Posts by Slicker

Mersenneplustwo home page
log in
1) Message boards : Number crunching : Windows App (Message 63)
Posted 4764 days ago by Profile Slicker
GMP is faster than MPIR. The only place it would make sense to use MPIR is for a 64-bit Windows build since that is the only platform on which MPIR is faster than GMP for the WEP algorithm, but since Windows apps are not planned at this time, it makes sense to stick with GMP. I couldn't check MPIR on the MAC OS X 64-bit version because I don't have a 64-bit MAC on which to run it, but I assume it would be similar to Linux where GMP is king.
2) Message boards : Number crunching : Windows App (Message 62)
Posted 4765 days ago by Profile Slicker
Exactly what I was looking for. I'm investigating whether MPIR which branched off from GMP is any faster than GMP. I haven't built it on Linux or OS/X before and since compiler options tend to cause GMP to give improper results at times, I'm guessing the same is true for MPIR. That means the more data with which to test, the better.

Have you considered a 64-bit MAC version since "Intel 64-bit Mac OS 10.5 or later" is 64 bit? I'm guessing that the 64-bit GMP is considerably faster than the 32-bit version and that Leopard/Snow Leopard would get a big boost from running the 64-bit version. Having a 64-bit version may attract more MAC users.
3) Message boards : Number crunching : Windows App (Message 60)
Posted 4765 days ago by Profile Slicker
So far so good. The output matches when no factors are found. Is there a previously run WU that found a factor that I can run in standalone mode so I can check that as well?
4) Message boards : Number crunching : Windows App (Message 59)
Posted 4766 days ago by Profile Slicker
I'm running some standalone tests now to make sure they come up with the same results as the Linux version.

Also, to get the BOINC percentage of the task complete to show up in BOINC manager:

in factorize64.C add :
extern int boinc_fraction_done(double);

and then in factor_using_random_wep, within the main loop add:

if (numtrials%100 == 0) //check every 100 iterations
{
boinc_fraction_done((double)numtrials / (double)T);
}

The boinc fraction done takes very little cpu time, so updating it every 100 verses 1000 iterations isn't going to slow it down noticeably and users will appreciate being able to tell how long the task will run before completion.
5) Message boards : Number crunching : Windows App (Message 57)
Posted 4767 days ago by Profile Slicker
I have a suspicion that the WEP-M+2 uses a variation of the following for the application:

random base web

If that is the case, it should be fairly easy to create a Windows version of the app. If you are interested, please send me a PM. I may be able to assist as the initial Collatz applications used GMP so I have a Windows w/ GMP environment already set up.




Return to WEP-M+2 Project main page


Copyright © 2023 M+2 Group