[QUOTE=pinhodecarlos;307899]Greg released, in 30 Jul 2012, the Linux 14e Lattice Sieve for AMD x86_64 or Intel EM64T CPU for NFS@Home, why not adding it to RSALS? Also available the following versions of the siever:

Mac OS 10.4 or later running on Intel
Intel 64-bit Mac OS 10.5 or later
FreeBSD on Intel-compatible 64-bit[/QUOTE]

Short answer: see [url]http://www.mersenneforum.org/showpost.php?p=306539&postcount=434[/url]

Longer answer: RSALS and NFS@Home use differing custom code to pass parameters to the siever. Our binaries aren't quite compatible.

I pointed [B]Mumps [MM][/B] to mersenneforum.org because if I understood well his private message he wants to help on post-processing. I told him I could help him under windows environment but not on linux.
What are the commands to run msieve under linux? Or how should I transmit to him how to run msieve under linux when I am a windows guy?

Getting back to the Cunningham composites in post 459 ( [url]http://www.mersenneforum.org/showpost.php?p=307754&postcount=459[/url] )...

Is
[code]n: 491296955542200529296298766407559388556005204618800597588204335685546761247869465840040321782365870638927430080063823667120739625660666450613330547424846830585129145599149537144369187250964237
m: 30912680532870672635673352936887453361
type: snfs
c6: 121
c3: 11
c0: 1
skew: 0.449644313022609
rlim: 67108863
alim: 67108863
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.6
alambda: 2.6[/code]

the proper polynomial for 11,327-, and

[code]n: 175936771974981772283182692394074213578398696418629500563272697258850587853092454731715044160330704227661645146926522496559416459706894451885803992947006494311389000244584321313611309094964002385703577602867598966770499
m: 30912680532870672635673352936887453361
type: snfs
c6: 121
c3: -11
c0: 1
skew: 0.449644313022609
rlim: 67108863
alim: 67108863
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.6
alambda: 2.6[/code]

the proper polynomial for 11,327+ ?

I produced that manually, from the algebraic factorizations 11^327-1 = (11^109-1)*(11^218+11^109+1) and 11^327+1 = (11^109+1)*(11^218-11^109+1). m=11^36.

[QUOTE=debrouxl;307950]Getting back to the Cunningham composites in post 459 ( [url]http://www.mersenneforum.org/showpost.php?p=307754&postcount=459[/url] )...[/QUOTE]
Yes, both look good.

[QUOTE=pinhodecarlos;307948]I pointed [B]Mumps [MM][/B] to mersenneforum.org because if I understood well his private message he wants to help on post-processing. I told him I could help him under windows environment but not on linux.
What are the commands to run msieve under linux? Or how should I transmit to him how to run msieve under linux when I am a windows guy?[/QUOTE]

I don't think a precompiled linux version is available, so he needs to download the source and compile it. If his computer has the appropriate development tools, then

svn co [url]https://msieve.svn.sourceforge.net/svnroot/msieve/trunk[/url] msieve
cd msieve
make x86_64 LARGEBLOCKS=1

If he runs into problems that zlib isn't available, then in Ubuntu do

sudo apt-get install zlib1g-dev
make clean
make x86_64 LARGEBLOCKS=1

Otherwise, he can compile without zlib support with

make clean
make x86_64 LARGEBLOCKS=1 NO_ZLIB=1

Once he has the msieve binary, the remaining steps are the same as Windows. Download the number.dat.gz, number.ini, and number.fb files from RSALS or NFS@Home. If the binary doesn't have zlib support, gunzip the number.dat.gz file. Then use

./msieve -s number.dat -l number.log -i number.ini -nf number.fb -nc -t 4 -v

assuming 4 cores. Adjust as necessary.

Plus, if he has a memory-challenged computer (by today's standards, that is, say just 4Gb; some of my old ones are! nothing to be ashamed of), then he would want to build another binary; both may be useful at different times:
[CODE]...
mv msieve msieveLB
make clean
make x86_64
mv msieve msieveSB
[/CODE]

Another thought: if [FONT=Courier New]grep ZLIB_VERS /usr/include/zlib.h[/FONT] will show [FONT=Courier New]#define ZLIB_VERSION "1.2.3[/FONT]", then either update zlib (this is of course a preferred solution) or don't use it (make x86_64 NO_ZLIB=1)

[QUOTE=Batalov;307960]Another thought: if [FONT=Courier New]grep ZLIB_VERS /usr/include/zlib.h[/FONT] will show [FONT=Courier New]#define ZLIB_VERSION "1.2.3[/FONT]", then either update zlib (this is of course a preferred solution) or don't use it (make x86_64 NO_ZLIB=1)[/QUOTE]

Ubuntu supplies
#define ZLIB_VERSION "1.2.3.4"

Does this count? :smile:

That's not good :-) (it does "count" 1. 2. 3. 4. but that's exactly how that deprecated library works with streams - byte by byte. 1. 2. 3. 4.)

>=1.2.5 is good. (internally works with reasonable chunks; that's pretty much the only patch; the later versions are all good)

His machines are a bunch of:

[LIST][*]AuthenticAMD AMD Opteron(TM) Processor 6220 [Family 21 Model 1 Stepping 2] (16 processadores)[/LIST][LIST][*]AuthenticAMD Six-Core AMD Opteron(tm) Processor 8435 [Family 16 Model 8 Stepping 0] (24 processadores)[/LIST] [LIST][*]AuthenticAMD Six-Core AMD Opteron(tm) Processor 2435 [Family 16 Model 8 Stepping 0] (12 processadores)[/LIST]all with 64 Gig of RAM.

In that case I recommend using the MPI version. Do the computers have an MPI implementation installed? If so, use

make x86_64 MPI=1 LARGEBLOCKS=1 NO_ZLIB=1
then unzip the dat file and run the postprocessing in 3 stages:
./msieve -s number.dat -l number.log -i number.ini -nf number.fb -nc1 -v
mpirun -np 24 ./msieve -s number.dat -l number.log -i number.ini -nf number.fb -nc2 4,6 -v
./msieve -s number.dat -l number.log -i number.ini -nf number.fb -nc3 -v

For the first and third stages, you can run the MPI binary on a single core without using mpirun. For the second stage, use
mpirun -np 12 ... -nc2 3,4 -v for 12 cores
mpirun -np 16 ... -nc2 4,4 -v for 16 cores
mpirun -np 24 ... -nc2 4,6 -v for 24 cores

Those are interesting computers :smile:
No wonder the grid's power jumped up, if he attached several of those, and others attached many / large computers as well...

In the Debian unstable repositories, I see two MPI implementations: mpich2 and openmpi (preferably 1.6, I guess).