PFGW 4.0.4 (with gwnum v30.10) Released

[rogue]

You can d/l 3.6.0 from here.

The changes include:

- Upgraded Windows builds to use MPIR 2.4.0 and renamed header to mpir.h, not wingmp.h.
- When using the number_primes feature, read pfgw.log and pfgw-prime.log to build list of values to be skipped from the ABC input.

I hope to include gwnum v27.x in the next release.

[rogue]

I've released a Mac only build of pfgw 3.6.1 at sourceforge. This build will address the issue of the 32-bit pfgw hanging on Mac. AFAICT, this does not affect pfgw64 or pfgw32 on other platforms. It appears to be due to a miscompile of GMP 5.0.1 (using llvm-gcc). I changed my environment to use clang, which correctly builds GMP 5.0.2 (make check failed when using llvm-gcc).

[rebirther]

If someone is using pfgw in any project with AVX and without Bulldozer at the moment pls test this one here built with gwnum 27.3. Speedup is around 39-56% (2-4 cores) on a Sandybridge. 32/64bit for windows are available. Its working great with SR5 project.

Pls report any time difference / speedup vs. the older app + project. Thx!

http://www.bc-team.org/downloads.php?cat=7

[rogue]

You can d/l 3.6.2 from here.

As many have been waiting, this is the official release with gwnum v27, using gwnum v27.4. There are no other changes.

I've only posted the windows build. The Mac build will follow later today and hopefully Steven Harvey can get the linux build done soon as well.

I'm looking for someone who is interested in doing some coding for pfgw. pfgw is lacking checkpointing for primality tests. This is an issue when trying to prove primality of large GFNs, which can take a number of days. I've been too busy with other development and life in general to work on it.

[Harvey563]

The Linux builds are done and available at the same location.

Steven

[rogue]

The Mac builds are now posted as well.

[rogue]

If anyone has been running a non-official release of pfgw with gwnum v27. I have some bad new for you. You probably have to re-run tests with the official release. gwnum v27 (and especially gwnum v27.2) has bugs. AFAIK, all of them (minus Bulldozer) have been addressed, which is why I released 3.6.2.

Please PM (or e-mail) me and I'll try to help you determine the scope of the problem, i.e. determine which tests you need to re-run.

[rogue]

You can d/l 3.6.3 from here.

This has AMD Bulldozer support as it uses gwnum v27.5. Nothing else has changed.

I've only posted the windows build. The Mac build will follow later today and hopefully Steven Harvey can get the linux build done soon as well.

[Harvey563]

The linux versions 3.6.3 are now at Sourceforge too.

StevenHarvey

[rogue]

Mac builds are now available as well.

[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.

SCRIPT

DIM n, 21-1
DIM k, 1
DIM max_k, 1000000
DIM max_n, 1000
DIM pfound, 0
DIM mfound, 0
DIM both, 0
DIMS tmpstr

OPENFILEAPP p_file,plus.txt
OPENFILEAPP m_file,minus.txt
SETS tmpstr,ABC $a*2^$b-1 & $a*2^$b+1 & $a*2^$b+5
WRITE p_file,tmpstr
SETS tmpstr,ABC $a*2^$b-1 & $a*2^$b+1 & $a*2^$b-5
WRITE m_file,tmpstr

LABEL next_n
SET n, n+1
IF (n>max_n) then GOTO end
SET k, -1
SET pfound, 0
SET mfound, 0
SET both, 0
LABEL next_k
SET k, k+2
IF (k>max_k) then GOTO next_n
IF (both) then GOTO next_n
SETS tmpstr,%d*2^%d-1;k;n
PRP k*2^n-1,tmpstr

IF ((ISPRP+ISPRIME)==0) then GOTO next_k
SETS tmpstr,%d*2^%d+1;k;n
PRP k*2^n+1,tmpstr

IF ((ISPRP+ISPRIME)==0) then GOTO next_k

LABEL plus
IF (pfound) then GOTO minus
SETS tmpstr,%d*2^%d+5;k;n
PRP k*2^n+5,tmpstr
IF (ISPRP+ISPRIME) then GOSUB write_p_file

LABEL minus
IF (mfound) then GOTO next_k
SETS tmpstr,%d*2^%d-5;k;n
PRP k*2^n-5,tmpstr
IF (ISPRP+ISPRIME) then GOSUB write_m_file

GOTO next_k

GOTO next_n

LABEL write_p_file
SETS tmpstr,%d %d;k;n
WRITE p_file,tmpstr
SET pfound, 1
IF (mfound) then SET both, 1
RETURN

LABEL write_m_file
SETS tmpstr,%d %d;k;n
WRITE m_file,tmpstr
SET mfound, 1
IF (pfound) then SET both, 1
RETURN

LABEL end

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.