20070422, 03:05  #265  
Jun 2003
1,579 Posts 
Quote:
I am not sure how the pollard Rho will be faster than SPH or baby or giant step method. The wiki page says the run time is the same. Btw, the SPH can be applied to pollard rho also. edit: Also pollard rho cannot be restricted to a range of n. Though a modification can help. Read http://citeseer.ist.psu.edu/cache/pa...1computing.pdf On the other hand Pollard rho can be used for a range of n, if you plan to use SPH, where only the order will be reduced. Though I could be wrong on this one. Read ( http://www.freedc.org/forum/showthr...ht=pollard+rho )post #20 Quote:
I think there are enough values to test for each smooth t. If t is very smooth and the sieve is fast then these 4 trillion values can be tested in the same time it takes to test 12 billion values and find the extra factors. Last fiddled with by Citrix on 20070422 at 03:30 

20070425, 12:50  #266 
May 2005
2^{3}·7·29 Posts 
Do srsieve and sr(x)sieve remove algebraic factors?
I have received following messages while sieving with srsieve: Code:
WARNING: 4*3^n1 has algebraic factors. 
20070425, 15:44  #267  
Dec 2006
41_{8} Posts 
Quote:
Guess those will not be removed automatically. Last fiddled with by thommy on 20070425 at 15:44 

20070426, 04:32  #268 
Mar 2003
New Zealand
2205_{8} Posts 
No they are not removed. Only some simple algebraic factors are detected by srsieve as a consequence of checking whether any subsequences consist of generalised Fermat numbers, there may well be others that are not detected. sr[125]sieve don't check for any algebraic factors.

20070506, 00:25  #269 
Mar 2003
New Zealand
13×89 Posts 
sr1sieve 1.0.23, sr2sieve 1.4.40
Some more improvements to the SSE2 code: mulmods now done in groups of 8 or 16 instead of groups of 4. (32bit SSE2 machines only).
I don't know whether this is faster on anthing other than the Northwood P4, any benchmarks will be a help. Here are some times for my 2.9GHz Northwood P4 at p=100e12: Code:
19k SoB.dat 68k riesel.dat 237k sr5data.txt    1.4.39: 293 kp/s 161 kp/s 77 kp/s 1.4.40: 377 kp/s 194 kp/s 85 kp/s 
20070506, 06:49  #270 
May 2005
11001011000_{2} Posts 
I'm sieving 4*3^n1 using sr1sieve linux.x8664 executable and the speed increased from 5.84M to 5.95M (~2%).
CPU = C2D 4300 @ 2.4 GHz Last fiddled with by Cruelty on 20070506 at 06:51 
20070506, 13:43  #271 
Sep 2004
B0E_{16} Posts 
I'm sieving 43046721*2^n+1 using sr1sieve and sieve speed decreased from 30.6 Mp/s to 29.2 Mp/s.
CPU = AMD 64 3000 2.0 GHz 
20070507, 08:12  #272 
Apr 2003
2^{2}·193 Posts 
Sieving on a PSP only (11K) SoB.dat with a Pentium M (Centrino) with 1.6GHz.
Version 1.4.39: around 350kp/sec Version 1.4.40: Around 333kp/sec By the way jjsieve on the same range/dat file combination 331kp/sec 
20070507, 12:28  #273 
I quite division it
"Chris"
Feb 2005
England
31·67 Posts 
Speed up of about 4% on Core2Duo E4300 running sr1sieve on both cores.
(k = 55 and k = 105) 
20070508, 01:44  #274 
Mar 2003
New Zealand
13·89 Posts 
Thanks for the benchmarks everyone, I don't have any solutions for those suffering a slowdown yet, so I'll keep the old versions available for download.
I have discovered that the Windows binary is reporting elapsed time when it should be reporting CPU time. I have fixed this in sr2sieve version 1.4.41. In sr2sieve 1.4.42 I have implemented some of the ideas from the SSE2 code for other 32bit machines. It is faster on my P3 but no change on P2, probably because of the slow L2 cache on the P2. Here are some times for my 600MHz Coppermine P3 at p=100e12: Code:
19k SoB.dat 68k riesel.dat 237k sr5data.txt    1.4.40: 92 kp/s 53 kp/s 30 kp/s 1.4.42: 100 kp/s 58 kp/s 31 kp/s Code:
length = 1000, iterations = 10000, b = 2, p = 4611686018427387817: : 57.812 million mulmods per second. CMOV: 54.953 million mulmods per second. SSE2: 106.400 million mulmods per second. 
20070508, 04:33  #275 
Apr 2003
772_{10} Posts 
Benchmarks for a P3, the Pentium M and a P4 HT 3.06GHz will come on thursday . I have no time now as i am on my way to England to visit a customer the next two days.
Lars 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Very Prime Riesel and Sierpinski k  robert44444uk  Open Projects  587  20161113 15:26 
Sierpinski/ Riesel bases 6 to 18  robert44444uk  Conjectures 'R Us  139  20071217 05:17 
Sierpinski/Riesel Base 10  rogue  Conjectures 'R Us  11  20071217 05:08 
Sierpinski / Riesel  Base 23  michaf  Conjectures 'R Us  2  20071217 05:04 
Sierpinski / Riesel  Base 22  michaf  Conjectures 'R Us  49  20071217 05:03 