|
2016-01-15, 15:48
|
#1 |
|
Jan 2016
2·3 Posts |
I'm trying to speedup AES MSIEVE factoring using CUDA build but...
I'm trying to speedup AES MSIEVE factoring using CUDA build but I dont see any success its same speed like with non cuda version.
How to force using Cuda Nvidia GPU? Hello Dr Gladman could you tell me is there special command that I can know that your compiled cuda version of msieve is using cuda at all? Like to show me that GPU is used, and that library is used, I don't see that anything is used. I'm watching GPU utilization and its 0% and I see that my cpu is 4x25% which means I'm using my CPU. I installed Cuda toolkit and I don't have microsoft visual studio (toolkit installation complained about visual studio not present) I dont know do I need to install Visual Studio in order to use msieve cuda? Where did I make mistake for not utilizing my Nvidia GPU? I'm on windows 7 sp1 64bit Here is the paste of the actual msieve: --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- C:\Msieve>msieve -v -e 0x2210FEE51674DC85D6B7CA8CE373518BAE3A6B7A21C70C976B8C26CBF8CF84CB8525B982142C2E5885FD967459E9041E82BB210D562C0FE3B4C035256A871DEF Msieve v. 1.52 (SVN unknown) Thu Jan 14 22:26:43 2016 random seeds: a36c8158 e6eec901 factoring 1784201585035392943382886765412274918625109709538884734462486341468956 46678669551048971047831265739010761443728254596771467332559243468940097505174638 5391 (154 digits) searching for 15-digit factors searching for 20-digit factors searching for 25-digit factors 200 of 214 curves completed 214 ECM curves searching for 30-digit factors 425 of 430 curves completed 430 ECM curves searching for 35-digit factors P+1 stage 1 factor found commencing quadratic sieve (108-digit input) using multiplier of 3 using VC8 32kb sieve core sieve interval: 47 blocks of size 32768 processing polynomials in batches of 5 using a sieve bound of 5375099 (186667 primes) using large prime bound of 806264850 (29 bits) using double large prime bound of 10756225367263650 (45-54 bits) using trial factoring cutoff of 54 bits polynomial 'A' values have 14 factors sieving in progress (press Ctrl-C to pause) 3444 relations (3243 full + 201 combined from 196287 partial), need 186763 --------------------------------------------------------------------------------------------------------------- Its painfully slow I was hoping to boost speed with the Cuda. Here are Aes and Bitcoin keys C:\Tesla>teslacrack.py Cannot decrypt ./file.pdf.vvv, unknown key Software has encountered the following unknown AES keys, please crack them first using msieve: 2210FEE51674DC85D6B7CA8CE373518BAE3A6B7A21C70C976B8C26CBF8CF84CB8525B982142C2E5885FD967459E9041E82BB210D562C0FE3B4C035256A871DEF found in ./file.pdf.vvv Alternatively, you can crack the following Bitcoin key(s) using msieve, and use them with TeslaDecoder: C30C9B81114913F821F91E037718880767020A0748250D38FFEB49C47C9E8334448E59797A95DFFE4702AA5ABB0068BDAD8C7D070B6B441B09A5AC0297C81CC8 found in ./file.pdf.vvv ----------------------------------- I attached the teslacrypt pdf file this should be RSA 4096 key personal identification ID: 794AA2F9BB5C19D2 With the tempo of factoring with msieve it should take 18-30 days to crack the number. Maybe I'm doing something wrong. Thanks in advance |
|
|
|
2016-01-15, 16:00
|
#2 |
|
I moo ablest echo power!
May 2013
22·463 Posts |
You should use YAFU. Follow the steps here: http://www.mersenneforum.org/showthread.php?t=20779
|
|
|
|
|
2016-01-15, 16:43
|
#3 | |
|
Jan 2016
68 Posts |
Quote:
Where is the problem beside missing rels0.dat, should I create empty rels0.dat, what about other errors? Last fiddled with by loopdemack on 2016-01-15 at 16:45 |
|
|
|
|
|
2016-01-15, 17:05
|
#4 |
|
Apr 2013
Durham, UK
1038 Posts |
I've factored it for you anyway - it took less than 3 minutes of ECM using 2 threads
http://factordb.com/index.php?id=1100000000814812720 |
|
|
|
|
2016-01-15, 17:12
|
#5 | |
"Ben"
Feb 2007
7·13·41 Posts |
Quote:
You can do that by putting a line in yafu.ini something like this: ggnfs_dir=C:\location\of\your\sievers |
|
|
|
|
|
2016-01-16, 07:44
|
#6 | ||||
|
Jan 2016
2·3 Posts |
Ok I did edit yafu.ini to something like
Quote:
in C:\Ecm, I extracted ecm644_win64_i7.zip I dont know why the first ecm path has to have ecm.exe and second ecm path doesn't have ecm.exe. I started yafu now and its seems to work painful slow *(0.00966 sec/rel)) but its work now. I think its extremely slow and I'm doing something wrong again, like wrong ecm or ggnfs version or something is wrong with the settings. Also could yafu work with GPU. Here is the yafu output for now Quote:
This is after 45 minutes its still working, I dont know how long it will work. I just want to know what I'm doing and am I doing it right. And when it finish should I expect the same numbers like factordb gave: http://factordb.com/index.php?id=1100000000814812720 Now the second question regarding factordb because I see that rob147147 used number Quote:
Now to return to factordb.com above number is: Quote:
I just want to understand everything because I want to make detailed tutorial, because there is non. |
||||
|
|
|
|
2016-01-16, 11:48
|
#7 |
|
Jan 2016
2×3 Posts |
Ok Yafu finished after 4 hours and some minutes but if I convert seconds i got 4 hours and 25 minutes, anyway eta calculation is dead wrong for almost double first eta was (filtering ETA: 2h 54m)
NFS elapsed time = 15318.3770 seconds. pretesting / nfs ratio was 0.00 Total factoring time = 15318.3820 seconds ***factors found*** P75 = 715992848178912796133932488667760309859728657842078534991619930743921203069 P33 = 439353422321288635964378942226547 ans = 1 Ok as I thought yafu shown only two numbers. |
|
|
|
|
2016-01-16, 14:55
|
#8 |
|
I moo ablest echo power!
May 2013
22×463 Posts |
Just as a heads-up, 0.009 sec/rel is quite fast--invert it and see that you're getting ~111 relations per second. Also, 4 or so hours isn't bad for a C108 number.
The Filtering ETA is an estimated time based on an assumed number of relations needed (here, about 4.5 million). If the assumed number is low, the actual time to finish can be longer. Note as well that the filtering ETA is ONLY for the relation-gathering step. You still have to build the matrix and carry out some linear algebra, which increases your total time. Glad you got everything working. |
|
|
|
|
2016-01-17, 02:49
|
#9 | |
|
Jan 2016
2×3 Posts |
Quote:
How can I know is it working ok? |
|
|
|
|
|
2016-01-17, 03:23
|
#10 | ||
|
Jan 2016
68 Posts |
One more thing yafu did factorized the numbers but it gave just two factors I've searched in the output and there are just those two factors?
Quote:
Quote:
|
||
|
|
|
|
2016-01-17, 16:07
|
#11 |
|
I moo ablest echo power!
May 2013
22×463 Posts |
You fed it a 108 digit number:
Code:
C:\yafu>yafu-x64.exe "factor(0x2210FEE51674DC85D6B7CA8CE373518BAE3A6B7A21C70C976B8C26CBF8CF84CB8525B982142C2E5885FD967459E9041E82BB210D562C0FE3B4C035256A871DEF)" -v -threads 4 nfs: commencing nfs on c108: 314573908204972170772685162835613740501161465222989282532110802340617739158508803543279344012292489789672743 |
|
|
|
 |
| Thread Tools | |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Speedup of trial factoring with prime95/mprime | Anonuser | Information & Answers | 13 | 2014-09-09 01:43 |
| 32-bit CUDA build? | f1pokerspeed | Msieve | 2 | 2013-12-30 01:14 |
| Windows x64 CUDA Build | Brian Gladman | Msieve | 12 | 2013-08-30 12:40 |
| Windows build for GPU CUDA code | Brian Gladman | GMP-ECM | 13 | 2013-05-13 15:00 |
| Mp factoring speedup question. | Fusion_power | Math | 11 | 2004-06-03 08:25 |
