GCP-ECM Pentium IV Cygwin binary : Usage question
 

Hello,
I downloaded this from :
[url]http://www.geocities.com/greatpsycho/[/url]

which is linked from :

[url]http://www.loria.fr/~zimmerma/records/ecmnet.html[/url]

From a DOS command line, it doesn't seem that the -c option is used in the calculation although the usage indicates [I]-c n[/I]: performs n runs for each input:
[I]
C:\math\ecm>ecm5 -c 100 3e6 < testFactor2.txt
GMP-ECM 5.0.3 [powered by GMP 4.1.2] [ECM]
Input number is 17267061499138678408786079717404735228863374683687241293 (56 digits)
Using B1=3000000, B2=4016636514, polynomial Dickson(12), sigma=2948287101
Step 1 took 24435ms
Step 2 took 28731ms

C:\math\ecm>ecm5 -c 10000 3e6 < testFactor2.txt
GMP-ECM 5.0.3 [powered by GMP 4.1.2] [ECM]
Input number is 17267061499138678408786079717404735228863374683687241293 (56 digits)
Using B1=3000000, B2=4016636514, polynomial Dickson(12), sigma=2881529914
Step 1 took 24144ms
Step 2 took 28962ms[/I]

But it is in the usage section and is accepted as input. I was able to get it to factor the number so I wonder if there's a default "c" which is being used. The logging doesn't indicate anything even with -v on.


Has anyone used this version?

Any help would be appreciated,
Thanks,
Grandpa

Hi, I found that if you are redirecting from a file then it must be in unix format with CR at the end of the line. If there is a CR and an LF or if there is nothing at all then it doesn't work.

That should be just LF at the end. "\n"

[QUOTE=marc]That should be just LF at the end. "\n"[/QUOTE]
Thanks :-) In other words the file needs to have an end of line character and be in unix format and not in DOS format.

Thanks folks. I believe I've got it working.

It seems a bit slow (I'm was running only PARI at the same time and killed that but the speed wasn't altered much).

It takes about 4 minutes per curve running with B1 = 3e6.

I'm using a Pentium M, 1.6Ghz w/ 512Mb RAM.

Any hints would be appreciated.

Thanks,
Grandpa

On a Pentium-M 1.7 GHz, it only takes 50 seconds. According to your results above, you also only need 53 seconds.

btw:
[code]GMP-ECM 5.0.3 [powered by GMP 4.1.2] [ECM]
Input number is 17267061499138678408786079717404735228863374683687241293 (56 digits)
Using B1=3000000, B2=4016636514, polynomial Dickson(12), sigma=802453689
Step 1 took 23423ms
Step 2 took 27630ms
********** Factor found in step 2: 71237436024091007473549
Found probable prime factor of 23 digits: 71237436024091007473549
Probable prime cofactor 242387464553038099079594127301057 has 33 digits
[/code]

:smile:

242387464553038099079594127301057 is indeed prime, btw

Sorry for the confusion, I switched to a 180-digit number:

278490841076279407188741439481565179355926725853710201632331620642982062389901741890579963524423782637435949041666525000702723914662388812510545494307250950777886431051612811386531

(the 43rd term in the Euclid-Mullin sequence).



-Grandpa

Oh, I re-verified a similar metric for the 56 digit number Myst Walker.

I guess 4 minutes is how long it should take on a machine like this.

I think the Pentium M core is more similar to a Pentium 3 than to a Pentium 4. Maybe try a P3 binary and see if it is faster.

Alex

Indeed the Pentium M core is similar to the Pentium 3 core, only extended with some extra's like sse2 if i am not mistaken and some extra registers. But NOT the netburst technologie.

I made private binaries for the P-M as well as the P3, they don't differ much. But here in this topic you can read how to compile easily your own binary of GMP-ECM, with depending on cygwin and its dll's...

[url]http://www.mersenneforum.org/showthread.php?t=3089[/url]



ps..
You can download the bin's here (they will stay atleast a week or so on this spot), altho they work perfectly on my systems i cannot give any certainty that they work on yours ;)

[url]http://1202.org/CarolKynea/ECM503_K7.exe[/url]
[url]http://1202.org/CarolKynea/ECM503_P3.exe[/url]
[url]http://1202.org/CarolKynea/ECM503_P4.exe[/url]
[url]http://1202.org/CarolKynea/ECM503_PM.exe[/url]

Thanks for the links and advice, bot. I tried all but the K7 exe and the metrics were slightly slower.

If anyone's interested, I installed the P4 cygwin binary on my desktop (which is a P4 1.8 Ghz with a skimpy 256Mb of RAM) and have seen slightly slower figures (Step1 is a bit faster, Step2 about 25% slower) than the numbers for the Pentium M. I guess it's just the Cygwin binary that isn't so fast.

I have to build this myself and see what happens.