Rechenkraft.net e.V.

I have stopped the work generator because IMO it does not make sense to continue the batch that runs with the original ("naval") dictionary set, where upgraded dictionary sets have already failed with similar number of target restarts.

IMO it's better to try with another dictionary set and tweaked app.
Both cases, however, require major upgrades on the server side - for performance reasons, I have removed some of the functionalities it originally had and now I have to add them back. It's going to take 2-4 weeks because my job leaves me very little spare time. I'm also going to clean the database from any leftovers, there are still over 20k workunits in the db, probably most of them are broken WUs that accumulated over time (at some point there was a sudden failure of SSD drive where the database filesystem is located, it was replaced by RAID5 setup with the data restored from backup, but ~4500 workunits/results ended in messed up state). I'll wait a few days and then all the orphaned workunits / results will be removed.

Also, all future (larger) batches will be split into a group of smaller sub-batches. While the batch is running, the server requires access to all the results that were returned and if the database gets to big, performance suffers. It also makes creating 'offline' backups harder due to the amount of data in the db dump.

In a few days I'll post downloadable results from the previous batches in case anyone wants to take a look.

I have just finished a GPU version of the Enigma breaking software. I am still testing the program and exploring the possibilities of further speed improvement, but since TJM is now working on a major upgrade on the server side:

http://www.enigmaathome.net/forum_threa ... d=835#4420

- it might be a good time to include support of GPU in Enigma@home, so I am releasing the 64-bit Windows binaries of my program for testing.

The software is here:

https://github.com/VE3NEA/enigma-cuda

To see if it works, just run the included run.bat file.

On my Nvidia GTX 1070, this program breaks the well-known PBNXA message (181 char) in 10 minutes by iterating over all wheel orders.

The source code will be released once I finish my experiments and clean it up.

If you have a descent Nvidia card, please give this software a try and let me know if there are any issues. Also, please let us all know how much time it takes the program to break PBNXA on your video card.


Cheers,

Alex Shovkoplyas VE3NEA

The server requires BOINC server code upgrade to support GPU app,
Per-app credits, application selection and a couple of other things.

I have scheduled the upgrade for this weekend, the server will be stopped on saturday, in worst case it might even go completely offline for a couple of hours.

It used app_info + tweaked client configuration to run. There was only one tester other than me.

Later today I'll post the app_info that requires no mods on the client side, combined with a first batch of test workunits. This will be the first batch with credits set somewhere near the production workunits. Until now all the GPU test workunits had credits set to 1 or 2 per task, just to see if validator works.