Re: Mersenne: Latest version p95v21.exe ??
At 06:35 PM 8/14/2002 -0400, Nick Glover wrote: >I'm curious, where does the roundoff checking every iteration come >in? This explains how the program chooses the FFT size, but what are the >conditions for when roundoff checking for every iteration is used? Based >on my understanding, it seems to me that if the soft crossover calculation >determines that an exponent can't be run at the smaller FFT size, then >maybe it should just do the exponent at that FFT size with roundoff >checking every iteration instead of moving to a larger FFT size. Does the >program currently do this? If prime95 decides to use the smaller FFT size (the thousand iterations average error was below the 0.241 to 0.243 threshold), then the LL test uses the smaller FFT size with error checking every iteration. This will protect you from roundoff errors up to 0.6. _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Latest version p95v21.exe ??
I'm curious, where does the roundoff checking every iteration come in? This explains how the program chooses the FFT size, but what are the conditions for when roundoff checking for every iteration is used? Based on my understanding, it seems to me that if the soft crossover calculation determines that an exponent can't be run at the smaller FFT size, then maybe it should just do the exponent at that FFT size with roundoff checking every iteration instead of moving to a larger FFT size. Does the program currently do this? Thanks. At 17:55 14/08/02 -0400, George Woltman wrote: >Now the gotcha. In v22.8, FFT crossovers are flexible. If you test an >exponent within 0.2% of a crossover point, then 1000 sample iterations >are performed using the smaller FFT size and the average roundoff >error calculated. If the average is less than 0.241 for a 256K FFT or >0.243 for a 4M FFT, then the smaller FFT size is used. > >Brian Beesley has been a great help in investigating revised crossover >points and analyzing the distribution of round off errors. We noticed >that consecutive exponents can have a pretty big difference in average >roundoff error (e.g. one exponent could be 0.236 and the next 0.247). >This is why I elected to try the flexible approach described above. The >0.241 to 0.243 average was chosen hoping for about 1 iteration in a >million generating a roundoff error above 0.4. We might change the 0.241 >to 0.243 constants with more data - it is hard to get enough data points >to accurately measure 1 in a million occurrences. > >One downside is the server does not know which FFT size is used and >will credit you based on the v21 x87 crossovers. Thus, if you are a lucky >person, you might get "bonus" CPU credit where you test the exponent >at a smaller FFT size and the server credits you based on the larger >FFT's timing. Nick Glover [EMAIL PROTECTED] Computer Science, Clemson University "It's good to be open-minded, but not so open that your brains fall out." - Jacob Needleman _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Latest version p95v21.exe ??
At 08:11 PM 8/13/2002 -0700, Anurag Garg wrote: > Is it possible to find out what the FFT crossovers for x87 CPUs >are with the latest version? Yes and no. This table shows FFT size, v21 x87 crossover, v22.8 x87 crossover, v21 SSE2 crossover, v22.8 SSE2 crossover. As you can see v22 is more liberal with x87 crossovers and more conservative with SSE2 crossovers. 262144 5255000 5255000 5185000 5158000 327680 652 6545000 6465000 6421000 393216 776 7779000 769 7651000 458752 904 9071000 897 8908000 524288 1033103810241018 655360 1283128912721265 786432 1530153415161507 917504 1785178917661755 1048576 2040204620182005 1310720 2535253925092493 1572864 3015301929922969 1835008 3510352034863456 2097152 4025403039783950 2621440 5000500249354910 3145728 5940595158925852 3670016 6910693668656813 4194304 7930793078367791 Now the gotcha. In v22.8, FFT crossovers are flexible. If you test an exponent within 0.2% of a crossover point, then 1000 sample iterations are performed using the smaller FFT size and the average roundoff error calculated. If the average is less than 0.241 for a 256K FFT or 0.243 for a 4M FFT, then the smaller FFT size is used. Brian Beesley has been a great help in investigating revised crossover points and analyzing the distribution of round off errors. We noticed that consecutive exponents can have a pretty big difference in average roundoff error (e.g. one exponent could be 0.236 and the next 0.247). This is why I elected to try the flexible approach described above. The 0.241 to 0.243 average was chosen hoping for about 1 iteration in a million generating a roundoff error above 0.4. We might change the 0.241 to 0.243 constants with more data - it is hard to get enough data points to accurately measure 1 in a million occurrences. One downside is the server does not know which FFT size is used and will credit you based on the v21 x87 crossovers. Thus, if you are a lucky person, you might get "bonus" CPU credit where you test the exponent at a smaller FFT size and the server credits you based on the larger FFT's timing. _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Latest version p95v21.exe ??
At 10:41 PM 8/13/2002 +, Russel Brooks wrote: >I'm adding a new pc running GIMPS and went to the web site where >p95v21 was available. Isn't this version a bit downlevel? Try version 22.8 at ftp://mersenne.org/gimps/p95v228.zip. This is the last version 22 beta. If all goes well it will be the official release in a few weeks. Let me know if you have problems. _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Latest version p95v21.exe ??
I'm adding a new pc running GIMPS and went to the web site where p95v21 was available. Isn't this version a bit downlevel? I'm running a PIII 866MHz, I'll upgrade Prime95 if a later version is better. Cheers... Russ DIGITAL FREEDOM! --> http://www.eff.org/ _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
