Yafu

[mnh001]

Concerning this threading, does the program automatically thread to multiple cores or is it something that one has to set up oneself beforehand?

[bsquared]

Originally Posted by mnh001

Concerning this threading, does the program automatically thread to multiple cores or is it something that one has to set up oneself beforehand?

Invoke the program with the -threads N switch.

E.g.:

yafu-32k-Win32.exe "siqs(rand(80))" -threads 4

[mnh001]

Ah, I see. Thanks.

[em99010pepe]

Originally Posted by bsquared

Available here.

I've had hardly any time to spend on yafu development lately, so bchaffin took it upon himself to redo the threading architecture in SIQS. He contributed a more efficient architecture and linux code, I kludged in a windows equivalent, and the result is much better siqs threaded performance. I've seen 30% or more speed improvement with all 4 cores of a phenom II vs. previous versions. bchaffin reports similar numbers.

em99010pepe, I'm curious to see if this solves the "losing cores" problem you've been seeing.

bsquared, I still didn't have time and free machines to test it but I expect by next Wednesday to run a few small composites with Yafu. Is there a new flag command that I should be aware to force all cores?

[bsquared]

Originally Posted by em99010pepe

bsquared, I still didn't have time and free machines to test it but I expect by next Wednesday to run a few small composites with Yafu. Is there a new flag command that I should be aware to force all cores?

No, it's the same as before... use -threads N where N is the number of cores in the processor.

[Andi_HB]

Wow Yafu1.25 (yafu-64k-x64) pushed the Crossover point on my
Core i3-350M to 102.3 digits running 1 thread.
With 2 threads the crossover point is at 107 digits.
Looks like the tune command only use 2 threads for QS but not for nfs if it`s configured in the yafu.ini ?

With 1 thread:
tune_info=Intel(R) Core(TM) i3 CPU M 350 @ 2.27GHz,WIN64,2.91156e-005,0.205855,19.3537,0.0747647,102.274,2323.04

Regards Andi_HB

[bsquared]

Originally Posted by Andi_HB

Wow Yafu1.25 (yafu-64k-x64) pushed the Crossover point on my
Core i3-350M to 102.3 digits running 1 thread.
With 2 threads the crossover point is at 107 digits.
Looks like the tune command only use 2 threads for QS but not for nfs if it`s configured in the yafu.ini ?

With 1 thread:
tune_info=Intel(R) Core(TM) i3 CPU M 350 @ 2.27GHz,WIN64,2.91156e-005,0.205855,19.3537,0.0747647,102.274,2323.04

Regards Andi_HB

That sounds about right for 1 thread on a win64 system. And yes, you should run the tune function single threaded otherwise it won't be a fair comparison to NFS (which is always single threaded within tune()).

You might also get better QS performance if you use the 32k version of yafu on your core i3 instead of the 64k version. If not, that would be interesting to know about.

[Andi_HB]

I Have tested now the Core I3 with the 32k Version.
It`s a little bit faster - the crossover point is now nearly 103 digits.

tune_info=Intel(R) Core(TM) i3 CPU M 350 @ 2.27GHz,WIN64,3.4573e-005,0.202483,17.6735,0.0748184,102.962,2214.85

[bsquared]

Great! It's a little faster, as expected.

[Andi_HB]

Is it possible to start nfs in yafu only with a msieve.dat.p ?
Yafu was searching a poly and i have stoped the polysearch.
After restart yafu with nfs(same N) it begins the polysearch from beginning.
But i have saved the msieve.dat.p before.

[bsquared]

Originally Posted by Andi_HB

Is it possible to start nfs in yafu only with a msieve.dat.p ?
Yafu was searching a poly and i have stoped the polysearch.
After restart yafu with nfs(same N) it begins the polysearch from beginning.
But i have saved the msieve.dat.p before.

Yes, but unfortunately not in an automated way.

You will have to find a polynomial in the .p file to use and manually create a ggnfs.job file.

Restarting with your N value should then see the .job file and commence sieving.

Here is an example ggnfs.job file:

n: 1303393746190602410130747675575270928834473492253595455992458315153116912083217001313161399
skew: 1065989.62
Y0: -8584497716068613471624
Y1:  324846034081
c0: -136172357314030026532285545
c1: -198415467127858630068
c2:  237666829470893
c3:  70578472
c4:  240
rlim: 1260000
alim: 1260000
lpbr: 25
lpba: 25
mfbr: 50
mfba: 50
rlambda: 2.300000
alambda: 2.300000

Starting from the rlim: line down are parameters you will have to supply. The rest come from the .p file for your chosen polynomial.

I've put it on my todo list to be able to restart with a .p file only.