Running Glucas correctly?
 

Newbie question.

I've recently switched to a mac after running prime95 on a Windows machine. I've downloaded and installed the binary on my Powerbook G4 1Ghz with Panther, but the documentation was so poorly written I'm unsure how to exactly proceed.

I figured that I needed to create a glucas.que file with the exponent I'm testing (33603839) followed by three zeros. I did that. I then ran the command

./Glucas -q glucas.que

Which supposedly begins a LL test and writes the results to a file called glucas.out. Doing cat glucas.out shows that the program is performing iterations on this exponent and saving its work to some unknown file. Using "top," I can see that when no other program is active, Glucas uses something like 75% of the CPU cycles. It's doing something.

Is it indeed doing a LL test? If so, must I check the glucas.out file in 2 months or so to see if I successfully found the first 10 million digit prime? Would the glucas.out file be what I send off to the GIMPS folks? Also, roughly how many iterations are required to check a 10 million digit prime--I'd like to know roughly how much progress I'm making.

Thanks. Any other tips on how to use this arcanely documented program would be greatly appreciated.

I have no experiences with glucas or even nix/nux, so I can answer only one of your questions: your exponent is 33603839, so 33603838 iterations have to be done, i think.

I believe the file you want is called "results" (unless you specified a different name with the -r option).

glucas.out would be for verbose output.

In the "results" file, look for lines of the form:
[code]
M1753 is not prime. Res64: A984D3CA0516EC2D. G29: 234882C56FA40728,116
[/code]
which indicate successful completion of a Lucas-Lehmer test. These lines need to be manually submitted (for instance, e-mailed to George).

It would be nice to have a version of Glucas that automatically interacts with PrimeNet in set-it-and-forget-it style.

Thanks for answering my question about iterations.

Unfortunately, there isn't any file named "results" in the Glucas directory. Doing ls, I come up with only the following files:

.DS_Store
ChangeLog
Glucas
README
g33603839
glucas.ini
glucas.out
glucas.que
s33603839
selftest.ini
selftest.out
selftest.res
t33603839

Neither g33603839, s33603839, nor t33603839 appear to be readable.

Checking the glucas.out file, I found lines like these:

Iter. 20000 ( 0.06%), Err= 0.219, 1180.65 user 63% CPU (0.941 sec/iter).
M33603839. Saved file at iteration 20480. Res64: C3C44CEF927646BE.
M33603839. Saved file at iteration 24576. Res64: CBBCDBA9D5CFBEE4.
M33603839. Saved file at iteration 28672. Res64: 13A6F9983078D942.
Iter. 30000 ( 0.09%), Err= 0.219, 6014.46 user 73% CPU (0.824 sec/iter).
M33603839. Saved file at iteration 32768. Res64: 2D4ACD83ACE0F9A4.
M33603839. Saved file at iteration 36864. Res64: C980FE10721BD1BB.
Iter. 40000 ( 0.12%), Err= 0.203, 6036.98 user 72% CPU (0.843 sec/iter).
M33603839. Saved file at iteration 40960. Res64: 0CD56CA2BCB1927A.
M33603839. Saved file at iteration 45056. Res64: 87C6134D4E4CAFC6.
M33603839. Saved file at iteration 49152. Res64: 97FFF089D5DCF33B.
Iter. 50000 ( 0.15%), Err= 0.203, 6039.62 user 73% CPU (0.823 sec/iter).
M33603839. Saved file at iteration 53248. Res64: 36904F5A74986FCB.
M33603839. Saved file at iteration 57344. Res64: 9C3B91F33CC2782C.
Iter. 60000 ( 0.18%), Err= 0.203, 6012.83 user 73% CPU (0.827 sec/iter).
M33603839. Saved file at iteration 61440. Res64: 22824C68A42BAE6A.
M33603839. Saved file at iteration 65536. Res64: 0CBD831AFBB91155.
M33603839. Saved file at iteration 69632. Res64: D124AFC238295C48.
Iter. 70000 ( 0.21%), Err= 0.203, 6009.12 user 74% CPU (0.817 sec/iter).

These look to show total progress and seconds per iteration. Multiplying 33603838 (remember that minus 1) by .817 gives 27454335.646. Dividing that number by 86,400 (number of seconds in a day) gives 317.76!!

Am I correct in assuming that it will take ten and a half months to check this one number assuming my computer stays on 24 hours a day, or are my calculations off?

Am I also correct in assuming that no results file has been made because the LL test has not been completed yet? Glucas.out is the best file I have giving anything that remotely looks like results.

[QUOTE][i]Originally posted by Unregistered [/i]
[B]
Am I correct in assuming that it will take ten and a half months to check this one number assuming my computer stays on 24 hours a day, or are my calculations off?
[/B][/QUOTE]

They look good to me. Perhaps you should consider testing a smaller exponent.

Your time estimation seems to be correct.
You may consider doing doublechecks or regular LL-Tests rather than searching for a 10,000,000 Digit Prime.
Smaller exponents mean less iterations and faster iterations, so tests will complete a lot faster.

Wow, that's disappointing.

On the prizes page, it states that a Pentium IV 2Ghz machine can test a 10 million digit prime in only 2 months with a roughly 1 in 250,000 chance of success.

I thought, um, that G4 processors were supposed to be just as fast, if not faster than Pentium IVs. Why does it now look to be 5 times slower? Is prime95 just that much better of a program?

Also, if we have no formula for determining the frequency of prime numbers, how can we estimate a roughly 1 in 250,000 chance of success? For all we know, it could be higher or lower, right?

I'd only like to test 10 million digit primes because of the greater monetary reward.

Keyboard goof. Hit the wrong button Please ignore.

You probably don't have a "results" file yet because you haven't yet completed an exponent.

Perhaps if you stop and restart the program, you might get a line in "results" like:

[code]
Restarting from iteration 25664 .Exponent 5896207. Shifted 576562
[/code]

You might try adding
[code]
Verbose_flag=1
[/code]
to your glucas.ini file for more frequent output.

[QUOTE][i]Originally posted by Unregistered [/i]
[B]
I thought, um, that G4 processors were supposed to be just as fast, if not faster than Pentium IVs. Why does it now look to be 5 times slower? Is prime95 just that much better of a program?

Also, if we have no formula for determining the frequency of prime numbers, how can we estimate a roughly 1 in 250,000 chance of success? For all we know, it could be higher or lower, right?
[/B][/QUOTE]

Actually we do have some fairly good estimates of the frequency of prime numbers, and Mersenne primes in particular. See:
[url]http://www.utm.edu/research/primes/notes/faq/NextMersenne.html[/url]

G4 processors are fast. However, Prime95 is written in super-optimized handcrafted x86 assembly code that has steadily improved over 8 years and 23 versions, whereas Glucas is a C program.

It's not that Glucas is slow... it's just that Prime95 is very, very fast.

Glucas may yet have its time in the sun, since future clients may run on all sorts of exotic hardware: game consoles, graphics cards, floating-point coprocessors like ClearSpeed. So we will definitely need a strong C client program going forward.

[QUOTE]However, Prime95 is written in super-optimized handcrafted x86 assembly code that has steadily improved over 8 years and 23 versions, whereas Glucas is a C program.[/QUOTE]

That and the SSE2.