Testers needed for PPS LLR

Well, new server or not, we are going to attempt to implement the Proth Prime Search (PPS) project. It's time to get back into the fun of finding primes. However, we promise to save the server if it becomes too much. :D

Currently we are testing PPS LLR WU's. The form is k*2^n+1 for all odd k; 4<k<1200 and all n; 200K<n<5M. The first tests will be very quick as we are starting at n=200K. n will continue to increase, thus tests will slowly get longer.

To begin with, there may be many WU downloading errors. This is a result of the first batch of work that was released. They will eventually be processed and go away...but don't be surprised if you get some. :)

To help test, please select Proth Prime Search in your PrimeGrid preferences section.

Thank you for your help! And good luck finding those primes.

p.s. All primes found now are too low to enter The Largest Known Primes Database. We'll have to wait until n>340K (as of 1 September 2008) before the primes become large enough.

NOTE: With the short WU's and as more people select PPS LLR to test, you may see this message. It's just the server not quite keeping up with demand. Most likely you'll get work within the next two requests.

|PrimeGrid|Message from server: No work sent
|PrimeGrid|Message from server: No work is available for PPS LLR
|PrimeGrid|Message from server: No work available for the applications you have selected. Please check your settings on the web site.

Also, on fast machines you may even get this message (daily limit 800/core):

|PrimeGrid|Message from server: (reached daily quota of 800 results)

If this happens (1600 for duals & 3200 for quads), which it will on fast machines, please consider adding PPS Sieve (64 bit app only) or PSP Sieve (32 & 64 bit app) to your active projects.

UPDATE: Limit has been increased to 2000 WU's /core (total 4000 WU's for dual & 8000 WU's for quad)

Update: 2 Sep 2008 16:45 UTC

The generosity of testers is always very nice to see. We currently have plenty of testers. However, please feel free to still attach if you’d like to sample this project.

With the popularity of PPS LLR, we'd like to emphasize the need for sieving. PPS Sieve is available for those who are interested…especially those with 64 bit OS’s and dual, quad, and dual quad core machines. :D

If you wish to participate in both sides of PPS, please consider attaching your 32 bit OS machines to LLR and your 64 bit OS machines to Sieve.

We’d like to keep PPS LLR active at least up to n=500K initially. However, in order to do that, our main focus MUST still be on sieving. Those users with the appropriate resources, please consider the Sieve over LLR. If you wish to participate in the Sieve, see this thread.

Finally, for those PG users familiar with PG's LLRNet, we still need some help clearing PPS work up to n=200K. Currently we're at n=191K. It would be nice to finish up all work in a few days and complete this phase of the project. If you're interested, please see the Project Staging Area.

Thank you for your help!

Update: 6 Sep 2008 16:30 UTC

The testing is going well...so much so that we like to repeat the request from the previous update.

We'd like to emphasize the need for sieving. PPS Sieve is available for those who are interested…especially those with 64 bit OS’s and dual, quad, and dual quad core machines. :D

If you wish to participate in both sides of PPS, please consider attaching your 32 bit OS machines to LLR and your 64 bit OS machines to Sieve.

We’d like to keep PPS LLR active at least up to n=500K initially. However, in order to do that, our main focus MUST still be on sieving. Those users with the appropriate resources, please consider the Sieve over LLR. If you wish to participate in the Sieve, see this thread.

I've signed up!

Is the plan still to complete up to 200k via the non-BOINC LLRnet stuff that has been ongoing for a while? If so, I'll keep crunching that too. Otherwise, I'll kill that.

(Will there be BOINC credit and/or Prime Finder points for the non-BOINC work at some stage?)
____________

PPS LLR credits must adjust more to the values of other subprojects (32bit)

Planned to add this to the account page?

I've signed up!

Is the plan still to complete up to 200k via the non-BOINC LLRnet stuff that has been ongoing for a while? If so, I'll keep crunching that too. Otherwise, I'll kill that.

(Will there be BOINC credit and/or Prime Finder points for the non-BOINC work at some stage?)

I'll join PPS LLR once I've finished sieving for 321 on my PC :)
____________

I will carry on with port300 until it's complete.
____________

I will carry on with port300 until it's complete.

I've signed up, if the server can keep my computer supplied with work, I'll stay signed up, otherwise I'll head back to my sieving

On my 2.8GHz dual-core, a test takes 1:09.

I was having issues keeping work, so I bumped it to maximum.

Update

The generosity of testers is always very nice to see. We currently have plenty of testers. However, please feel free to still attach if you’d like to sample this project.

With the popularity of PPS LLR, we'd like to emphasize the need for sieving. PPS Sieve is available for those who are interested…especially those with 64 bit OS’s and dual, quad, and dual quad core machines. :D

If you wish to participate in both sides of PPS, please consider attaching your 32 bit OS machines to LLR and your 64 bit OS machines to Sieve.

We’d like to keep PPS LLR active at least up to n=500K initially. However, in order to do that, our main focus MUST still be on sieving. Those users with the appropriate resources, please consider the Sieve over LLR. If you wish to participate in the Sieve, see this thread.

Finally, for those PG users familiar with PG's LLRNet, we still need some help clearing PPS work up to n=200K. Currently we're at n=191K. It would be nice to finish up all work in a few days and complete this phase of the project. If you're interested, please see the Project Staging Area.

Thank you for your help!

Note: first post updated with this information
____________

With the most recent challenges, the app heated up systems dramatically. Is this a less-intensive app?

With the 24 hour deadline, BOINC goes into panic mode and reports the tasks as soon as they are done. Wouldn't it be better to set the deadline to 2 days, and get several tasks reported at the same time? I think this will stress the server a lot less, as explained here.

So we're finding primes already according to my account page.

You can tell what the next question is can't you? :-)

The primes found column in http://www.primegrid.com/stats_pps_llr.php doesn't list any, and the Top Prime Finders pages don't include any either.
____________

On the Server stats site under "primes found" no primes are displayed. on my user home page 4 primes found is displayed. Is that possible?

EDIT: lol 27 Seconds too late...

And what is this?
03.09.2008 12:01:15|PrimeGrid|Message from server: No work sent
03.09.2008 12:01:15|PrimeGrid|Message from server: No work is available for PPS LLR
03.09.2008 12:01:15|PrimeGrid|Message from server: (reached daily quota of 3200 results)

And what is this?
03.09.2008 12:01:15|PrimeGrid|Message from server: No work sent
03.09.2008 12:01:15|PrimeGrid|Message from server: No work is available for PPS LLR
03.09.2008 12:01:15|PrimeGrid|Message from server: (reached daily quota of 3200 results)

Would be better....
my Quad seems could crunch ~8600 of this WU´s per day ;-)

My quad is now out of work :-(
Other PG sub projects will give me no work too. So i switched to other projects.

My quad is now out of work :-(
Other PG sub projects will give me no work too. So i switched to other projects.

The primes found column in http://www.primegrid.com/stats_pps_llr.php doesn't list any, and the Top Prime Finders pages don't include any either.

I noticed that you seem to be releasing k one at a time, which is a great way to prevent the server from getting bogged down. What I don't know what the stats mean. Specifically, it lists total tasks, but I presume that is the total number of n to be tested for the k on the server.

I have some request on the stats (because I like numbers just like everyone else). Could this page show the min/max n for each k on the server? Is there a way to show how many n are not yet loaded onto the server for the corresponding k? Could the Total line at the bottom sum the Total Tasks and Done Tasks?

Also, it would be nice to see Max N on the Cullen and Woodall projects because it is clear (to me at least) that not all k values for the specified range are loaded onto the server.

Finally I hope that the primes and GFN divisors are reported to Proth Search or by e-mail to Wilfrid Keller and myself so that we can keep ProthSearch up to date. I suspect that you have found many new unreported primes.

I noticed that you seem to be releasing k one at a time, which is a great way to prevent the server from getting bogged down. What I don't know what the stats mean. Specifically, it lists total tasks, but I presume that is the total number of n to be tested for the k on the server.

I have some request on the stats (because I like numbers just like everyone else). Could this page show the min/max n for each k on the server? Is there a way to show how many n are not yet loaded onto the server for the corresponding k? Could the Total line at the bottom sum the Total Tasks and Done Tasks?

Also, it would be nice to see Max N on the Cullen and Woodall projects because it is clear (to me at least) that not all k values for the specified range are loaded onto the server.

Just had a batch of PPS LLR's error out with the following message:

<core_client_version>5.10.20</core_client_version>
<![CDATA[
<message>
- exit code -1073741502 (0xc0000142)
</message>
<stderr_txt>
BOINC LLR 5.09 wrapper: starting
Major OS version: 5; Minor OS version: 1
called boinc_finish

</stderr_txt>
]]>


Got some new ones and will report back if this continues...

____________
141941*2^4299438-1 is prime!

This web page shows that there were extensive research done for most of these numbers. Why are we redoing these ranges?
http://www.prothsearch.com/

Are we searching for missed ones?

Update

The testing is going well...so much so that we like to repeat the request from the previous update.

We'd like to emphasize the need for sieving. PPS Sieve is available for those who are interested…especially those with 64 bit OS’s and dual, quad, and dual quad core machines. :D

If you wish to participate in both sides of PPS, please consider attaching your 32 bit OS machines to LLR and your 64 bit OS machines to Sieve.

We’d like to keep PPS LLR active at least up to n=500K initially. However, in order to do that, our main focus MUST still be on sieving. Those users with the appropriate resources, please consider the Sieve over LLR. If you wish to participate in the Sieve, see this thread.

We now see prime counts on the server pages and in Top Prime Finders - thank you!

Also re: 2000WU/core daily limit - I assume this will drop back down to 800WU/core once the runtime of the PPS LLR WUs exceeds 108 seconds?

____________

Can anyone clarify what 'min remaining number' on the Server status page means with relation to the values of k/n?

Update
If you wish to participate in both sides of PPS, please consider attaching your 32 bit OS machines to LLR and your 64 bit OS machines to Sieve.

What to do for that?

• Lucas-Lehmer-Riesel test (WIKI)
  
• Download LLR by Jean Penné
  

Using Venues/Locations

You can use venues/location, i.e. home, work, school. For example, you can assign all your 64 bit computers to home and all your 32 bit computers to work. Then go to your PrimeGrid preferences page and select Edit PrimeGrid preferences.

Then choose the SIEVE projects you want for "Default computer location" home and select Update Preferences. Repeat process for LLR projects but this time choose a different "Default computer location".

How to assign Venues/Locations

Go to Your Account/Computers on this account and View your computers. Select each computer and assign it to a location: Home, Work, School. Do this for each computer you want to assign.


____________

Looks like the PPS LLR queue has started a second pass at k=5 again - presumably, this is n=220-240 or similar. However, it never gave out k=1051 or greater, which is confusing... What gives?
____________

Looks like the PPS LLR queue has started a second pass at k=5 again - presumably, this is n=220-240 or similar. However, it never gave out k=1051 or greater, which is confusing... What gives?

Right, also all green with 100%, after that yellow=in progress???

Thanks for the update John. I'm not very worried about this btw - was just caught by surprise.

Of course, my ideal would be to issue new WUs strictly by n value, though I suspect there may well be some performance/locality benefits by issuing by k value depending on the database schema (which is opaque to us!) :-)
____________

Thanks for the new stats format for PPS LLR. It took a little while to get used to the different colors, but I appreciate the additional information that it provides

Agreed - and it looks like the new WU allocation is by increasing n - which is perfect. Thanks!
____________

As noted on the front page, the first Proth Primes are starting to trickle in. Although many are still double check, this is an indication that we are reaching more ranges that have yet to be searched. The higher the n, the more open the ranges will be...and thus more "new" primes.

All prime finders will receive an email notification informing them of the "new" prime and instructions on how to report.

If you notice a new prime on your account but did not receive an email notification, that means it was a double check and has already been reported to the Prime Pages. Therefore, please do not report it.

We do not reveal any "new" primes until they are reported. Therefore, if you see the prime, it means it is reported.

Also, many have questioned why so many double checks? This effort is a complete double check of all the work that the project Proth Search has conducted. This means that we can be quite sure that all primes have been found for odd k<1200 up to our current n level. Additionally, all Fermat divisibilities are known for that range as well. There will be no need to double check these ranges in the future.

See this post for some additional information: http://www.primegrid.com/forum_thread.php?id=1103&nowrap=true#11439

Good Luck finding new primes!
____________

And proof that the question was timely!

2296 887*2^450283+1 135552 L654 2008

BTW, what happens when we reach n=500k ?
Do we have further ranges reserved?
____________

And proof that the question was timely!

2296 887*2^450283+1 135552 L654 2008

BTW, what happens when we reach n=500k ?
Do we have further ranges reserved?

But you who run 64bit OS stay on sieve work.


/Lennart

Yes, we will Lennart, accept during the PSP Sieve challenge next wednesday - sunday ;)

Btw: We are missing your extraordinary help on Proth Sieving the last weeks :(
Did you forgot to change your preferences on your herd of bit64's back from Proth LLR to Proth Sieve ;) ?

Cheers

And proof that the question was timely!

2296 887*2^450283+1 135552 L654 2008

2228 403*2^457660+1 137772 L153 2007

Hang in there! It's taken a while to "prime" the pump but the water is starting to flow. :D

In just the past 24 hours, 7 new primes were found (17 in the past week). We expect to see new primes on a daily basis now.

Hopefully this trend will continue.
____________

Yes-- the PPS is getting interesting-- it took 7000 WU's to get my first blood-- Now I've had a taste I am switching all my "old clunker's" over to PPS (see if we can reach that December target you wanted)---

ps- I have downloaded that Wubi-Ubantu you suggested-- havn't had time to set it up yet--

DoES

Yeah, just found another one...

I've mentioned this before on chat but is looks like better place.

We have "Other recent record primes". Then, we have Top Prime Finders.
Can we have time log of found primes? Can look like Top Prime Finders but sorted by date. Just an idea....
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

Yeah, just found another one...

I've mentioned this before on chat but is looks like better place.

We have "Other recent record primes". Then, we have Top Prime Finders.
Can we have time log of found primes? Can look like Top Prime Finders but sorted by date. Just an idea....

Yes-- the PPS is getting interesting-- it took 7000 WU's to get my first blood-- Now I've had a taste I am switching all my "old clunker's" over to PPS (see if we can reach that December target you wanted)---

ps- I have downloaded that Wubi-Ubantu you suggested-- havn't had time to set it up yet--

DoES

maybe you can start with this:
http://www.primestats.net/primegrid/charts.php?project=PPS_LLR&field=foundperday#stats

We have "Other recent record primes". Then, we have Top Prime Finders.
Can we have time log of found primes? Can look like Top Prime Finders but sorted by date. Just an idea....

Thanks for replay, John.
I can see that such an idea is already there and will be implemented in (hopefuly near) future...

And thanks for additional information, will try...
____________
My stats
Badge score: 1*1 + 5*1 + 8*3 + 9*11 + 10*1 + 11*1 + 12*3 = 186

Hang in there! It's taken a while to "prime" the pump but the water is starting to flow. :D

In just the past 24 hours, 7 new primes were found (17 in the past week). We expect to see new primes on a daily basis now.

Hopefully this trend will continue.

Not many reportable primes around 500K-- seems the last new one was a few days ago (187*2^497926+1)-- I did a recheck on 635*2^500689+1 yesterday-- Hopefully it will get better as we climb the 500K range further

DoES

Sadly, there will be a dry spell from 500K-540K as this range has already been searched. This is PrimeGrid's last "big" double check area. The good news is that beyond 540K, most of the ranges 300<k<1200 are unsearched. For 64<k<300, we need to get to n~800K to open up a lot of those areas. For 32<k<64, we will need to reach n~1M and for k<32 it is n>1.5M.

I was wondering why the same number is showing up as the "Min remaining number" (being 615*2^ 465547 +1) on the status page. It appears to have two tests done, both in December. Did their residues not match?

I was wondering why the same number is showing up as the "Min remaining number" (being 615*2^ 465547 +1) on the status page. It appears to have two tests done, both in December. Did their residues not match?

John, is it known what ranges are to be done automagically nuder BOINC and what on for example Beta test PRPNet?

Range stats for PPS LLR contain only BOINC ranges.
Any plan to merge all PPS LLR into single db?
Or even into primestats?

Sadly, there will be a dry spell from 500K-540K as this range has already been searched. This is PrimeGrid's last "big" double check area. The good news is that beyond 540K, most of the ranges 300<k<1200 are unsearched. For 64<k<300, we need to get to n~800K to open up a lot of those areas. For 32<k<64, we will need to reach n~1M and for k<32 it is n>1.5M.

Anyone want to comment on this PPS Number---

615*2^465547+1-- Oldest WU remaining---

Been sitting for 3 days now completed but unvalidated--
____________
Member of AtP

Shown here is an Australian native rat (Ratus Kickarsus)

Are there longer WUs now in the seed?
I have on my Core Duo T2600
two WUs with 21' 55" runtime instead the average of 13' 30"
On my Core 2 Duo T8300 i also have two WUs with 7' 52" instead of the average 5' 00".
Just curious...
I try to fiddle out the proper WUs in my WU-list and will report them back if needed.

Here are the links coming...

84270749
84268884
84272179
84272081
____________

[NB I haven't looked at this in any detail, but...]

Usually, when an LLR WU takes ~50% longer than previously (and there aren't other things causing this such as other tasks running on your machine), it is because the program has had to move up to the next size FFT band - and the cost in runtime is usually around 50%. For the PPS LLR WUs, you'll find this happens slightly earlier for the WUs with a higher k (e.g. 1000+) than for the lower values of k (e.g. 5), but will eventually spread downwards to cover all k. And then we wait for the next crossing of FFT size boundary and the runtime increases again...

Perhaps someone in the know could confirm if we actually are in the territory of a change in FFT size at the moment?
____________

Another one on the T2600 and three on the T8300.
I opt against the "other applications"-theory as my clients are not obliged to any other work besides boinc.
Only occasionally surfing - but not in the daytime when i am at work - they are so to speak "home alone" :)
So i think it is the FFT-Size
____________

Another one on the T2600 and three on the T8300.
I opt against the "other applications"-theory as my clients are not obliged to any other work besides boinc.
Only occasionally surfing - but not in the daytime when i am at work - they are so to speak "home alone" :)
So i think it is the FFT-Size

(...)
EDIT--A closer look shows the C2D is getting a few as well -- its just not claiming as much credit--

(...)
EDIT--A closer look shows the C2D is getting a few as well -- its just not claiming as much credit--

Yep, the 64-bit Core 2 Duos (or quads) are getting penalized a lot by pps llr.
My C2D 6400 doing pps sieve has a RAC of ~ 2350 wheras my C2D T8300 doing pps llr only achieves a RAC of ~ 1300, i switched my clients back to sieving, i am wasting clock-cycles... ;)

It's not that 32 bit PPS LLR is penalizing your machines. It's that 64 bit PPS Sieve is optimizing your machines.

A 64 bit LLR application does not exist...and it doesn't look likely in the near future...unless a genius comes along.

Therefore, your 64 bit machine uses the 32 bit LLR app...which is baseline for crediting.

When you have a box that produces an odd result -- You can put it down to Murphy (Law) -- But when you have 8 very different hosts (and other users) all showing the same thing-- Welllll!! you have to be curious--

I see that the WU lenght is generally steadily incteasing -- what I (and others) are seeing are anomilies in the rate of increase--

For the PPS LLR WUs, you'll find this happens slightly earlier for the WUs with a higher k (e.g. 1000+) than for the lower values of k (e.g. 5), but will eventually spread downwards to cover all k. And then we wait for the next crossing of FFT size boundary and the runtime increases again...

Vato is 100% correct. :) PPS LLR is approaching an FFT cross over.

All WU's that roadrunner_gs lists here have k's > 1165.

PPS LLR queue is empty.
And just as the WUs that had an estimated completion time that was compatible with TPS and the other LLR WUs were coming up :-(
____________

Queue is being refilled as I write this message.
____________

Queue is being refilled as I write this message.

Excellent - thanks Rytis!
____________

Congratulations!!!

All work for n<500K has been completed. This is a very nice milestone. Congratulations to everyone from sieving to LLRing.

I dare not speculate on when the next 500K will be completed (up to 1M). Common sense says it will take a while...but as more and more participants join the search, we may see it by the end of 2009! Yes, I said it, "end of 2009" or sooner. :D

Now that the seeds been planted, time to start watering. :D

Best of Luck everyone!!!
____________

Sadly, there will be a dry spell from 500K-540K as this range has already been searched. This is PrimeGrid's last "big" double check area. The good news is that beyond 540K, most of the ranges 300<k<1200 are unsearched. For 64<k<300, we need to get to n~800K to open up a lot of those areas. For 32<k<64, we will need to reach n~1M and for k<32 it is n>1.5M.

Sadly, there will be a dry spell from 500K-540K as this range has already been searched. This is PrimeGrid's last "big" double check area. The good news is that beyond 540K, most of the ranges 300<k<1200 are unsearched. For 64<k<300, we need to get to n~800K to open up a lot of those areas. For 32<k<64, we will need to reach n~1M and for k<32 it is n>1.5M.

Seems there is the odd reportable in the current range --- I just found this in my inbox

165*2^509812+1 Divides xGF(509808,10,7)

Keep crunching guys -- could be some others missed

Actually, I should have said "mostly" searched. Your prime came from a gap I overlooked. If you visit this page, Intervals searched for primes k.2^n + 1 with k < 300, and go to k=165, you'll see the gap.

Congratulations on your find!

If anyone is interested --From the data on the link supplied by John --

Given our current range of n = 500k + (now approx 516K)

Reportables might be found at

K = 97 (500K < n< 700K)
K = 135 (500K < n< 600K)
K = 165 (500K < n< 600K)
K = 167 (500K < n< 600K)
K = 1191 to 1199 (500K < n)

When n > 520K (with a few exclusions)

K = 999 to 1189 (n > 520K )

Become potential reportables--

The current range is not totally barren and will improve as we pass n > 520K -- Then as John stated n > 540K opens further ranges of K

DoES
____________
Member of AtP

Shown here is an Australian native rat (Ratus Kickarsus)

If anyone is interested --From the data on the link supplied by John --

Given our current range of n = 500k + (now approx 516K)

Reportables might be found at

K = 97 (500K < n< 700K)
K = 135 (500K < n< 600K)
K = 165 (500K < n< 600K)
K = 167 (500K < n< 600K)
K = 1191 to 1199 (500K < n)

When n > 520K (with a few exclusions)

K = 999 to 1189 (n > 520K )

Become potential reportables--

The current range is not totally barren and will improve as we pass n > 520K -- Then as John stated n > 540K opens further ranges of K

DoES

Thanks very much for this research. It's GREAT seeing the open ranges! :) As you noted in Meebo Chat, we are at n>520K so more k's are now available for "new" primes.

Thanks very much for this research. It's GREAT seeing the open ranges! :) As you noted in Meebo Chat, we are at n>520K so more k's are now available for "new" primes.

John

I have just noticed a strange event has occured in the PPS range stats -- in the 10 min prior to 4.04 am UTC 3 primes entered the database yet the total counter increased by 2 (1665 to 1667)

one is K=233 and oddly the other 2 are both in the K=57 range--

I seems the database may have counted the 2 in 57 as 1 prime---

I can be fairly certain of this as I have a simple little program that checks this database then date stamps & notifies me when new PPS primes are listed--

EDIT-- I should add that the primes and their finders and recorded -- only the the count for this range is wrong--

When I checked 2 were listed but there are definitly 3 primes there

DoES

Thanks for the info. At times, the counts may be off. Not exactly sure why or how this happens. However, no primes are missed...only the total counts are off.