Preparing N=333333

[biwema]

Originally Posted by MooooMoo

edit: There needs to be some kind of system (maybe a factor verification script?) to deter cheaters. This is quite important, since part of the credit for finding the twin will go to the top siever. Suggestions on preventing/deterring cheating are welcome.

Yes, that is a good point.
If we synchronize the sieving effort by exchanging the factors removed (that is already supported by NewPGen), they can quickly be checked when merging.
That way we can also prevent destroying the candidate file in case someone tries to merge a factor file of a different exponent. (NewPGen adds everything without any information about the exponent into the NewPGen.del file even if there were already factors of a previous search inside. verifying could sort out the valid factors if the user mix up information here)

[smh]

I would like to reserve:

205T - 215T

[MooMoo2]

As many of you have probably noticed, there is a new thread to reserve ranges for sieving: http://www.mersenneforum.org/showthread.php?t=6858

Enjoy.

[smh]

How do you create a new input file using only the removed numbers file?

NewPgen has the option to merge files, but it looks like this uses only output files

BTW, anyone has any luck running newpgen on a dual processor machine (not just with this input file, but any sieving)?

Once i start the sieving, the program just exits. This happend on 3 computers (all Windows 2003). I tried two cingle cpu machines, which both worked fine. (only 1 is sieving though)

[VBCurtis]

smh-
I run two instances of NewPGen on a Core2Duo by starting one copy, going into task manager and setting affinity of that instance to CPU1 only; then starting a second instance, setting its affinity to CPU0. I had difficulty like yours when I did not set affinity. Also note that LLR can set affinity in a menu within the program; running LLR on CPU1 and NewPGen without setting affinity works fine.

To set affinity within task manager (at least in XP Pro), right-click on a process from within the "processes" tab.
Happy Hunting.
-Curtis

[gribozavr]

Just out of interest, I started sieving on my Celeron. I got DSM_LogDeleted = 1 in newpgen.ini, but I see no newpgen.del file. Anyway, removing rate was 5.8 k/sec.

Newpgen was about 1.5-2 times slower (factors checked/sec) than at 195000. For n=195000 25G range newpgen says it uses fastarray, and for n=333333 100G it uses array. Maybe that is the cause?

[biwema]

Originally Posted by gribozavr

Just out of interest, I started sieving on my Celeron. I got DSM_LogDeleted = 1 in newpgen.ini, but I see no newpgen.del file. Anyway, removing rate was 5.8 k/sec.

Newpgen was about 1.5-2 times slower (factors checked/sec) than at 195000. For n=195000 25G range newpgen says it uses fastarray, and for n=333333 100G it uses array. Maybe that is the cause?

What version / Operating system are you using?

My tests (Windows XP, Vers 2.82, SSE2) always says using array.
Both 195000 and 333333 and a small array as well as a large array
(0.3 KByte for tests, 292.5MB for the 100G file)

How much are you sieving per second (increasing of p per second)?

[gribozavr]

$ uname -a
Linux server 2.4.27-custom8jul2005 #1 Птн Июл 8 11:07:36 EEST 2005 i686 GNU/Linux
$ cat /proc/cpuinfo
processor       : 0
vendor_id       : GenuineIntel
cpu family      : 6
model           : 11
model name      : Intel(R) Celeron(TM) CPU                1200MHz
stepping        : 1
cpu MHz         : 1202.751
cache size      : 256 KB
fdiv_bug        : no
hlt_bug         : no
f00f_bug        : no
coma_bug        : no
fpu             : yes
fpu_exception   : yes
cpuid level     : 2
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 mmx fxsr sse
bogomips        : 2398.61

Sieving 2700M-25G n=195000 says "using array", uses 192Mb of RAM, rate is 16777200 p/sec. I used the following command to capture two screenshots delayed by 1 sec:

(import -window root s1.bmp &); sleep 1; import -window root s2.bmp

[R. Gerbicz]

Originally Posted by gribozavr

Sieving 2700M-25G n=195000 says "using array", uses 192Mb of RAM, rate is 16777200 p/sec.

Are you using Newpgen version 2.82 ?

Another thing: 16777200 is about 2^24=16777216, so this isn't the correct p/sec. time probably. It would be a little better to sieve say 10G range and from times get the p/sec rate=10000000000/used_seconds is a good approximation.

Yesterday I've done also a test, the two speeds were almost same.
But I think that in theory n=333333 can be slower by 5-7% if it is using one more squaremod operation than n=195000.

[ValerieVonck]

Ok I have downloaded the zip file, how do I set a range to sieve from to
in NewPgen?

Thank you

[biwema]

I don't know the exact difference between the windows and linus version of NewPGen. Maybe that can explain some issues.
For example 192M array for a 25G range is quite a lot (the 100G range needs 292.5M ) maybe there is a different type of array used under linux if there is plenty of memory. (As already discussed before, the time difference should not differ that much if p is large)

Also windows prints an output every 2^24. so sieving some G is more accurate. I have around 80M/s with 3.4GHz P4. Probably some AMD architectures are even faster.

Set a range:
Start of the range: best use a Hex editor and set the starting value into the first line. Normally a Hex editor does not need to load the whole file into memory and is therefore very fast.

End of the range:
Newpgen provides an option to set a manual p max. It needs to be set before the sieving process starts otherwise it does not stop there.