20160210, 12:45  #12 
Sep 2002
Database er0rr
F17_{16} Posts 
Okay, I think I can replace those routines with calls to gwtobinary32 and binary32togw and use some array arithmetic to achieve the mod calculation...
Last fiddled with by paulunderwood on 20160210 at 12:54 
20160212, 06:06  #13  
Sep 2002
Database er0rr
3,863 Posts 
Quote:
Last fiddled with by paulunderwood on 20160212 at 06:08 

20160216, 00:46  #14  
"Mark"
Apr 2003
Between here and the
2×5×643 Posts 
Quote:


20160216, 00:46  #15 
"Mark"
Apr 2003
Between here and the
2×5×643 Posts 
Completed both to n=400,000 and continuing.

20160216, 02:53  #16 
P90 years forever!
Aug 2002
Yeehaw, FL
7632_{10} Posts 

20160216, 18:37  #17 
Sep 2002
Database er0rr
3863_{10} Posts 
I have got the time down to 5:40 minutes (for a buggy program). This means the array arithmetic is taking 2:10 minutes, and I only have 3040 seconds to play with. I really doubt I can beat Georges generic modular reduction.

20160229, 03:03  #18 
"Mark"
Apr 2003
Between here and the
2×5×643 Posts 
Add (2^442042+1)^22 as a new prime!

20160325, 23:37  #19 
"Mark"
Apr 2003
Between here and the
6430_{10} Posts 
Completed to 475,000. Nothing new, but continuing

20160407, 22:01  #20 
"Mark"
Apr 2003
Between here and the
14436_{8} Posts 
Completed to 505,000. Nothing new, but continuing

20160408, 22:10  #21 
"Mark"
Apr 2003
Between here and the
191E_{16} Posts 
I have a command line siever that works, but it slower than MultiSieve for base 2. I couple of reasons that it is slower is that I have generic logic for base 2 and poorly coded hash map code. Fortunately I think the non base 2 code is much faster the MultiSieve. The negative with MultiSieve is that it misses some factors, but that isn't a big enough issue to use the new code (yet).

20160412, 01:05  #22 
"Mark"
Apr 2003
Between here and the
2×5×643 Posts 
Here is a Windows build of my sieving code. I (ahem) borrowed parts of the discrete log code from srsieve so I won't release the source until I clean it up and remove references to that. This code runs about 10% faster than MultiSieve (for p < 1e9), unless you are logging factors. That slows it down a bit based upon my testing, but as all factors are doublechecked, that shouldn't be a big deal. I've stated before MultiSieve misses 10% to 15% of the factors due to some bug, but as cksieve is faster there is no reason to continue using MultiSieve for sieving nearsquare primes. One core should be able to sieve about 150G for a range of 500,000 n per day. Unlike MultiSieve, this code is base agnostic, so sieving shouldn't be impacted when sieving a larger base. There is multithreading code in the source, but I know it doesn't work so I don't know if I'm going to try to fix it or not.
Last fiddled with by rogue on 20200924 at 19:47 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Carol / Kynea Primes  rogue  And now for something completely different  253  20211015 08:39 
Carol / Kynea Coordinated Search  Reservations  rogue  And now for something completely different  294  20210830 08:07 
Search primes of form 2*n^n ± 1  JeppeSN  And now for something completely different  27  20180412 14:20 
Factorial primes search?  flava  Open Projects  18  20101204 05:24 
Why Search for these Huge Primes?  Unregistered  Math  8  20050427 00:55 