PFGW 4.0.4 (with gwnum v30.10) Released - Page 12

rogue · 2012-08-20, 18:08

Quote:

Originally Posted by henryzz

I think I have found a bug in 64-bit pfgw. If I run the following script then it seems to work modulo 2^32 when running the 64-bit version. It runs properly with the 32-bit version of pfgw.

Another frustrating feature/bug I found when writing this script is if you PRP a number and it is trivially prime then ISPRP is not set to 1. If a number is prime then it should also be prp.

When I have a chance, I'll take a look.

As for the second issue, the documentation isn't clear on what happens if you do a PRP test for a number that is trivially prime. I'll probably change the code to do as you suggest.

rogue · 2012-08-21, 20:05

I found the first problem. It affects 64-bit Windows builds only when working with numbers between 32 and 64 bits in size. This bug does not affect Mac or Linux builds.

rogue · 2012-08-21, 20:39

I've posted the Windows build of pfgw here. This build fixes the problems discovered by henryzz.

I intend to upgrade to gwnum v27.7. I will post Linux and Mac builds at that time.

henryzz · 2012-08-22, 13:36

If I am testing a file with the header:
ABC $a*2^$b-1 & $a*2^$b+1 & $a*2^$b+5
then if I find prime for all of these three forms then it outputs - Complete Set - in pfgw.out. Is there a way for getting it to stop testing after the first complete set of primes like there is for a single prime? Could this make it into the coming release?

rogue · 2012-08-22, 14:03

Quote:

Originally Posted by henryzz

If I am testing a file with the header:
ABC $a*2^$b-1 & $a*2^$b+1 & $a*2^$b+5
then if I find prime for all of these three forms then it outputs - Complete Set - in pfgw.out. Is there a way for getting it to stop testing after the first complete set of primes like there is for a single prime? Could this make it into the coming release?

None that I can think of. I don't think this would be an easy change to make.

The only alternative I can think of is to use a pfgw script.

henryzz · 2012-08-22, 23:00

Quote:

Originally Posted by rogue

None that I can think of. I don't think this would be an easy change to make.

The only alternative I can think of is to use a pfgw script.

Looking at the code it doesn't look impossible but just complex. I might give it a try sometime.

rogue · 2012-09-18, 18:34

I have posted pfgw 3.6.6 over at sourceforge. Right now only the Windows build is available, but I hope to have the Mac and Linux builds available in the next day or two.

The only change is that pfgw 3.6.6 uses gwnum 27.8. This release of gwnum fixes an infinite loop that occurs with small AVX FFTs of the form k*b^n+c where abs(c) != 1.

henryzz · 2012-11-21, 00:37

I was just running tests to see how long each test would take on a number to determine the sieve depth. I ended up running a full 35 minute test to get a time. Is there a way to get iteration times in pfgw so I can extrapolate?
Another issue I can across is that I ended up doing quite of lot of tests to discover where the fft boundaries are. Is there an easier way pfgw can allow you to access this information?

rogue · 2012-11-21, 01:27

Quote:

Originally Posted by henryzz

I was just running tests to see how long each test would take on a number to determine the sieve depth. I ended up running a full 35 minute test to get a time. Is there a way to get iteration times in pfgw so I can extrapolate?
Another issue I can across is that I ended up doing quite of lot of tests to discover where the fft boundaries are. Is there an easier way pfgw can allow you to access this information?

You can try to use the -m benchmark, but I haven't used it in a while, so I don't know how accurate it is.

pfgw does not have code to determine where the FFT boundaries are. I don't know how easy that would be to code. I might need some help from George (Prime95) for that.

Often what I do is about 10% of the PRP test, then extrapolate from that.

henryzz · 2012-11-21, 09:20

Quote:

Originally Posted by rogue

You can try to use the -m benchmark, but I haven't used it in a while, so I don't know how accurate it is.

pfgw does not have code to determine where the FFT boundaries are. I don't know how easy that would be to code. I might need some help from George (Prime95) for that.

Often what I do is about 10% of the PRP test, then extrapolate from that.

How do you measure the time for that 10%? PFGW outputs no timing output until the test is finished.

rogue · 2012-11-21, 13:35

Quote:

Originally Posted by henryzz

How do you measure the time for that 10%? PFGW outputs no timing output until the test is finished.

Let me restate that. I use a clock and run the PRP test for up to two minutes. I then ^C PFGW then restart to see which bit it restarts at. I can then estimate the time to do a full PRP test.

I agree that it would be better if PFGW could do that itself, but the challenge is the change in FFT boundaries. PFGW has no knowledge of that.

2012-08-21, 20:05	#123
rogue "Mark" Apr 2003 Between here and the 3·2,447 Posts	I found the first problem. It affects 64-bit Windows builds only when working with numbers between 32 and 64 bits in size. This bug does not affect Mac or Linux builds. Last fiddled with by rogue on 2012-08-21 at 20:05

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
A possible bug in LLR/PFGW while using GWNUM (no bug in P95)	Batalov	Software	77	2015-04-14 09:01
PFGW 3.2.0 has been Released	rogue	Software	94	2010-09-14 21:39
PFGW 3.2.3 has been Released	rogue	Software	10	2009-10-28 07:07
PFGW 3.2.1 has been released	rogue	Software	5	2009-08-10 01:43
PFGW 3.1.0 has been Released	rogue	Software	25	2009-07-21 18:13

2012-08-21, 20:39	#124
rogue "Mark" Apr 2003 Between here and the 3×2,447 Posts	I've posted the Windows build of pfgw here. This build fixes the problems discovered by henryzz. I intend to upgrade to gwnum v27.7. I will post Linux and Mac builds at that time.

2012-08-22, 13:36	#125
henryzz Just call me Henry "David" Sep 2007 Liverpool (GMT/BST) 3·23·89 Posts	If I am testing a file with the header: ABC $a2^$b-1 & $a2^$b+1 & $a*2^$b+5 then if I find prime for all of these three forms then it outputs - Complete Set - in pfgw.out. Is there a way for getting it to stop testing after the first complete set of primes like there is for a single prime? Could this make it into the coming release?

2012-09-18, 18:34	#128
rogue "Mark" Apr 2003 Between here and the 3·2,447 Posts	I have posted pfgw 3.6.6 over at sourceforge. Right now only the Windows build is available, but I hope to have the Mac and Linux builds available in the next day or two. The only change is that pfgw 3.6.6 uses gwnum 27.8. This release of gwnum fixes an infinite loop that occurs with small AVX FFTs of the form k*b^n+c where abs(c) != 1.

2012-11-21, 00:37	#129
henryzz Just call me Henry "David" Sep 2007 Liverpool (GMT/BST) 3×23×89 Posts	I was just running tests to see how long each test would take on a number to determine the sieve depth. I ended up running a full 35 minute test to get a time. Is there a way to get iteration times in pfgw so I can extrapolate? Another issue I can across is that I ended up doing quite of lot of tests to discover where the fft boundaries are. Is there an easier way pfgw can allow you to access this information?