SIQS - problem with solving linear algebra

[Dome]

Originally Posted by VBCurtis

Dome-
Your english is better than many on this forum, even somewhat regular posters. You write quite well, and the important parts about code or math are very clear. We're glad you used the forum, and I enjoy threads like this as it keeps reminding me that learning an algorithm like SIQS is within my intellectual reach if I find a cure for (personal) laziness.

Thank you, your compliment means a lot to me! Yeah, laziness is my big enemy too. I'm glad that there is community that is interested in factorization, because sometimes it's really hard to find the correct answers at the articles or books.

[bsquared]

Originally Posted by Dome

So true, sometimes it's making me crazy or totally hopeless, strong will is really required. :-D

Well, you factored pretty big number with your SIQS. I'll take it as a challenge! I'll try to get as closer as possilble, but still a lot of work to do.

Sincere good luck to you on that challenge! You will need to move past Gaussian solvers and understand block Lanczos or Wiedemann if you're going to shoot for C100+. Or at least understand it enough to adopt Jason's awesome implementation in msieve, like I and others have done .

[xilman]

Originally Posted by bsquared

Sincere good luck to you on that challenge! You will need to move past Gaussian solvers and understand block Lanczos or Wiedemann if you're going to shoot for C100+. Or at least understand it enough to adopt Jason's awesome implementation in msieve, like I and others have done .

RSA-129 was done with MPQS and structured Gauss back in 1994 because the other algorithms weren't available then.

That said, BL or BW is indeed the way to go these days for C100+ numbers.

[Dome]

Originally Posted by bsquared

Sincere good luck to you on that challenge! You will need to move past Gaussian solvers and understand block Lanczos or Wiedemann if you're going to shoot for C100+. Or at least understand it enough to adopt Jason's awesome implementation in msieve, like I and others have done .

That's true, if I want to factorize numbers like 100+ digits big, Gaussian elimination has to be changed. I implemented Guassian only because it's easy to implement and it's easy to understand it. SIQS is my project for the master's thesis and I need to implement it quickly, because I have only one semester to make it working, do measurements and describe my implementation, so I'm in rush (At my university we work on our thesis only two semesters, first semester we are learning the theory, and in the second one we are implementing the practical part). But after school I would like to make my SIQS better.