Why V5 has no factoring stats from 610-720M? - Page 5

Prime95 · 2008-05-05, 20:31

Quote:

Originally Posted by antiroach

are you also submitting the results when no factor has been found ? (from looking at the recently cleared results it does not seem like you are doing this)

no factor results are not needed - the extra microseconds for prime95 to retest to 2^43 is much less than the TCP/IP and database overhead of sending the no factor results.

I've been working on the manual results reporting page. However, I cannot get the file upload to work. So for now Batalov's slow method is as good as it gets.

Batalov · 2008-05-05, 20:56

Quote:

Originally Posted by antiroach

are you also submitting the results when no factor has been found ? (from looking at the recently cleared results it does not seem like you are doing this)

No, they are factored too low, they should be available for factoring no matter if they were reported as "no factor to 2^43" or if they are "factored to 0" (as they are now). I estimated that even without "no factor to 2^43" messages I will be sending URL messages until two weeks from now at this rate...

With them it would have taken take two months. I am afraid the network adapter will fry.

Really, my idea was to avoid network traffic of trivial back-and-forth communications, so that you guys will get "enriched ore" - by giving away those files to be fed to the server directly, but since George earlier wrote that there's no such way, I simply wanted to see how much time it takes to feed all the *.out43 participants via worktodo.txt files into 25.6 Prime95 client. Prime95, of course, revalidates them and enhances them to 2^50 on the fly and then (as it turns out) the bottleneck still remains to send the results over the network (takes apparently about a day for every 10M range). I will probably stop this silly business after 800M or 810M. Or if it is useful, let me know.

P.S. At least 1) I am in San Diego and the host of the server is (or is it?) in San Diego - I might be in a better position to do the silly networking, 2) I do only half of network - I don't request the assignments; I put them into worktodo.txt; the server tried to "accession" them; but I stop restarted and the process went on without "registering" a million assignments.
___________

P.P.S. I completely agree with the message below, too.

antiroach · 2008-05-05, 21:19

now i see how it makes sense to not send the "no factor" results. Since the work has already been done I believe it is useful to keep submitting the results, if they are always ahead of the current factoring efforts as handed out by the server. If the server catches up (or passes) the values you are submitting then all the work will be duplicated unnecessarily.

antiroach · 2008-05-05, 22:07

Quote:

Originally Posted by Batalov

P.P.S. I completely agree with the message below, too.

Would using multiple prime95 clients to perform the submissions speed up this process?

CCol · 2008-05-06, 03:46

Yes, I would vote to keep submitting the results. This has to be a good test of the V5 server.

Batalov · 2008-05-06, 04:21

OK, even though I was about to say that it is rather useless (more ruminations below), even though entertaining procedure. A learning experience.

I found how I was able to avoid "registering" the assignments, though! (At least I think so) and I freely share this bug, so that it will be patched and impossible to use. I have a hunch that this happens when I put Unix-style line breaks in the worktodo.txt file for use in a Windows Prime95.exe! The program devours the worktodo.txt file for quite some time, but "thinks" that it is a single line, doesn't break and doesn't misparse it, only _does not_ proceed to register, but goes straight to work... Otherwise I would have given up long ago, but I got lucky from the very start, because I don't use

Windows (hate it really), except for Prime95.exe... I prepared the worktodo.txt files on my real computers, of course. You know what I mean. awk, grep, perl ...

I realized that by inserting the forgotten line "[Worker #1]" and saving it in WordPad - I made it a Windows file and then the program starts registering ad nauseum without even doing anything useful...

I want to load a next bunch (810-814M), and then do (815-819M) "honestly" (i.e. all numbers, not the ones with pre-discovered factors). I want to compare the running times and the stresses to different parts of the computer... I suspect that I got it all wrong from the beginning. I.e. there's no pain in LMH-TF'ing... for each "easy" number that slips right through with a factor there's another three that take a few seconds. So - one gets a healthy mix for normal network casual conversation with the server. As opposed to a monster list that I created for my own punshment and burning the network adapter... But you're right it is a good extreme test. For both server and the client.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Overall Stats	Fred	PrimeNet	7	2016-02-10 14:48
CSVs for stats available + New combined stats	opyrt	Prime Sierpinski Project	3	2010-05-31 08:13
Stats	em99010pepe	Sierpinski/Riesel Base 5	66	2008-02-06 19:42
"Factoring only" results / stats in account details page	schneelocke	PrimeNet	3	2004-01-07 22:12
Factoring stats?	bayanne	Software	1	2002-10-02 14:23

2008-05-05, 21:19	#47
antiroach Jun 2003 2²×61 Posts	now i see how it makes sense to not send the "no factor" results. Since the work has already been done I believe it is useful to keep submitting the results, if they are always ahead of the current factoring efforts as handed out by the server. If the server catches up (or passes) the values you are submitting then all the work will be duplicated unnecessarily.

2008-05-06, 03:46	#49
CCol Jul 2007 5·7 Posts	Yes, I would vote to keep submitting the results. This has to be a good test of the V5 server.

2008-05-06, 04:21	#50
Batalov "Serge" Mar 2008 San Diego, Calif. 281D₁₆ Posts	OK, even though I was about to say that it is rather useless (more ruminations below), even though entertaining procedure. A learning experience. I found how I was able to avoid "registering" the assignments, though! (At least I think so) and I freely share this bug, so that it will be patched and impossible to use. I have a hunch that this happens when I put Unix-style line breaks in the worktodo.txt file for use in a Windows Prime95.exe! The program devours the worktodo.txt file for quite some time, but "thinks" that it is a single line, doesn't break and doesn't misparse it, only _does not_ proceed to register, but goes straight to work... Otherwise I would have given up long ago, but I got lucky from the very start, because I don't use Windows (hate it really), except for Prime95.exe... I prepared the worktodo.txt files on my real computers, of course. You know what I mean. awk, grep, perl ... I realized that by inserting the forgotten line "[Worker #1]" and saving it in WordPad - I made it a Windows file and then the program starts registering ad nauseum without even doing anything useful... I want to load a next bunch (810-814M), and then do (815-819M) "honestly" (i.e. all numbers, not the ones with pre-discovered factors). I want to compare the running times and the stresses to different parts of the computer... I suspect that I got it all wrong from the beginning. I.e. there's no pain in LMH-TF'ing... for each "easy" number that slips right through with a factor there's another three that take a few seconds. So - one gets a healthy mix for normal network casual conversation with the server. As opposed to a monster list that I created for my own punshment and burning the network adapter... But you're right it is a good extreme test. For both server and the client.