'recovering' known factors to check if they're really new - it seems there is no automatic mechanism for adding new factors. With only four having been discovered this century, there would not be much compelling reason to do so. (I would have calculated the list of factors myself if it had been necessary).

To the reply by James - I don't exactly like making assumptions, but with no access to the code I necessarily will have to. In this case it seemed pretty safe, as also the acceptance of P+1 without apparent issue shows that P-1 should be equally possible (I'm not sure about PRP and TF; the former would be nice). The P-1 is, of course, a particular issue with me, but should not be dismissed on that ground.

[James Heinrich]

Originally Posted by Andrew Usher

shows that P-1 should be equally possible (I'm not sure about PRP and TF; the former would be nice). The P-1 is, of course, a particular issue with me

If it's important to you, please provide a P-1 result line that you tried to submit but failed to be accepted and I can look into it further.

[Madpoo]

Originally Posted by Andrew Usher

Well, that shows why Primenet must have the known factors - one doesn't want a human to have to look at each of those results 'recovering' known factors to check if they're really new - it seems there is no automatic mechanism for adding new factors. With only four having been discovered this century, there would not be much compelling reason to do so. (I would have calculated the list of factors myself if it had been necessary).

To the reply by James - I don't exactly like making assumptions, but with no access to the code I necessarily will have to. In this case it seemed pretty safe, as also the acceptance of P+1 without apparent issue shows that P-1 should be equally possible (I'm not sure about PRP and TF; the former would be nice). The P-1 is, of course, a particular issue with me, but should not be dismissed on that ground.

I don't think that's why there are so many duplicates. I did a little looking at "factor found" results on F12 last night and of the 10K+ total results like that (reporting a factor), 5138 of them were from a single user finding the same handful of factors over and over again. This user clearly was in a position to know about existing factors yet kept churning them out time and time again. Another user turned in 1469 of those, another did 898, another with 795 etc.

To me that reeks of user error, much more so than the ones who have turned in the same factors less than 10 times each (or certainly the 12 wonderful users who only reported one factor for F12).

For me (and I don't speak for George or James), mersenne.org is about Mersenne primes, so while Fermat work is interesting to some, I never spent any time personally on doing any improvements related to that on the website. One, I don't understand it as well, and two it seemed like scope creep. But if George and/or James wanted to include anything more about those, I'll pitch in, it just wasn't my own priority. LOL

[kriesel]

Originally Posted by Madpoo

For me ... mersenne.org is about Mersenne primes, so while Fermat work is interesting to some, ... it seemed like scope creep.

(Bold added) Amen to that. And mersenneforum.org has what seems to me a greater degree of scope creep. Some breadth is useful to the mission of finding Mersenne primes; efficiently eliminating composite factor candidates as part of TF code for example.

[Andrew Usher]

(I will here have to repeat statements I have made in my P-1 thread: https://mersenneforum.org/showthread.php?t=28281.)

First, I have no such P-1 result; I never submitted any (as I thought at the time of no reason to do so) and after seeing Batalov's work going much farther than mine ever would, stopped all such effort and deleted all results obtained, so as not to be tempted into using any more cycles on it. I never was concerned specifically with my own P-1, anyway (regardless of its merit), but with presenting the most useful results for everyone, to which may now be added: presenting the results honestly. We now have pages saying that P+1 but not P-1 has been performed on Fermats, which is clearly untrue. That as well as the direct testimony from Jan S tells me that it is not being accepted (note: after the updates in the new post https://mersenneforum.org/showpost.p...3&postcount=16 PRP is no longer of any especial relevance, only P-1).

'User error' is something we have to deal with; some people will submit results not including (all) known factors, regardless of the kind of number and work. It isn't, as admitted, the main reason for a clean-up, and is not normally cleaned up on Mersennes (though I'd have no problem with this being done).

As for 'scope creep' in general: that horse has already left the barn. Adding Fermats at all was an example of it, yes, though as cxc pointed out in the other thread, there are only 18 (or 19, if F30 is added) and that's very little to the size of the database. Adding anything beyond the original mission of GIMPS to simply classify Mersennes as prime or not was even more, and yet it happened, much of that codified by the change to v5, 15 years ago. This really hurts no one, much as 'scope creep' of the forum, which can indeed be compared to it, and also happened for similar reasons. These are also driven directly by user demand.

[chalsall]

Originally Posted by Andrew Usher

As for 'scope creep' in general: that horse has already left the barn.

Thanks for that reminder.

One of the first things I have my people read is Brooks' The_Mythical_Man-Month. It just makes comms amongst geeks faster.

Please ask if there are any questions.