Testing....

[kar_bon]

you can delete the "do_tosend.awk" in your distribution!

[gd_barnes]

Outstanding and fast work Max! I can feel it. The end of our long, tortured wait is nearing an end for a server that will run a more up-to-date version of LLR.

I'm beat tonight and am going to bed shortly but my Monday afternoon will be dedicated to attempting to break it in every way possible. I'll also fiddle around by loading different size tests in the server. I may even set up a 2nd test server to make it easier to go back and forth between the 2. I'll probably load some super small n=5K or 10K tests in there and run a gob of cores on it just to see what happens. :-)


Gary

[MyDogBuster]

Ian, can you please post here some more details.

I ran some CRUS stuff thru a test. 4 cores - range was from n=2.5K to n=10K about 9500 tests in all. I had 2 cores doing 5 tests each in a cache and the other 2 cores doing 100 tests each in a cache. I stopped them many times and restarted them with no problems. I even made sure I had 2 cores looking for work at the same time. I even tested all 4 error messages at the end of the script and each message showed up when it should have and the tests restarted where they should have. Found 3 primes and ran them against PFGW to verify.

I also ran about 1500 tests against the NPLB setup that came with the download. There I had a cache of 100 tests. Even found a prime on that one.

At the present time I don't have any cores available to test anymore. I'm trying to cleanup my CRUS stuff so that I can go back to NPLB and the upcoming k>300<400 drive. Time to go prime hunting again.

[kar_bon]

@Ian:
thanks for the information and an amazing wide range of testing. thanks.

@Max;
i had a closer look to you WIN-perl script.
if i understand perl correct, the function checkForPrimes will write found primes in the dedicated file when $individualPrimeLog = true. but you test to if($line != "") so if i'm right every time (workfile.res exists when cllr is done but if no prime is found, contains only the header line!) the header will be backuped, too!
i've not tested it running yet, but will do today.
nice quick work!

[mdettweiler]

Originally Posted by kar_bon

@Max;
i had a closer look to you WIN-perl script.
if i understand perl correct, the function checkForPrimes will write found primes in the dedicated file when $individualPrimeLog = true. but you test to if($line != "") so if i'm right every time (workfile.res exists when cllr is done but if no prime is found, contains only the header line!) the header will be backuped, too!
i've not tested it running yet, but will do today.
nice quick work!

That's what this is for a few lines earlier:

  # Open workfile.res and throw out the first line (the NewPGen header) - we don't need it
  open(RES, "workfile.res");
  $foo = <RES>;

It reads the first line of workfile.res to the variable $foo, which isn't used later.

Meanwhile, I just uncovered an actual bug: looking at the if($line != "") part (which is intended to copy off any line in workfile.res after the header that isn't blank), I see that it should be if($line ne "") instead (you use eq and ne instead of == and != for strings in Perl). It's been a little while since I last used Perl for anything big, as you might have noticed.

I'll have a fix online in a moment.

Edit: I've uploaded the fix, same places as before. Note that this prime-logging code is one area that I haven't been able to test, so be prepared for possible errors. If I can I'll do some additional testing of my own on it today.

[kar_bon]

so when will we start a major test?

say 10000 kn-pairs with one third at about n=5k,100k and 500k or something like this and many primes, too.
other testcase: at rieselprime, there's a file with all primes. what about a piece of that.
what about testing for twins: different header not tested yet with the script!

[mdettweiler]

Originally Posted by kar_bon

so when will we start a major test?

say 10000 kn-pairs with one third at about n=5k,100k and 500k or something like this and many primes, too.
other testcase: at rieselprime, there's a file with all primes. what about a piece of that.
what about testing for twins: different header not tested yet with the script!

Agreed, all of those would be good test cases. Gary, what do you say we try to do a big test later today with all of those worktypes? We could load them all in together (under the twin header, to test that).

We've already tested Karsten's batch script quite a bit, though it surely won't hurt to test it more. I'll test both it and my Perl script on my Windows box here.

[gd_barnes]

OK, guys, here's the timeframe:

It's 1:45 PM CST here now. I've had a good night's rest and am going to go eat some breakfast now. :-)

At about 2:15 PM I'll start copying the latest client to all of my Linux machines. I've got the equivalent of about 3 Linux quads on CRUS right now so I won't bother those for now. But that still leaves 8 Linux quads for this test. This process will take take about 30 mins.

So at ~2:45-3 PM, it will take me ~20-45 minutes to set up the servers with the proper test cases. I'll also create a 2nd test server for this, which will save quite a bit of time and allow specific situations to be tested more easily if there is a problem with one of them.

So the bottom line is that I'll start the cores up around 3:30 PM CST. That would be around 10:30 PM in Germany.

Nice work on the test cases Karsten. That's exactly what I like to see. I will be testing every one of those situations. I'll put something like 60-70 primes at the beginning of one of the tests so that each core is likely to get more than one prime. I want to see it write to the prime file with more than one prime per core all combined into one big primes file on a machine. That will be interesting. It's important that it works if primes come in from 2 different cores at the same time. I don't want to see it overlay itself if that happens and inadvertantly drop a prime.


Gary

[gd_barnes]

Originally Posted by mdettweiler

Behold: the long-awaited Linux/Perl version of the script is ready! I've tested it somewhat heavily on a Windows machine, though I haven't yet tested it on Linux. I'm pretty sure it will work, though; there's not really any OS-specific code in there and it works great on Windows. Gary and Ian, feel free to stress test the heck out of it.

Here's download links for Windows and Linux versions. The only difference between the two is which OS the LLR and LLRnet binaries are compiled for; swap those out and you can quite easily turn one into the other. All necessary instructions are included in readme.txt.
Windows: http://www.noprimeleftbehind.net/dow...0.61-win32.zip
Linux: http://www.noprimeleftbehind.net/dow...61-linux32.zip

OK, I'm a little confused. Will we be using Karsten's or Max's script for the Window's client? Regardless, I'll only be testing the Linux client this afternoon.


Gary

[kar_bon]

i think there's one reason a prime would not be written in the general primes.txt:

if one client found a prime and writes it into that file it will be locked and a second client found a prime the same moment could not write the same file: perhaps an error like 'file not found' should occur.
i'm thinking of a later optimization about the primes.txt:
show the client in that file,too!
how to: the llr-clientconfig.txt contains an option 'serviceName' and i named them like "LLRnet-G 995x" with x the number of the core/client. so the primes.txt looks like:

[2010-02-18 00:25:38] LLRnet-G 9951 found 2001 58406

i'm with the test with my Vista Quad. hope all test are positive, so we can provide it to our standard users (not much by now, but if this works, perhaps there'll be more again).

[kar_bon]

Originally Posted by gd_barnes

OK, I'm a little confused. Will we be using Karsten's or Max's script for the Window's client? Regardless, I'll only be testing the Linux client this afternoon.

i think, we should test all 3 versions! i'm with the normal DOS-script.
the reason i used this: perl has to be downloaded and installed first, an the customizing could call others errors a non-expert user could not handle!
so the normal DOS-version is ready to go on WIN-platforms without any external programs!