mersenneforum.org (https://www.mersenneforum.org/index.php)
-   PSearch (https://www.mersenneforum.org/forumdisplay.php?f=13)

 bc 2003-04-10 13:06

PRP queries

Some queries on PRP (the program)

2. Can anyone explain what the "time per bit" is proportional to? Obviously it varies with CPU speed, but it seems to vary per number, where I would have expected it to be constant. I don't know how this stuff works, but I would like to know when I benchmark hardware changes I've got some baseline to work on.

Cheers and Happy Hunting

 wfgarnett3 2003-04-11 02:43

Hi bc,

http://www.mersenne.org/gimps

prp.zip for Windows, and prp.tgz for Linux You do have the latest version of PRP.

Time per bit in milliseconds is based on the FFT range. Let's look at the very similar Prime95 page (which deals with mersenne numbers and not proth numbers):
http://www.mersenne.org/bench.htm

For exponents of 6465000 to 7690000 for mersenne numbers, the FFT range is 384K, which is the amount of memory (if I understand correctly) that is used. So all numbers in that range will have the same time per bit. Now, again if I am understanding correctly, towards the top of the range more memory is starting to be needed so for the next range of 7690000 to 8970000, the FFT is 448K. And since the FFT is bigger and more memory is needed, the time per bit for numbers in this range is slower than the previous range. And so on :) Proth numbers are similar in regards to FFT.

Regards,
william

 1260 2003-04-23 13:59

Another question: When will the option for primality testing of expressions be implemented?

--Ray

 wfgarnett3 2003-04-25 05:30

Hi Ray,

Primality is hard to prove for general numbers. For easily factorable numbers when you add or subtract one, try:
http://www.primeform.net/openpfgw/

Mersenne, Proths, Riesels, and Generalized Fermat's are all factorable 100% when you add or subtract one, that's why they are easy to prove; they use Wilson's Theorm (I think). For any random odd number, the program to use is:

http://www.ellipsa.net/

But it takes a long time to use Primo to prove an odd number prime; but it does allow any odd integer.

Regards,
william

 1260 2003-04-28 10:17

Thanks William. I'm now using version 0.4 update 4 of PrimeForm because it has an option to check an expression. Primo can only test up to 2^40960 and the numbers I am testing are over 2^300000.

 paulunderwood 2003-04-28 17:21

You should use WinPFGW which is over three times faster than PrimeForm :surprised: :

http://www.teamprimerib.com/pfgw/20030108_Win_Dev_Alpha_PFGW_WinPFGW.zip

WinPFGW allows the use of expressions by way of an "ABC2 file" -- please see the documentation. HTH

 1260 2003-05-01 13:16

Yes. I've now switched to WinPFGW. It's really faster. Thanks.

 All times are UTC. The time now is 08:56.