A GPU version of Enigma software

Message boards : Number crunching : A GPU version of Enigma software

To post messages, you must log in.

Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 · Next

AuthorMessage
Helix Von Smelix

Send message
Joined: 31 Oct 07
Posts: 36
Credit: 1,198,682,898
RAC: 0
Message 5956 - Posted: 20 Jul 2017, 14:42:41 UTC - in response to Message 5955.  

Are there issues with work distribution? I haven't gotten any new wu's since this morning.

Same here
ID: 5956 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile TJM
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 25 Aug 07
Posts: 843
Credit: 267,994,998
RAC: 0
Message 5959 - Posted: 20 Jul 2017, 17:02:56 UTC - in response to Message 5956.  

It's fixed already, I did a stupid mistake patching the work generator to support test app and it placed all the workunits in the wrong queue.
M4 Project homepage
M4 Project wiki
ID: 5959 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
JugNut

Send message
Joined: 16 Mar 13
Posts: 24
Credit: 125,506,046
RAC: 0
Message 5960 - Posted: 20 Jul 2017, 17:30:02 UTC - in response to Message 5959.  

Still nothing....
ID: 5960 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alex VE3NEA

Send message
Joined: 22 Oct 16
Posts: 36
Credit: 128,745
RAC: 0
Message 5961 - Posted: 20 Jul 2017, 17:35:59 UTC

A critical bug in enigma-cuda was found by @matszpk - thank you! Because of this bug, there was an 1/256 chance of missing the correct solution. The bug is now fixed, please download the new executables.
ID: 5961 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Empie

Send message
Joined: 5 Mar 10
Posts: 12
Credit: 40,797,951
RAC: 0
Message 5962 - Posted: 20 Jul 2017, 18:03:41 UTC

Do we have to recheck the results again due to this bug? Glad it's found.
ID: 5962 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alex VE3NEA

Send message
Joined: 22 Oct 16
Posts: 36
Credit: 128,745
RAC: 0
Message 5963 - Posted: 20 Jul 2017, 18:12:17 UTC

Please wait a minute, that fix is not final.

I don't think there is a need to re-check anything because each candidate key is tried many times, by doing multiple restarts.
ID: 5963 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Empie

Send message
Joined: 5 Mar 10
Posts: 12
Credit: 40,797,951
RAC: 0
Message 5964 - Posted: 20 Jul 2017, 18:30:26 UTC

Ok, clear. Thanks.
ID: 5964 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile TJM
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 25 Aug 07
Posts: 843
Credit: 267,994,998
RAC: 0
Message 5970 - Posted: 20 Jul 2017, 21:41:05 UTC - in response to Message 5963.  

Please wait a minute, that fix is not final.

I don't think there is a need to re-check anything because each candidate key is tried many times, by doing multiple restarts.


Let me know when the app is ready to update, I've prepared release 1.10 that's just waiting for new executables.
M4 Project homepage
M4 Project wiki
ID: 5970 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alex VE3NEA

Send message
Joined: 22 Oct 16
Posts: 36
Credit: 128,745
RAC: 0
Message 5976 - Posted: 21 Jul 2017, 18:42:38 UTC

It looks like my yesterday's posting did not pass through. The bug is fixed now, and the new build is ready for download.
ID: 5976 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
oh2hyt

Send message
Joined: 14 Jul 09
Posts: 53
Credit: 705,427,365
RAC: 0
Message 5995 - Posted: 24 Jul 2017, 21:05:42 UTC

I noticed following results at front page. How come these have different score?

g3_test100_5: score: 2172362 settings: B:TO:MDV:423:ADEHGYIMKNLROZQVTXUW
anxpanzxgruppexvierxsiegfriedsiegfriedtonixdivxstehtseitxeinszwoxsiebenxeinseinsnullnullxuhrmitanfae found by europe64
g3_test100_6: score: 33562 settings: B:TO:MDV:423:ADEHGYIMKNLROZQVTXUW
anxpanzxgruppexvierxsiegfriedsiegfriedtonixdivxstehtseitxeinszwoxsiebenxeinseinsnullnullxuhrmitanfae found by mimeq
ID: 5995 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alex VE3NEA

Send message
Joined: 22 Oct 16
Posts: 36
Credit: 128,745
RAC: 0
Message 5996 - Posted: 25 Jul 2017, 1:53:48 UTC - in response to Message 5995.  

Perhaps these two WU's use different dictionaries. The score is computed based on the n-gram frequencies in those dictionaries.
ID: 5996 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Werinbert

Send message
Joined: 22 Jul 13
Posts: 7
Credit: 10,011,525
RAC: 0
Message 6020 - Posted: 31 Jul 2017, 0:05:55 UTC

Ok, I set my computer to get the new GPU tasks. Three were downloaded and immediately failed. As these are the test apps do I need to download some additional software (something was mentioned along that lines a while back)?

Link to one task: http://www.enigmaathome.net/result.php?resultid=286264924
ID: 6020 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Empie

Send message
Joined: 5 Mar 10
Posts: 12
Credit: 40,797,951
RAC: 0
Message 6038 - Posted: 2 Aug 2017, 19:29:40 UTC - in response to Message 6020.  
Last modified: 2 Aug 2017, 19:34:06 UTC

Ok, I set my computer to get the new GPU tasks. Three were downloaded and immediately failed. As these are the test apps do I need to download some additional software (something was mentioned along that lines a while back)?

Link to one task: http://www.enigmaathome.net/result.php?resultid=286264924


They fail with exit code 195. This has been mentioned one page back in this topic. I see them also, but most of the wu's go well for me.
ID: 6038 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Werinbert

Send message
Joined: 22 Jul 13
Posts: 7
Credit: 10,011,525
RAC: 0
Message 6047 - Posted: 3 Aug 2017, 0:29:00 UTC - in response to Message 6038.  

Thanks Empie for the response and likely solution. I set Enigma to no new tasks thinking the problem was my computer. I have now resumed getting tasks and they don't seem to be getting the 195 error.
ID: 6047 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Skivelitis2
Avatar

Send message
Joined: 16 May 15
Posts: 7
Credit: 10,001,739
RAC: 0
Message 6068 - Posted: 3 Aug 2017, 14:55:12 UTC

All GPU apps immediately failing with exit code -148 (can't run app).
Win XP SP3 32 bit....GT 730 1GB.
Any ideas?
ID: 6068 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
matszpk
Project tester
Volunteer developer
Volunteer tester

Send message
Joined: 20 May 11
Posts: 10
Credit: 105,914,725
RAC: 0
Message 6211 - Posted: 11 Aug 2017, 17:27:50 UTC

Hi. I wrote enigma-opencl optimized in assembler for Radeon GPU's. First tentative timings are 2x or more faster than pure OpenCL version.
Today or tomorrow, I publish this optimized version in github :)
ID: 6211 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
matszpk
Project tester
Volunteer developer
Volunteer tester

Send message
Joined: 20 May 11
Posts: 10
Credit: 105,914,725
RAC: 0
Message 6212 - Posted: 11 Aug 2017, 19:01:48 UTC

Optimized for GCN (Radeon GPU's) enigma-opencl is here: https://github.com/matszpk/enigma-cuda/tree/gcn-experim
ID: 6212 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alex VE3NEA

Send message
Joined: 22 Oct 16
Posts: 36
Credit: 128,745
RAC: 0
Message 6213 - Posted: 11 Aug 2017, 19:46:08 UTC - in response to Message 6212.  

You may want to include my fix to the bug that you found, your own fix does not fully solve the problem. Here is the code:

https://github.com/VE3NEA/enigma-cuda/blob/master/enigma-cuda-lib/cuda_code.cu#L698-L719
ID: 6213 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
matszpk
Project tester
Volunteer developer
Volunteer tester

Send message
Joined: 20 May 11
Posts: 10
Credit: 105,914,725
RAC: 0
Message 6233 - Posted: 12 Aug 2017, 18:54:12 UTC - in response to Message 6213.  
Last modified: 12 Aug 2017, 19:11:18 UTC

VE3NEA: Thank you for advice. I apply your changes to my version to CUDA code. However changes in OpenCL code is obsolete (in my opinion after further analysis). Code find best result correctly and store correct index of result. If I forgot about some things (about my fix), please explain me.
ID: 6233 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Alex VE3NEA

Send message
Joined: 22 Oct 16
Posts: 36
Credit: 128,745
RAC: 0
Message 6236 - Posted: 13 Aug 2017, 1:17:05 UTC - in response to Message 6233.  

The problem with the original code was that Result.index was used to both point to the solution in the array and to find the rotor positions for which that solution was found. After your fix, the code finds the best result, but, since it changes the index, the rotor positions are lost. After my fix, the index is never changed.
ID: 6236 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 . . . 6 · 7 · 8 · 9 · 10 · 11 · 12 · Next

Message boards : Number crunching : A GPU version of Enigma software




Copyright © 2024 TJM