GPU sieving drive part II: k<10000 n=2M-3M

[mdettweiler]

Hi all,

Our first "bigger and better GPU sieving drive", for k<10000 and n<2M, has been completed to p=75T and handed off to PrimeGrid BOINC which is making astounding progress on it (~60T/day! ). The priority for manual sieving is now on the n=2M-3M range, which is as before being done as a joint effort between NPLB and PrimeGrid's PST forum.

The difference this time is that we'll be doing both the Riesel and Proth sides (k*2^n-1 and k*2^n+1) in the same sieve. Recent innovations in tpsieve (the twin-prime cousin of ppsieve) allow it now to sieve non-twin searches concurrently on the +1 and -1 sides without any speed loss! We will, therefore, be using tpsieve for this sieve instead of ppsieve. The two programs are very similar and if you're familiar with one the other should be a breeze.

This sieve will likely go into BOINC at some point as well; currently the idea is to do this at p=75T as with the 2M sieve. This will of course depend on whether BOINC is yet done with the 2M sieve at that point, though at the rate they're going there's a good chance they will be.

As before, the PST forum's reservation thread will be the "primary" one; I will reserve ranges from there for NPLB and distribute them here. The reservations table for NPLB's ranges ONLY follows below; see PST's thread for the overall canonical list.

P-range       reserved by     status
  0T- 20T     PST forum
 20T- 21T     vaughan         complete
 21T- 30T     PCZ             complete
 30T- 50T     PST forum
 50T- 70T     PCZ             complete
 70T- 90T     PST forum
 90T-110T     PCZ             complete
110T-140T     PST forum
140T-143T     PCZ             complete
143T-145T     Mathew Steine   complete
145T-160K     PCZ             complete

This post will be followed shortly with instructions for running tpsieve on this range. Please read them anew as they will be slightly different than the old ppsieve instructions.

Max

[mdettweiler]

To participate in this sieve, follow these steps:

-Download the appropriate version of tpsieve for your computer.
All tpsieve versions can be downloaded at ppsieve software as before. The CPU version of tpsieve is bundled along with CPU ppsieve (ppsieve-bin-*.zip). For CUDA, download "tpsieve-cuda-*.zip".
Note: CUDA users will need the cudart-2.0 libraries in a place where tpsieve can see them. They are from a slightly older-than-latest version of the CUDA toolkit, so they probably didn't come with your installed CUDA toolkit. For Windows, download cudart.dll and plop it in your tpsieve folder. For Linux, download libcudart.so.2 for 32-bit or 64-bit and stick it in your CUDA libraries directory (probably /usr/local/cuda/lib/ or /usr/local/cuda/lib64/--at least that's where it is on Gary's machine).

There is not currently an OpenCL/ATI version of tpsieve; hopefully one will be available soon. (If you're a real diehard, you might try running ppsieve twice, one each for the Proth and Riesel side...it will be twice as slow but it's at least something. Let me (Max) know if you need help on how to do that.)

-Extract tpsieve to its own folder somewhere.

-Reserve a range from this thread.

-Run the appropriate command line for your tpsieve version and range. Quoted from John at the PST forum:

This is a dat-less sieve. Remember to modify factor file name to ppr3M_##G-##G.txt

tpsieve(version) -p##G -P##G -k 5 -K 9999 -n 2M -N3M -ffppr3M_##G-##G.txt -t# -M2 -q

-When the range is complete, upload your factors to http://pgllr.mine.nu/sieves/ppse/factors/. If you have trouble uploading, email your factors to max@noprimeleftbehind.net and I'll get them to where they need to go.

-Rinse, lather, and repeat!

Max

[vaughan]

20000G - 21000G Reserved by vaughan

[vaughan]

20000G - 20500G in progress

20500G - 20500G completed and uploaded as zip file

[mdettweiler]

Originally Posted by vaughan

20000G - 20500G in progress

20500G - 20500G completed and uploaded as zip file

Uh...might you mean 20T-20.5T is complete and 20.5T-21T is in progress?

[vaughan]

I'm running 20T - 20.5T on a slower GPU a GTX260. I don't know its progress yet as I selected the *boinc* application exe not the non-boinc one. If I get a chance I'll log in remotely and stop and restart the client to check progress.

The faster GTX460 has already finished the other half of the range 20.5T - 21T.

[PCZ]

I have been sieving the k7001 - k7999 on GPU's for a while now.
Seems a bit pointless though if that range is covered already by efforts such as this.
I can pull the ranges of interest out of the combined files, assuming they remain available.
So my GPU's would be better utilised helping in the initial sieves before they go out on the grid.

Anyways i will stop rambling and get on to reserving an initial range.

PCZ
21T-24T

[vaughan]

20000G - 20500G completed and zip file uploaded

[mdettweiler]

Originally Posted by PCZ

I have been sieving the k7001 - k7999 on GPU's for a while now.
Seems a bit pointless though if that range is covered already by efforts such as this.
I can pull the ranges of interest out of the combined files, assuming they remain available.
So my GPU's would be better utilised helping in the initial sieves before they go out on the grid.

Anyways i will stop rambling and get on to reserving an initial range.

PCZ
21T-24T

Yes, the files will be publicly posted throughout the sieving drive and at its conclusion. The sieve files have thusfar been posted in the PST forum; see the bottom of this thread for the latest n<2M sieve file at any given time. The 2M-3M sieve has not yet progressed far enough to have any useful sieve files posted, but they will be similarly posted in the PST forum.

When each of these sieves has reached optimal depth I'll download the final file and mirror it on the noprimeleftbehind.net server.

[PCZ]

23T-24T Finished
Found 1493625 factors
Uploaded the file successfully, i think !!

21T-23T will be finished in about 3hrs.




Reserving 24T-27T

[mdettweiler]

Originally Posted by PCZ

23T-24T Finished
Found 1493625 factors
Uploaded the file successfully, i think !!

21T-23T will be finished in about 3hrs.




Reserving 24T-27T

Yep, it made it--just not quite to the right place. You put it in http://pgllr.mine.nu/sieves/ppse/, instead of http://pgllr.mine.nu/sieves/ppse/factors/. You'll want to re-upload it to the latter location (since that's where the PST folks will be looking for the factors).