Re: Mersenne: Optimizing P4 usage
- Original Message - From: "Gerry Snyder" <[EMAIL PROTECTED]> To: "mer" <[EMAIL PROTECTED]> Sent: Monday, January 14, 2002 5:37 AM Subject: Mersenne: Optimizing P4 usage >At home I have 2 PC's, both of which devote their spare cycles to GIMPS. >One is a 1.3 GHz P4 running W98, and the other a dual 1 GHz P3 running >linux. One of the P3's is doing LL testing, and the other is doing >mostly ECM, with a little trial factoring thrown in. If I'm not mistaken, you'd be better to have one of your dual processors doing just TF. [...] > My question is whether it is worth the trouble to shift the trial and > P-1 factoring of the next one to one of the P3 processors (the non-LL > one). It might lead to one extra LL test in two years. Definitely move the TF to your P3. A P4 doing integer work is wasted. > The worktodo file for the P4 has: > > Test=current,68,0 Why hasn't your current assignment had a P-1? You should suspend testing to do this. My system with 459MB available memory would choose B1=375000, B2=8156250 (5.17% chance of success) for a 33M first-time assignment (i.e a successful P-1 saves two LL tests) and B1=18, B2=297 (3.78%) for a 33M doublecheck assignment. Your system - with a different amount of available memory - might choose different limits, but whatever they are, you should choose limits somewhere between them, closer to the first-time, or doublecheck values according to how far through the first time check you are. > Test=next,60,0 > > It should be noted that the W95 machine does not have enough RAM for > phase 2 of the P-1 factoring, but the linux machine does. Do P-1 on whichever machine can give it the most memory, and always give it as much as you can. 340MB or thereabouts is optimal for a 6-7M exponant. GOK how much a 33M exponant would take. > Any suggestions? If you've got less than 512MB on your linux box, email the exponant(s) to me, (after you've TFed it) and I'll P-1 it for you. I can do a 6-7M P-1 in about 3 hours, so I'd guess it'd take a couple of days. > TIA, Gerry Daran G. _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Optimizing P4 usage
On 13 Jan 2002, at 21:37, Gerry Snyder wrote: > At home I have 2 PC's, both of which devote their spare cycles to GIMPS. > One is a 1.3 GHz P4 running W98, and the other a dual 1 GHz P3 running > linux. One of the P3's is doing LL testing, and the other is doing > mostly ECM, with a little trial factoring thrown in. > > The only part of Prime95 that uses the SSE2 P4 instructions is the LL > testing. Because of the huge speedup this gives, I would like to keep > the P4 machine doing nothing but LL tests. It is now about a month away > from finishing its first 10 megadigit candidate. > > My question is whether it is worth the trouble to shift the trial and > P-1 factoring of the next one to one of the P3 processors (the non-LL > one). It might lead to one extra LL test in two years. You can do even better than that: Suppose the three processors are as follows: A wants to run trial factoring - if you run LL or P-1 on this, B slows down due to memory bus contention B doesn't mind running P-1 C wants to run LL only. Firstly you want to make sure that C doesn't have SequentialWorkToDo=1 in prime.ini (else TF & P-1 will be run before the current LL completes). Also make C's DaysOfWork greater than the time estimated to run TF & P-1 on A & B. Let C get a new assignment Test=,N1,P and let N2 be the required factoring depth for that exponent. Edit the new assignment in C's worktodo.ini, replacing N1 by N2 and P by 1. If N1 < N2, trial factoring is required. Edit A's worktodo.ini file, adding a new line at the top Factor=,N1. Stop & restart mprime on A to get the assignment started. If a factor is found, remove the assignment from C's worktodo.ini file & let C get a new assignment. Whatever happens, let A report the result of the trial factoring. If A fails to find a factor and P=0, P-1 factoring is required. Edit B's worktodo.ini file, adding a new line at the top Pfactor=,N2,0. Stop & start mprime on B to get the assignment started. Let B report the result of the P-1 factoring. If a factor is found, remove the assignment from C's worktodo.ini file & let C get a new assignment. Regards Brian Beesley _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Re: Mersenne: Optimizing P4 usage
Hi, At 09:37 PM 1/13/2002 -0800, Gerry Snyder wrote: >At home I have 2 PC's, > >My question is whether it is worth the trouble to shift the trial and >P-1 factoring of the next one to one of the P3 processors You only need to do the trial factoring on the P3. P-1 testing on the P4 also uses the SSE2 instructions. As to whether this is worth the trouble, only you can answer that question! -- George _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
Mersenne: Optimizing P4 usage
Hello, Gurus: At home I have 2 PC's, both of which devote their spare cycles to GIMPS. One is a 1.3 GHz P4 running W98, and the other a dual 1 GHz P3 running linux. One of the P3's is doing LL testing, and the other is doing mostly ECM, with a little trial factoring thrown in. The only part of Prime95 that uses the SSE2 P4 instructions is the LL testing. Because of the huge speedup this gives, I would like to keep the P4 machine doing nothing but LL tests. It is now about a month away from finishing its first 10 megadigit candidate. My question is whether it is worth the trouble to shift the trial and P-1 factoring of the next one to one of the P3 processors (the non-LL one). It might lead to one extra LL test in two years. The worktodo file for the P4 has: Test=current,68,0 Test=next,60,0 It should be noted that the W95 machine does not have enough RAM for phase 2 of the P-1 factoring, but the linux machine does. Any suggestions? TIA, Gerry -- mailto:[EMAIL PROTECTED] Gerry Snyder, AIS Director & Symposium Chair, Region 15 RVP Member San Fernando Valley, Southern California Iris Societies in warm, winterless Los Angeles--USDA 9b-ish, Sunset 18-19 my work: helping generate data for: http://galileo.jpl.nasa.gov/ _ Unsubscribe & list info -- http://www.ndatech.com/mersenne/signup.htm Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers