Searching for some beta testers after the easter weekend - Page 3

Brucifer · 2006-04-18, 04:13

Nice to hear the good results Lars!! :) So you've got the systems until tomorrow night then. :)

Bruce

ltd · 2006-04-18, 17:01

Good news.

We can close the first round of beta tests. Everything ran very smooth.
I will let the beta server continue to run cause i have some more scripts to install but there is no need anymore to push it with force.

Many thanks to the people who helped to stress the server:

Brucifer,Dagger,japelprime,thommy3 and usa

I will now start to think about the best way of moving the main DB over to the new server.( That means with the least effort and outage)
I keep you all informed.

Lars

Citrix · 2006-04-18, 19:16

Hi Lars,

Were any primes found in the process? Also could you update the getting started thread + other guides hosted on your website about the URL to the new server instead of the old one.

Thanks

ltd · 2006-04-18, 20:10

There were no primes so far but i think we will find one further down the road with the next beta tests i plan.

The links within the getting startet pages on the new server are already pointing to www.psp-project.de.

For the thread here in the forum is it possible that you make the update?
I am fighting with a 1GB backup of the database and hope to have it installed sometime tonight.

Lars

Citrix · 2006-04-19, 00:58

I took care of the getting started page, if you make anymore changes, let me know so I can update the getting started thread. I will try to update the page on geocities server, though it is a useless page, it just duplicates the getting started thread.

Citrix

edit:-

I changed the main page to point to the getting started thread. It might be a good idea to have www.psp-project.de, to host the main psp page, and then have the stats on a sub page. A wiki for PSP might also be helpful on the mersennewiki site, but since creating the wiki will take a lot of time, I will leave that for the future.

Citrix · 2006-04-19, 13:55

A typo on stats page

" Open n per block(Shown upto 2.5Mil):"

Should be 3.0 M

I did some calculations and according to them we should find a prime under 2.4M. So I think we should speed our efforts in the PRP section for a month or so.

Citrix

hhh · 2006-04-19, 18:53

Quote:

Originally Posted by Citrix

I did some calculations and according to them we should find a prime under 2.4M. So I think we should speed our efforts in the PRP section for a month or so.
Citrix

A priori, I disagree. I just don't trust ANY calculations.

What if we push the PRP to 2.4M and don't find a prime? Is it overdue and we should push up to 2.5M? This is an eternal argument; we know almost nothing about the prime distribution; I personally believe it is a kind of memory-less thing; and I feel waiting for a prime is like waiting for overdue numbers in roulette.

Anyway, I have evidence from calculus against your strategy, too.

Consider the two scenarios
we only sieve for a while and restart PRP later,
we PRP right now until a prime, and restart sieving then with the optimized dat.

If we sieve first, we waste some processor time by the sieving of numbers for which we could have a prime very soon. But this time doesn't count at 100%, as the sieve speed is less than propotional to the dat size.

An the other side we save good PRP time on all the k's, which is counting 100%, as we have to do the tests sooner or later.

Doublecheck PRP is not yet an issue, as the error rate is really low (I'm currently running some tests).

And now, all these considerations go out of the window, as everybody is free in this project to do what he wants, and every help is highly appreciated, be it PRP or sieve.

But: IMHO, sieve is still more worth for the project as a whole then sieve, as it finds more factors than PRP tests in the same time. (No english, but one can understand, I guess).

And Citrix, I hope you find that prime you're searching for.

Cheers, Yours H.

ltd · 2006-04-19, 19:04

Quote:

Originally Posted by Citrix

A typo on stats page

" Open n per block(Shown upto 2.5Mil):"

Should be 3.0 M

I did some calculations and according to them we should find a prime under 2.4M. So I think we should speed our efforts in the PRP section for a month or so.

Citrix

Error corrected.

For the prime my personal feeling ( no i have no arguments) is that it will come somewhere between 2.6M and 2.8M.

Lars

Citrix · 2006-04-19, 19:35

hhh,

If we knew for sure that there was a prime to be found under 2.4 M would you keep on sieving and never PRP or PRP only when 1 test is left. I personally think it would be easier to just PRP first find the prime and then sieve faster.

As for knowing for sure, the probability of finding a prime increases as we do more and more tests with non-prime results. But the probabibility never reaches 100%, so you can never be sure.

I think the project will benefit from a prime than sieving, so I am putting my computers on PRP. Other people are free to choose what they see best.

I hope I or someone else finds the prime soon.

Citrix

hhh · 2006-04-19, 21:30

Of course sieving has only to be considered as long as the factor throughput is higher than the PRP throughput (as a rough approximation). We are still away from changing this situation.
That is only to justify my point.
But we agree: who wants the prime, must PRP, and that's what makes the project work, after all.
H.

2006-04-18, 17:01	#24
ltd Apr 2003 1404₈ Posts	Good news. We can close the first round of beta tests. Everything ran very smooth. I will let the beta server continue to run cause i have some more scripts to install but there is no need anymore to push it with force. Many thanks to the people who helped to stress the server: Brucifer,Dagger,japelprime,thommy3 and usa I will now start to think about the best way of moving the main DB over to the new server.( That means with the least effort and outage) I keep you all informed. Lars Last fiddled with by ltd on 2006-04-18 at 17:01

2006-04-19, 00:58	#27
Citrix Jun 2003 2·7·113 Posts	I took care of the getting started page, if you make anymore changes, let me know so I can update the getting started thread. I will try to update the page on geocities server, though it is a useless page, it just duplicates the getting started thread. Citrix edit:- I changed the main page to point to the getting started thread. It might be a good idea to have www.psp-project.de, to host the main psp page, and then have the stats on a sub page. A wiki for PSP might also be helpful on the mersennewiki site, but since creating the wiki will take a lot of time, I will leave that for the future. Last fiddled with by Citrix on 2006-04-19 at 01:09

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Weekend GT730 Project	masser	GPU Computing	17	2016-11-15 01:55
MemtestG80 and MemtestCL: Memory Testers for GPUs	Jeff Gilchrist	Hardware	0	2010-08-19 16:32
I'm Dreaming of a White Easter	xilman	Lounge	8	2008-03-24 19:39
Happy Easter!	mfgoode	Lounge	4	2007-04-11 03:51
Attention PRP Testers (with SSE2 non-P4 machines)	axn	Sierpinski/Riesel Base 5	5	2006-10-12 05:08

2006-04-18, 04:13	#23
Brucifer Dec 2005 139₁₆ Posts	Nice to hear the good results Lars!! :) So you've got the systems until tomorrow night then. :) Bruce

2006-04-18, 19:16	#25
Citrix Jun 2003 2·7·113 Posts	Hi Lars, Were any primes found in the process? Also could you update the getting started thread + other guides hosted on your website about the URL to the new server instead of the old one. Thanks

2006-04-18, 20:10	#26
ltd Apr 2003 2²×193 Posts	There were no primes so far but i think we will find one further down the road with the next beta tests i plan. The links within the getting startet pages on the new server are already pointing to www.psp-project.de. For the thread here in the forum is it possible that you make the update? I am fighting with a 1GB backup of the database and hope to have it installed sometime tonight. Lars

2006-04-19, 13:55	#28
Citrix Jun 2003 2×7×113 Posts	A typo on stats page " Open n per block(Shown upto 2.5Mil):" Should be 3.0 M I did some calculations and according to them we should find a prime under 2.4M. So I think we should speed our efforts in the PRP section for a month or so. Citrix

2006-04-19, 19:35	#31
Citrix Jun 2003 62E₁₆ Posts	hhh, If we knew for sure that there was a prime to be found under 2.4 M would you keep on sieving and never PRP or PRP only when 1 test is left. I personally think it would be easier to just PRP first find the prime and then sieve faster. As for knowing for sure, the probability of finding a prime increases as we do more and more tests with non-prime results. But the probabibility never reaches 100%, so you can never be sure. I think the project will benefit from a prime than sieving, so I am putting my computers on PRP. Other people are free to choose what they see best. I hope I or someone else finds the prime soon. Citrix

2006-04-19, 21:30	#32
hhh Jun 2005 373 Posts	Of course sieving has only to be considered as long as the factor throughput is higher than the PRP throughput (as a rough approximation). We are still away from changing this situation. That is only to justify my point. But we agree: who wants the prime, must PRP, and that's what makes the project work, after all. H.