PFGW 3.8.3 (with gwnum v28.7) Released - Page 15

rogue · 2013-01-07, 13:45

I've updated the zip file. I know that both the 32-bit and 64-bit exes work, so if the zip file is corrupted, it might be a problem created by WinZip (which I used to create the zip file) or the upload.

pepi37 · 2013-01-07, 14:21

Now zip from sourceforge is ok, update 3.7.0 at Boinc Confederation

C:\Users\Phenom>C:\Users\Phenom\Desktop\pfgw64.exe -d -q"701*2^293561+1"
PFGW Version 3.7.0.64BIT.20130104.Win_Dev [GWNUM 27.8]

701*2^293561+1 is composite: RES64: [E03CF76F4D8B6048] (25.4841s+0.0001s)

C:\Users\Phenom>C:\Users\Phenom\Desktop\pfgw32.exe -d -q"701*2^293561+1"
PFGW Version 3.7.0.32BIT.20130104.Win_Dev [GWNUM 27.8]

701*2^293561+1 is composite: RES64: [E03CF76F4D8B6048] (30.4190s+0.0002s)

henryzz · 2013-01-11, 01:18

Quote:

Originally Posted by rogue

Factoring might be faster, not just because it has its own asm routine (instead of GMP), but because it can test two potential factors at a time in the asm routine.

I was referring to the factoring. I use the factoring quite a bit when starting of bases for CRUS. Has the default depth of factoring changed at all now it it is much faster?

rogue · 2013-01-11, 01:29

Quote:

Originally Posted by henryzz

I was referring to the factoring. I use the factoring quite a bit when starting of bases for CRUS. Has the default depth of factoring changed at all now it it is much faster?

I believe it has, but I would have to look at the code in depth to determine how much deeper it will factor. In any case, this change will significantly benefit uses who run the new bases script used with the version of pfgw. How much? I don't know. Someone would have to run 3.6.7 and 3.7.0 on the same range (to a low n) and see how much faster it is

henryzz · 2013-01-11, 01:42

Quote:

Originally Posted by rogue

I believe it has, but I would have to look at the code in depth to determine how much deeper it will factor. In any case, this change will significantly benefit uses who run the new bases script used with the version of pfgw. How much? I don't know. Someone would have to run 3.6.7 and 3.7.0 on the same range (to a low n) and see how much faster it is

I will do some comparison runs soon.

rogue · 2013-01-16, 00:39

I have released pfgw 3.7.1 to sourceforge. In this release the -m option has been replaced with -B. In other words, I have rewritten the benchmarking code. This is documented in pfgwdoc.txt as follows:

Code:

 -B   Benchmark
      This will start to test how fast pfgw works on your computer.  This option
      can take one or more of the following comma delimited parameters:
         gen   --> benchmark generic modular reduction on f!-1 for 100<=f<=1000000
         spec  --> benchmark special modular reduction on 3*2^n+1 for 100<=n<=10000000
         fact  --> benchmark trial factoring
         fft   --> benchmark all FFT sizes
         minf= --> override min f on generic modular reduction benchmark
         maxf= --> override max f on generic modular reduction benchmark
         minn= --> override min n on special modular reduction benchmark
         maxn= --> override max n on special modular reduction benchmark
         gexp= --> expression to use for benchmarking generic modular reduction, must use
                   the variable f, e.g. 500*f#+1, 3^(f*2)+4^f+1
         sexp= --> expression to use for benchmarking special modular reduction, must
                   be in the form k*b^n+/-c or b^n+/-c, e.g. 785*366^n-1

      Examaples
         -B                      --> with no parameters acts like -Bgen,spec
         -Bspec,fft              --> benchmarks all FFT sizes for special modular reduction on 3*2^n+1
         -Bspec,sexp=500*67^n-1  --> benchmarks special modular reduction on 500*67^n-1
         -Bmaxn=100000           --> benchmarks special modular reduction on 3*2^n+1 up to n=100000

Here is an example of the output:

Code:

pfgw64 -Bmaxf=50000,maxn=50000
PFGW Version 3.7.1.64BIT.20130115.Mac_Dev [GWNUM 27.8]

Benchmarking!  This may take several minutes.                                    

Benchmarking generic modular reduction for f!-1!                                    
[appx max f]   [num bits]   [iter time]  [PRP time]  [FFT used]                                    
--------------------------------------------------------------------------------                                    
       100          524       1281 ns      671 us     generic reduction FFT length 64                                    
       400         2886       7026 ns       20 ms     generic reduction FFT length 64                                    
       700         5612         10 us       56 ms     generic reduction FFT length 320                                    
      1000         8529         24 us      208 ms     generic reduction FFT length 512                                    
      4000        42099        118 us     4982 ms     generic reduction FFT length 896                                    
      7000        79320        254 us       20 s      generic reduction FFT length 4K                                    
     10000       118458        423 us       50 s      generic reduction FFT length 8K                                    
     40000       553809       2415 us       22 mi     generic reduction type-0 FFT length 12K, Pass1=48, Pass2=256                                    
     50000       708356       3019 us       35 mi     generic reduction Pentium4 type-1 FFT length 56K, Pass1=224, Pass2=256                                    
                                    

Benchmarking special modular reduction for 3*2^n+1!                                    
[appx max n]   [num bits]   [iter time]  [PRP time]  [FFT used]                                    
--------------------------------------------------------------------------------                                    
       100          101        748 ns       75 us     all-complex FFT length 32                                    
       400          401        253 ns      101 us     all-complex FFT length 32                                    
       700          701        294 ns      206 us     all-complex FFT length 32                                    
      1000         1001        412 ns      412 us     all-complex FFT length 32                                    
      4000         4001       1593 ns     6375 us     all-complex FFT length 48                                    
      7000         7001       2952 ns       20 ms     all-complex FFT length 192                                    
     10000        10001       4073 ns       40 ms     all-complex FFT length 384                                    
     40000        40001         19 us      793 ms     all-complex FFT length 512                                    
     50000        50001         31 us     1585 ms     all-complex FFT length 2K

and another

Code:

pfgw64 -Bfft,maxn=10000,sexp=12345*6^n-1
PFGW Version 3.7.1.64BIT.20130115.Mac_Dev [GWNUM 27.8]

Benchmarking!  This may take several minutes.                                    

Benchmarking special modular reduction for 12345*6^n-1!                                    
[appx max n]   [num bits]   [iter time]  [PRP time]  [FFT used]                                    
--------------------------------------------------------------------------------                                    
       180          478       1149 ns      549 us     FFT length 32                                    
       270          711       1811 ns     1288 us     FFT length 48                                    
       360          944       2802 ns     2645 us     FFT length 64                                    
       450         1176       3086 ns     3629 us     FFT length 80                                    
       530         1383       2952 ns     4082 us     FFT length 96                                    
       620         1616       3836 ns     6198 us     FFT length 112                                    
       710         1848       2810 ns     5192 us     FFT length 128                                    
       890         2314       6507 ns       15 ms     FFT length 160                                    
      1100         2857       7328 ns       20 ms     FFT length 192                                    
      1300         3374       7566 ns       25 ms     FFT length 224                                    
      1400         3632       8818 ns       32 ms     FFT length 256                                    
      1800         4666         11 us       53 ms     FFT length 320                                    
      2100         5442         11 us       62 ms     FFT length 384                                    
      2400         6217         16 us      100 ms     FFT length 448                                    
      2800         7251         18 us      136 ms     FFT length 512                                    
      3500         9060         24 us      221 ms     FFT length 640                                    
      4100        10611         25 us      266 ms     FFT length 768                                    
      4700        12162         34 us      423 ms     FFT length 896                                    
      5500        14230         40 us      582 ms     FFT length 1K                                    
      6800        17591         57 us     1009 ms     FFT length 1280                                    
      8100        20951         54 us     1135 ms     FFT length 1536                                    
      9400        24312         79 us     1922 ms     FFT length 1792                                    
     10000        25863         70 us     1833 ms     FFT length 2K

Right now I only have the Mac build posted at sourceforge. The Windows version will be available tomorrow and the Linux version will be ready whenever Steven Harvey has time to work on it.

unconnected · 2013-01-21, 10:24

Code:

-bash-3.2# ./pfgw32s
FATAL: kernel too old
Segmentation fault
-bash-3.2# uname -r
2.6.18-274.7.1.el5.028stab095.1
-bash-3.2# cat /etc/redhat-release 
CentOS release 5.8 (Final)

What minimum requirements for precompiled pfgw 3.7 from Sourceforge?

rogue · 2013-01-21, 13:35

Quote:

Originally Posted by unconnected

Code:

-bash-3.2# ./pfgw32s
FATAL: kernel too old
Segmentation fault
-bash-3.2# uname -r
2.6.18-274.7.1.el5.028stab095.1
-bash-3.2# cat /etc/redhat-release 
CentOS release 5.8 (Final)

What minimum requirements for precompiled pfgw 3.7 from Sourceforge?

I don't know. Steven Harvey (who is on the forum) builds the linux exe.

Dubslow · 2013-01-22, 06:07

Quote:

Originally Posted by unconnected

Code:

-bash-3.2# ./pfgw32s
FATAL: kernel too old
Segmentation fault
-bash-3.2# uname -r
2.6.18-274.7.1.el5.028stab095.1
-bash-3.2# cat /etc/redhat-release 
CentOS release 5.8 (Final)

What minimum requirements for precompiled pfgw 3.7 from Sourceforge?

Huh. It runs on 2.6.38 just fine.

Code:

uname -a
Linux Gravemind 2.6.38-16-generic #67-Ubuntu SMP Thu Sep 6 17:58:38 UTC 2012 x86_64 x86_64 x86_64 GNU/Linux

This is Ubuntu 11.04, which while an older distribution than CentOS 5.8, has a *only slightly* higher kernel. Hmm.

rogue · 2013-01-23, 14:08

I've released 3.7.2 which addresses a memory leak in the tree factoring code.

rogue · 2013-02-11, 18:51

I've released PFGW 3.7.3 over at sourceforge. A problem was found when using -f that caused it to crash. That problem was fixed.

2013-01-07, 14:21	#156
pepi37 Dec 2011 After milion nines:) 1,451 Posts	Now zip from sourceforge is ok, update 3.7.0 at Boinc Confederation C:\Users\Phenom>C:\Users\Phenom\Desktop\pfgw64.exe -d -q"7012^293561+1" PFGW Version 3.7.0.64BIT.20130104.Win_Dev [GWNUM 27.8] 7012^293561+1 is composite: RES64: [E03CF76F4D8B6048] (25.4841s+0.0001s) C:\Users\Phenom>C:\Users\Phenom\Desktop\pfgw32.exe -d -q"7012^293561+1" PFGW Version 3.7.0.32BIT.20130104.Win_Dev [GWNUM 27.8] 7012^293561+1 is composite: RES64: [E03CF76F4D8B6048] (30.4190s+0.0002s) Last fiddled with by pepi37 on 2013-01-07 at 14:21

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 3.2.2 has been Released	rogue	Software	20	2009-08-23 12:14
PFGW 3.2.1 has been released	rogue	Software	5	2009-08-10 01:43

2013-01-07, 13:45	#155
rogue "Mark" Apr 2003 Between here and the 1100011010010₂ Posts	I've updated the zip file. I know that both the 32-bit and 64-bit exes work, so if the zip file is corrupted, it might be a problem created by WinZip (which I used to create the zip file) or the upload.

2013-01-21, 10:24	#161
unconnected May 2009 Russia, Moscow 2593₁₀ Posts	Code: -bash-3.2# ./pfgw32s FATAL: kernel too old Segmentation fault -bash-3.2# uname -r 2.6.18-274.7.1.el5.028stab095.1 -bash-3.2# cat /etc/redhat-release CentOS release 5.8 (Final) What minimum requirements for precompiled pfgw 3.7 from Sourceforge?

2013-01-23, 14:08	#164
rogue "Mark" Apr 2003 Between here and the 1100011010010₂ Posts	I've released 3.7.2 which addresses a memory leak in the tree factoring code.

2013-02-11, 18:51	#165
rogue "Mark" Apr 2003 Between here and the 1100011010010₂ Posts	I've released PFGW 3.7.3 over at sourceforge. A problem was found when using -f that caused it to crash. That problem was fixed.