|
2009-03-03, 10:30
|
#1 |
Apr 2007
Spessart/Germany
2·34 Posts |
curious timings running ECM curves
Hello,
atm my computer run ECM curves with B1=11e6 at M2539 and M2543. For M2539 prime95 uses FFT 112. One curve needs ~225 sec for stage 1 and ~76 sec for stage 2. For M2543 prime95 uses FFT 128. One curve needs ~183 sec for stage 1 and ~72 sec for stage 2. Is it possible to run prime95 with the larger FFT size at the smaller M2539, too? I could save ~40 sec per curve, and I want to run 2000 curves or so. Matthias |
|
|
|
2009-03-03, 14:45
|
#2 |
|
P90 years forever!
Aug 2002
Yeehaw, FL
19×397 Posts |
You can force the larger FFT size. After the long assignment ID (or the string "N/A") in worktodo.txt insert the text ",FFT2=128" assuming this is an SSE2 machine. Use ",FFT=128" for non-SSE2 machines.
|
|
|
|
|
2009-03-03, 18:44
|
#3 |
|
Apr 2007
Spessart/Germany
101000102 Posts |
Thank you, this works fine. Indeed the curves about 40 sec faster now.
 best regards, Matthias |
|
|
|
 |
| Thread Tools | |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| New proggy and timings | axn | Operation Billion Digits | 1 | 2009-02-06 16:14 |
| Need GMP trial-division timings | ewmayer | Factoring | 7 | 2008-12-11 22:12 |
| 321 LLR timings | paulunderwood | 3*2^n-1 Search | 14 | 2008-04-17 22:27 |
| Curious about what people are running..... | schickel | Factoring | 6 | 2007-12-10 18:01 |
| Weird LLR FFT timings | MooooMoo | Riesel Prime Search | 2 | 2007-10-11 08:56 |
