Riesel base 3 reservations/statuses/primes

[gd_barnes]

On Jan. 12th, Micha reported completion of k=550M-560M to n=25K. k's remaining are now shown on the web pages.

[jasong]

I have just returned to this thread and read what is written. I'm not going to try to start anything up, although I'd like to thank KEP and XYZZY for trying to help.

I would like to reserve k=800M-900M, test them all to n=2000 using some scripts I've found in the forum, and then try to use sed and some scripts along with srsieve and LLR to take them on n=25,000. Unless I'm confused about what I've read, it will be approximately a year or so until all the small primes are found, so I'd like to set a date of 3/19/2009 for it to be unreserved if I become uncommunicative or you are not able to contact me. I do not intend to do either of those two things, I'm just trying to do things that won't make people unhappy. My crunching computer(a quad-core) is temporarily out of service(I think it has to do with the master/slave setting on the hard drive) so gdbarnes thinking it over for a few hours(even if it's been most of a day before you see this) won't prevent me from having anything to crunch.

I apologize for my behavior, although if one reviews our past correspondence, I'm sure it can be seen how I became offended. (This would involve a good deal of time passing between the time things started and ended, and I'm assuming that people simply want to move on)

Thank you.

[gd_barnes]

Originally Posted by jasong

Unless I'm confused about what I've read, it will be approximately a year or so until all the small primes are found,

By small primes, I take that to mean the primes n<=2000. Why would it take a year to find all the primes n<=2000 for k=800M-900M? I could probably do that in < 2 CPU days running one core. For that matter, why would it take a year to do k=800M-900M all the way to n=25000?

I have a better idea about reserving it: I won't reserve it now but you can go ahead and start it. After you've started, please let me know. I'll then tentatively reserve it for you to n=2000 only. Running one core, that should take only a few days or a week.

If you are able to complete it to n=2000 in < 1 week, then I'll reserve it to n=25000 for you.

Being non-communicative will not work. Please provide a status < every 2 weeks.

And finally: I still don't understand this insistance on reserving HUGE chunks of work. It is not an insult to one's intelligence to do things in smaller pieces until one has a full understanding of it. At one time, none of us understood any of this stuff. I would respect your efforts much more if they were done in small pieces. k=800M-810M would be an excellent size range for you to start with. That would be far more preferrable.


Gary

[jasong]

Originally Posted by gd_barnes

By small primes, I take that to mean the primes n<=2000. Why would it take a year to find all the primes n<=2000 for k=800M-900M? I could probably do that in < 2 CPU days running one core. For that matter, why would it take a year to do k=800M-900M all the way to n=25000?

I meant that with the Riesel base-3 k being as high as it is, that finding all the small primes for all the ks under the Riesel k, which is around 3 billion I think would probably take a while.

I have a better idea about reserving it: I won't reserve it now but you can go ahead and start it. After you've started, please let me know. I'll then tentatively reserve it for you to n=2000 only. Running one core, that should take only a few days or a week.

If you are able to complete it to n=2000 in < 1 week, then I'll reserve it to n=25000 for you.

Do I submit the file to you, or am I on the honor system? I'm assuming the first one, but I'm not sure.

Being non-communicative will not work. Please provide a status < every 2 weeks.

That wasn't something I planned on doing, I just figured I should deal with people's concerns by offering a termination condition.

And finally: I still don't understand this insistance on reserving HUGE chunks of work. It is not an insult to one's intelligence to do things in smaller pieces until one has a full understanding of it. At one time, none of us understood any of this stuff. I would respect your efforts much more if they were done in small pieces. k=800M-810M would be an excellent size range for you to start with. That would be far more preferrable.

If I can manage to make the proper scripts, then there will be significantly less processing time involved. If I fail to make the scripts, then I will follow your wishes for reservations from then on. :)

On my quad-core, going to n=2000 takes a little over a day. This leaves approximately 1000 unprimed ks(I haven't actually counted, I could be way off). I think a system of sieving, LLRing the file until the n-value is about 25% higher than when started, finding primes with the script, noting the primes in the LLR file, recording them in the proper fashion, purging the numbers from the LLR file, and repeating the process, especially with the aid of scripts, would greatly speed things up. It sounds complicated, but if I can manage to automate it, the final result should take mere seconds to perform.

See, all the output from LLR is in a specific form. It's made to be human readable, but all the important lines contain the words "probable prime". Because of the scripts I found in the CRU forum, I'll have all the unprimed ks in one file in standard form. So the script finds the probable prime line, edits out everything but the equation with the found n-value, makes sure the k value is in the unprimed list, sends it it to a file for the new primes and deletes the equation from the file that represents the unprimed ks. Obviously, I'll have a backup in case there's a screwup, but in the end, we'll have a file with all the k values that don't have primes for n<=2000, but do have primes for 2000<n<=25000. And we'll also have a file for the numbers that still need to have primes found for n>25000.

I know it sounds like a clusterf***, but if you've ever tried to write an essay on how to tie your shoes, prepare toast, or anything similar, you know a simple task can seem horrendously complicated. It's going to take me a good amount of time to figure things out, but it's simply a matter of documenting what's needed to be done and following through. There are probably a hundred or so individual things I need to accomplish, but when broken down properly, none of the individual tasks is beyond me.

[gd_barnes]

Originally Posted by jasong

I
Do I submit the file to you, or am I on the honor system? I'm assuming the first one, but I'm not sure.

Please submit it to me just this 1st time. If you get through it, you shouldn't have to submit it for future reservations. But...please don't send all the primes...too large. Just send n=500-2000. I can quickly run n=1-500 myself to make sure that everything is in order.

On my quad-core, going to n=2000 takes a little over a day. This leaves approximately 1000 unprimed ks(I haven't actually counted, I could be way off). I think a system of sieving, LLRing the file until the n-value is about 25% higher than when started, finding primes with the script, noting the primes in the LLR file, recording them in the proper fashion, purging the numbers from the LLR file, and repeating the process, especially with the aid of scripts, would greatly speed things up. It sounds complicated, but if I can manage to automate it, the final result should take mere seconds to perform.

Karsten has already done it. I haven't used it yet but Micha, Chris, and some others have. Karsten, Micha, or Chris, if you see this, is there enough documentation in the automated process that Karsten came up with that you could provide it here for Jason?

Jason, if the primes up to n=2000 look OK, I'll give you another shot here. Please just keep us posted on where you are at on a regular basis.


Gary

[kar_bon]

Originally Posted by gd_barnes

Karsten has already done it. I haven't used it yet but Micha, Chris, and some others have. Karsten, Micha, or Chris, if you see this, is there enough documentation in the automated process that Karsten came up with that you could provide it here for Jason?

i'm just working on some new enhancements of the two scripts i gave in this forum here

- combine the 2 scripts with the sieving for the LLR-part, so only one call will do the whole work (with small parameters to give)
- the LLR-part contains now a file with all ramainig k-values, so it's easier to report them
- including also the reduction for multiple of base (MOB)

this will done the next days, after testing for some cases.
i will post it then in the above thread.

[michaf]

Originally Posted by kar_bon

i'm just working on some new enhancements of the two scripts i gave in this forum here

- combine the 2 scripts with the sieving for the LLR-part, so only one call will do the whole work (with small parameters to give)
- the LLR-part contains now a file with all ramainig k-values, so it's easier to report them
- including also the reduction for multiple of base (MOB)

this will done the next days, after testing for some cases.
i will post it then in the above thread.

Oohh... exciting :)

Jason, I found that the most 'work' is the 'gathering' of results. Getting them in order, sorting out everything, those kind of things...
Small chunks make that way easier (my large chunk of 80M-size is divided in 9 manageable chunks of 10M in size).

Karstens scripts would greatly simplify all this, especially if sieving & prp'ing is combined :)

[Flatlander]

Originally Posted by michaf

Oohh... exciting :)

Jason, I found that the most 'work' is the 'gathering' of results. Getting them in order, sorting out everything, those kind of things...
Small chunks make that way easier (my large chunk of 80M-size is divided in 9 manageable chunks of 10M in size).

Karstens scripts would greatly simplify all this, especially if sieving & prp'ing is combined :)

Yes, we could do with some simplifying for this conjecture.
I have to be in the right mood to reserve a range here.

[jasong]

As far as splitting up the files are concerned, I can make them as small or as large as people want. It would simply be a matter of altering one of the files parameters. Of course, in my case, it would be a kludgy sort of method, since I'd simply make multiple copies of one of the files when I change the parameters.

Honestly, the reason I chose the portion of the project that I did is that I have a ton of free time and only one meatspace friend. If I'm more useful someplace else, I'm willing to go there. I'm sorry if I seem inconsistent, I felt like I needed to prove something a few days ago. Now that I have a second chance, I'm wanting to be more agreeable. If I actually start my assigned range with the n=500-2000 restriction, I'll post here, but I still need to get my quad-core working, so feel free to make a change if you catch me in time. :)

My quad-core is still being really weird, so I don't have anything to crunch on at this moment. I'm going to make a couple more attempts to get it to work, then maybe give up and go back to Windows XP if I can't get it working properly.

Edit: I'm installing Windows XP back on it. I'll futz with Linux some other time.

[gd_barnes]

So when do you actually plan to START the range, Jason? If your machine is not working properly, do you really want to work on this? I have big reservations about you doing it wondering if your machine will produce false primes, false composites, bad residues, etc.

To be clear: You'll need to run the primes for n=1-2000 but send me the primes for n=500-2000.

Once you start on the range, I'll reserve it for you up to n=2000. When I get the primes for that range verified, I'll reserve it for you up to n=25000 as long as you check in with us < every 2 weeks.

If you are only doing this to prove something to us, that is a poor reason. You have to face the man in the mirror. You have to do it because you like computers and you like math: There should be no other reason.

If you want to impress me, just fully complete the much smaller k=800M-810M in a timely and correct manner (like what Michaf is doing) and I'll gladly be more willing to reserve ranges for you in the future. Doing k=800M-900M is not necessary at all.

There were two things that I did not say in response to KEP's post earlier:

First KEP, your post was an outstanding one and had many great points. The world can use more compassion towards those with certain disorders. We all need to be more tolerant of our differences. Lack of tolerance towards people different from us is what starts many wars.

Second, contrary to the first point, about the worst place to expect compassion or tolerance from your fellow man is in a math forum. Many of us here are number crunchers, math geeks, and/or computer geeks. We know brute logic, enjoy the intellectual process of solving problems in math or with computers, don't show a soft side often, and I know of know females on the forum (likely for about all of these reasons). I don't want to over-generalize as there are many great helpful people here but...you can't expect most math/computer folks to feel sorry for people who have a lot of difficulty with what it is we are doing after making several attempts to help.

Jason, I believe that is why you are experiencing everyone being an _hole to you here. I want to believe you can do what you reserve but after 2-3 attempts, it's easier for me as an admin to just say: Do something else or check with others. That's because I simply don't have the time to spell out every detail and is why I referenced you to Karsten/Michaf/Chris here. But if one day, it sounds like you want to do this huge range and the next, you haven't even started because your computer is acting up, then what am I supposed to believe?

If all of this seems cold, that is the way it is in a math forum. There are plenty of "friendlier" sites out there but IMHO, this is the best laid out math forum out there so it is where I chose to find prime numbers.

Edit: I've already said no less than 3 times where else at CRUS you would be more useful. Try reading my prior posts please. I will not keep repeating myself.


Gary

[KEP]

Originally Posted by gd_barnes

First KEP, your post was an outstanding one and had many great points. The world can use more compassion towards those with certain disorders. We all need to be more tolerant of our differences. Lack of tolerance towards people different from us is what starts many wars.

Thanks for the kind words on my post. I guess you're right on the war remark, because most wars is caused by lack of respect towards other cultures or political standings and eventually the disrespect turns to anger and anger turns to war :(

Originally Posted by gd_barnes

Second, contrary to the first point, about the worst place to expect compassion or tolerance from your fellow man is in a math forum. Many of us here are number crunchers, math geeks, and/or computer geeks. We know brute logic, enjoy the intellectual process of solving problems in math or with computers, don't show a soft side often, and I know of know females on the forum (likely for about all of these reasons). I don't want to over-generalize as there are many great helpful people here but...you can't expect most math/computer folks to feel sorry for people who have a lot of difficulty with what it is we are doing after making several attempts to help.

This is exactly why Jasong has to take this advice: So Jasong, to avoid using up other peoples patience, you have to start coming up with a likely plan on how to tackle your reservation.

Once I started supporting CRUS, I really had no idea how long things took and even less did I understand how to do things. However I used a day or 2 to do some testings in the beginning, to find out how to do things. Now almost nearing my 1 year anniversary on this project, I have gotten quite an amount of experience, and this allow me to tell you how I do things now:

1. Consider the size of my reservation
2. Do some initial testings before reserving, to get a grasp on how long it will actually take me to complete the given reservation
3. Checks back occationally via e-mail or forum message and tells how far I've gotten and when I expect to have en ETA on my reservation

Regarding no. 3 it can be difficult to say since one can be forced to switch to slower computers during processing of the reservation or simpel malfunctions can make one suffer several setbacks :( (However it gives a good starting point when it comes to tell how long things is actually going to take)

Now finally I should say in support of Gary, that testing Riesel base 3 is in fact one of the more harder tasks that is availeable at CRUS. So Jasong the best advice regarding this base to you from me, is to in fact do following:

1. Reserve work no earlier than you're able to crunch it
2. Give some statements that actually tells that you're understanding what you're doing and that it is actually reasonable to expect a completion
3. Present facts and not guesses, since it causes confusion if one states different things from time to time (have done that myself earlier, sorry Gary ) or it causes one to seem unwilling or unable to complete what he/she is actually reserving
4. Have fun and plan for setbacks (e.g. power outtages or sudden reboots etc)

A good ETA Jasong for a 100M range to n=25K for Riesel base 3 is actually 2 months (with setbacks) and 1 month if you're lucky and suffers no setbacks.

Good luck on your task Jasong, I believe that you can actually do it, if you just don't give up or loose focus

Take care everyone.

Kenneth!