mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > Riesel Prime Search

Reply
 
Thread Tools
Old 2007-09-18, 21:19   #23
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

241458 Posts
Default

Quote:
Originally Posted by Kosmaj View Post
Gary

I'd like to thank you for double checking the 6 k's I mentioned.

As far as k<300 and n<100,000 are concerned, they have already been double checked on a professional, error-free hardware, and some k/ranges have been checked 3 or more times. You are clearly wasting you time there.
As of late last night, your 6 k's are up to n=281K with no errors. Of course only k=109 is running now up to n=350K. It should be done Wed. or Thurs.

OK, great for k < 300 and n < 100K. I knew it had been double-checked but didn't realize the extent to which it had been done. I've caught 1-2 errors of my own that had already been double-checked by doing a 'triple-check' so I figured it was worth a shot. I won't go any further with it.

Would you like me to double-check some k's < 300 with unusually large prime gaps at some point? Ones specifically that I saw (besides the infamous k=289 that I already checked) are for k=261 from n=185283 to 618918 and k=297 from n=82782 to 225274. (I only looked at k > 250 for large prime gaps.) If we're unsure about any prior testing range on those, I'd be glad to add whatever ranges are in doubt to my double-checking task list in this thread.

One more thing...do the double-checked ranges for the 6 k's that I did need to be reported to Prof. Keller? If he needs a results file, I can send it also.


Gary
gd_barnes is online now   Reply With Quote
Old 2007-09-18, 21:47   #24
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

33·383 Posts
Default

Quote:
Originally Posted by Anonymous View Post
Alas, my resources are very small (essentially consisting of one P4 3.2Ghz HT). However, given that the ranges to test are small enough, I might be able to do some LLR doublecheck testing for you. How big would the ranges be? I would prefer if they could be small enough so that they don't take more than a week apiece to test. My computer is on for, I would guess, an average of 8-9 hours a day--and I know from working on the k=105 search here that n's around 650,000 take about an hour apiece to test , so if based on those calculations you could come up with a range that would take approximately a week for me to process, then I might be able to do some.

If you'd like to get some more resources for doublechecking than my drop-in-the-bucket resources, then you could start a reservation thread here for doublechecking--making it clear at the beginning of the thread that this is only for non-overclocked systems and that in mostly, there won't be much of the glory of finding new primes--and I'm sure you'll attract a bunch of users who don't care as much about finding primes to be attached to their name and care more about what's in the best interests of the project. It worked over at Riesel Sieve--for a while they ran a separate LLRNet server for doublechecking, and from what I can tell there were a bunch of people with decently large resources crunching for it. (The doublechecking LLRNet server has now been replaced with an LLR component to their BOINC setup [although the BOINC part of the project is diverted from doublechecking work to assault a particular stubborn k at the moment].)

P.S.: You could always set up an LLRNet server for doublechecking. Since none of the first-pass projects here have an LLRNet server set up, you might be able to get some users that way who would rather a set-it-and-forget-it solution for, say, large farms that are too much work to maintain manually. Of course, overclockers would have to be left out, but I'm sure that there's plenty of users who don't overclock and would be glad to do doublechecking work simply because LLRNet could take care of it automatically for them.
Anon,

I'm not ignoring your post here. There's a lot in it and I need to 'play around' on the internet to attempt to understand some things. I'm still a relative 'newbie' at this (4 months). I have only a cursory understanding of what LLRnet and BOINC are. They sound like good ideas for a double-check effort.

As far as what ranges I could give you on a double-check effort, it would be for n=50K to 100K for as many k's as you could reasonably stand! Like I did for n <= 50K, I would be sieving the entire range of 300 < k <= 1001; 351 k's in all. It would be entirely up to the weight of the k's as to how long each k would take to LLR. So what I would probably do is give you an average testing time per candidate on a 3GH P4. Then you tell me about how many candidates that you would be willing to test. Then I'll give you an appropriate # of k's that have about that # of candidates.

If you don't mind me asking...How come you only run your machine 8-9 hours/day? You should be able to do most things on it and run both cores at the same time. On my main machine that I have most of my personal stuff on, if I need to run another program or process that needs some serious CPU cycles, I'll temporarily stop one of the cores from LLRing and then restart it after I'm done. All of my machines run 24 hours non-stop on either LLRing or sieving unless they finish in the middle of the night or while I'm at work, which they seem fond of doing! (Which is why I now give them much larger chunks of work to do at once.) If it's a home machine and you're gone during the day, the main key is to keep your place < then about 80 degrees. Just a thought...

I'll get back with you a little more on your thoughts on double-checking later on.


Gary
gd_barnes is online now   Reply With Quote
Old 2007-09-18, 23:10   #25
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

186916 Posts
Default

Note: I've split your post into multiple quote sections for easier reading.
Quote:
Originally Posted by gd_barnes View Post
Anon,

I'm not ignoring your post here. There's a lot in it and I need to 'play around' on the internet to attempt to understand some things. I'm still a relative 'newbie' at this (4 months). I have only a cursory understanding of what LLRnet and BOINC are. They sound like good ideas for a double-check effort.
No problem. I'm not much of an expert either; all that I know about prime search projects was gained from the message boards of this and the various other projects that I participate in, mostly over the course of the last couple of months. So you probably know more than me.

Quote:
Originally Posted by gd_barnes View Post
As far as what ranges I could give you on a double-check effort, it would be for n=50K to 100K for as many k's as you could reasonably stand! Like I did for n <= 50K, I would be sieving the entire range of 300 < k <= 1001; 351 k's in all. It would be entirely up to the weight of the k's as to how long each k would take to LLR. So what I would probably do is give you an average testing time per candidate on a 3GH P4. Then you tell me about how many candidates that you would be willing to test. Then I'll give you an appropriate # of k's that have about that # of candidates.
Okay, I see now. What I was meaning to say was that so that I don't accidentally overload myself, if I could do work in blocks of about 1 week's worth at a time, if that's not a problem. Sort of like how it's done with a distributed effort--get a range of x amount of numbers to test, do them, report them and get more. I generally don't like to get too much work piled up on me at a time, even if there are no "deadlines" to speak of--it's just a little more flexible for me if I do stuff in smaller chunks.

Quote:
Originally Posted by gd_barnes View Post
If you don't mind me asking...How come you only run your machine 8-9 hours/day? You should be able to do most things on it and run both cores at the same time. On my main machine that I have most of my personal stuff on, if I need to run another program or process that needs some serious CPU cycles, I'll temporarily stop one of the cores from LLRing and then restart it after I'm done. All of my machines run 24 hours non-stop on either LLRing or sieving unless they finish in the middle of the night or while I'm at work, which they seem fond of doing! (Which is why I now give them much larger chunks of work to do at once.) If it's a home machine and you're gone during the day, the main key is to keep your place < then about 80 degrees. Just a thought...
No, I don't mind you asking. But I think you misunderstood me when I said that I run my machine for 8-9 hours a day. I meant that my machine is only switched on for 8-9 hours during the day, with distributed computing projects using all the idle CPU time throughout that, while I use the computer too since they automatically back out of the way for as much or as little as necessary since they're set to idle priority level by default. Why don't I leave my computer on 24/7, you may ask? Because the electricity bills would be outrageous. The main reason why I run distributed computing on my computer is to put its idle CPU time to good use, not for it to become an expensive hobby. But yes, I do have distributed computing apps running on both cores whenever the computer is on--at the time, the configuration is BOINC set to use one core, and LLR doing k=105 work on the other. The doublechecking work would take the place of the k=105 work (I am almost done with my current range).

Quote:
Originally Posted by gd_barnes View Post
I'll get back with you a little more on your thoughts on double-checking later on.

Gary

Last fiddled with by mdettweiler on 2007-09-18 at 23:13 Reason: fixed some typos
mdettweiler is offline   Reply With Quote
Old 2007-09-21, 00:08   #26
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

33·383 Posts
Default 6 k's done

Kosmaj,

The double-checking of the 6 k's that you requested has now completed. The following final prime was confirmed:
203 248862

No missing or incorrect primes were found. Let me know if I need to report it anywhere. I attempted to attach a 2.44 MB results file, but even zipped, it was around 570 KB; too big for the size restriction here.


Gary
gd_barnes is online now   Reply With Quote
Old 2007-09-27, 08:48   #27
Kosmaj
 
Kosmaj's Avatar
 
Nov 2003

2·1,811 Posts
Default

Hi Gary,

Please mail the details including Ks, ranges you tested, and primes you confirmed to Wilfrid Keller. You can find his address on this page. You can also mention the range of k=289 you tested before. There is no need to send any results. He is interested only in k<300. And don't forget to sign your message with your real, full name.
Kosmaj is offline   Reply With Quote
Old 2007-09-28, 06:23   #28
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

33×383 Posts
Default Note sent and upcoming sieving by Anon...

Quote:
Originally Posted by Kosmaj View Post
Hi Gary,

Please mail the details including Ks, ranges you tested, and primes you confirmed to Wilfrid Keller. You can find his address on this page. You can also mention the range of k=289 you tested before. There is no need to send any results. He is interested only in k<300. And don't forget to sign your message with your real, full name.

I sent the note to Mr. Keller. A funny or annoying side note on this depending on how you look at it...After finding no prime for k=289 from my original range of n=260K-520K (other than the 2 already found), I went back and double-checked the range of n=90K-260K. No luck...a remarkably large primeless range. But this turned out to be a good thing. In looking at Mr. Keller's site, I see that there were a few small gaps in checking of ranges for n=100K-200K. So I went ahead and reported that I had double-checked the entire range of n=90K-520K. So now all of his gaps are filled up to n=520K.

On another note...Anon has kindly agreed to do some sieving for the range of 300<k<=1001 for the range of n=50K-100K for a double-check effort. Although sieving shouldn't take too long...I'm guessing 2 weeks or so on a 24-hour high-speed machine, the LLR will be rather substantial. I was planning on holding off on this until next year but now may possibly attempt to get some of it done by the end of this year. I may enlist the help of a few others on it if anyone cares to assist.


Gary
gd_barnes is online now   Reply With Quote
Old 2007-10-03, 21:10   #29
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

33×383 Posts
Default Any more double-checking for k < 300 ?

Kosmaj,

It looks like Prof. Keller's ranges page for Riesels has been updated to include the ranges that I double-checked.

In looking at his page, I see that there are many ranges that have not been checked. Are there any that you're particularly concerned about? I know you said that there were many ranges that you and someone else had checked but had not sent to him. Let me know.

Since I'll be working on k=5 from n=470K-500K here shortly, I won't have the CPU resources to do much in the way of double-checking until that is done. But at that point, I will probably plan on splitting 2 core's time between LLRing 300 < k <= 1001 from n=50K-100K and anything you deem important for double-checking for k < 300.


Gary
gd_barnes is online now   Reply With Quote
Old 2007-10-25, 15:39   #30
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

33·383 Posts
Default Now checking k=301-1001 for n=50K-100K

After getting a sieve file from Anon (thanks Anon!), I have started doing some LLRing for a double-check of the range 300 < k <= 1001 for 50K < n <= 100K.

This is a large LLR effort and I will work on it on-and-off for a while splitting time with k=5 and another sieving effort that I'm working on.

I split it up in 100 k pieces. I'm currently working on k=300-400 and k=400-500, one 100 k piece on each core of a dual-core machine. After about 10 hours on both cores, k=301, 303, 401, and 403 are complete. There were a total of 6 primes in the range for those k's and no errors were found.

So I don't inundate everyone with status reports, from this point forward, I'll plan on reporting statuses and any errors found after completion of each 100 k piece.


Gary

Last fiddled with by gd_barnes on 2007-10-25 at 16:35
gd_barnes is online now   Reply With Quote
Old 2007-10-25, 17:09   #31
mdettweiler
A Sunny Moo
 
mdettweiler's Avatar
 
Aug 2007
USA (GMT-5)

141518 Posts
Default

Quote:
Originally Posted by gd_barnes View Post
After getting a sieve file from Anon (thanks Anon!), I have started doing some LLRing for a double-check of the range 300 < k <= 1001 for 50K < n <= 100K.

This is a large LLR effort and I will work on it on-and-off for a while splitting time with k=5 and another sieving effort that I'm working on.

I split it up in 100 k pieces. I'm currently working on k=300-400 and k=400-500, one 100 k piece on each core of a dual-core machine. After about 10 hours on both cores, k=301, 303, 401, and 403 are complete. There were a total of 6 primes in the range for those k's and no errors were found.

So I don't inundate everyone with status reports, from this point forward, I'll plan on reporting statuses and any errors found after completion of each 100 k piece.


Gary
Considering that this is such a big LLR effort, I'd be glad to do some of it--for starters a file of about 20,000 candidates should be good. If this would work out, you can send it to me whenever you're ready.
mdettweiler is offline   Reply With Quote
Old 2007-11-06, 17:57   #32
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

33·383 Posts
Default

Quote:
Originally Posted by Kosmaj View Post
3) Thomas Ritschel and I began double cheking such intervals of all k's tested in the 4th Drive and we completed all of them except the following:
k=109, n=100-350k (k=109 is low-weight)
k=203, 100-260k
k=205, 100-250k
k=215, 110-250k
k=217, 100-260k
k=257, 130-260k

We haven't reported our completed ranges to Keller yet, but we found no missing primes.
Kosmaj,

Per the above, did you have a chance to report your completed double-checked ranges to Prof. Keller yet? If not, can you let me know what they are?

If Prof. Keller would be OK with it and it would save you some time, I could report them for you if you let me know which ranges were searched by whom. I'll make sure that I specify that to him.

Quote:
4) I'll publish a similar list of k's from the 5th Drive soon, but you can begin your double checking with 6 k's above. Others are welcome to join, but for double-checking please your most stable machine preferably with no over-clocking (or with very conservative OC settings), and with no history of LLR errors of any kind. It will be nice to report your results to W. Keller using your full name.
Did you post a list of double-checked k's from the 5th drive yet? You may have but I wasn't clear on what it would entail and may not have known what it was.

I'm nearly half done LLRing my double-check for 300 < k <= 1001 for 50K < n <= 100K. (Several missing primes as expected will be posted on Wednesday.) After completing it, I would like to get more of the ranges filled in on Prof. Keller's site for k < 300 but I don't want to double-check what you or Thomas have already done.

I can start one core now on sieving some double-check ranges for k < 300 if you can let me know.


Thanks,
Gary
gd_barnes is online now   Reply With Quote
Old 2007-11-08, 17:37   #33
gd_barnes
 
gd_barnes's Avatar
 
May 2007
Kansas; USA

1034110 Posts
Default

Quote:
Originally Posted by gd_barnes View Post
Kosmaj,

Per the above, did you have a chance to report your completed double-checked ranges to Prof. Keller yet? If not, can you let me know what they are?

If Prof. Keller would be OK with it and it would save you some time, I could report them for you if you let me know which ranges were searched by whom. I'll make sure that I specify that to him.



Did you post a list of double-checked k's from the 5th drive yet? You may have but I wasn't clear on what it would entail and may not have known what it was.

I'm nearly half done LLRing my double-check for 300 < k <= 1001 for 50K < n <= 100K. (Several missing primes as expected will be posted on Wednesday.) After completing it, I would like to get more of the ranges filled in on Prof. Keller's site for k < 300 but I don't want to double-check what you or Thomas have already done.

I can start one core now on sieving some double-check ranges for k < 300 if you can let me know.


Thanks,
Gary

Bump
Bump
gd_barnes is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Double checking of Results pinhodecarlos Prime Gap Searches 13 2017-12-09 06:07
What about double-checking TF/P-1? 137ben PrimeNet 6 2012-03-13 04:01
Double checking Unregistered Information & Answers 19 2011-07-29 09:57
Double-checking milestone? jobhoti Math 17 2004-05-21 05:02
Any glory in double checking? Quacky Lounge 5 2003-12-03 02:20

All times are UTC. The time now is 23:54.

Sat Apr 17 23:54:11 UTC 2021 up 9 days, 18:35, 0 users, load averages: 2.47, 2.33, 2.03

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.