View Single Post
Old 2013-06-05, 13:55   #7
WraithX
 
WraithX's Avatar
 
Mar 2006

23·59 Posts
Default

I have a few questions about command line options and parameters for msieve 1.51 with CUDA.

1) I know stage1_norm is used in -np1. Does it affect -nps or -npr?

2) I know stage2_norm is used in -nps and -npr. Does it affect -np1?

3) From the documentation, it looks like -t X is only used by -np1 and not -nps or -npr. Is this correct?

4) I've read that running -np1 multiple times, with a fixed set of arguments, will usually not produce the same results (because the search space is large and is split into pieces and random pieces are searched). I'm wondering if running -nps or -npr multiple times, with a fixed set of arguments, will always produce the same results? Or could those produce different polynomials because of different random internal states in msieve?

5) Are Murphy E scores directly comparable between degree 5 and degree 6 polynomials? I think not, but I want to be sure.

Thanks everyone for the links to help guide me on parameter selection. I have some ideas of what values to use, and I'll be test sieving those probably by the end of this month.

BTW, here are some of the best poly's I've found so far for my C210. I'm going to go back and try tweaking the norm's in -nps and -npr to see if I get anything better.
Code:
Degree-5 Polynomials:
skew 291528643.68, size 1.051e-20, alpha -7.678, combined = 9.961e-16 rroots = 3
skew 104279094.33, size 1.124e-20, alpha -6.911, combined = 1.038e-15 rroots = 5
skew 609572156.15, size 9.604e-21, alpha -9.019, combined = 9.234e-16 rroots = 5

Degree-6 Polynomials:
skew 1564743.15, size 2.403e-15, alpha -10.513, combined = 9.028e-016 rroots = 2
WraithX is offline   Reply With Quote