Msieve 1.43 feedback

[Jeff Gilchrist]

Originally Posted by 10metreh

This used to be the "matrix must have more columns than rows" error, but I think Jason added this feature, where it is detected beforehand, in 1.43.

I have also see this message when building a matrix on data that was *way* too oversieved. So I needed to cut down the # of relations to filter and then it had enough cycles.

So you might be too low or too high.

[Andi47]

Originally Posted by Jeff Gilchrist

I have also see this message when building a matrix on data that was *way* too oversieved. So I needed to cut down the # of relations to filter and then it had enough cycles.

So you might be too low or too high.

With 17.5M unique relations I was still tooo low ("filtering wants 1M more relations"), so I don't think that the gap between "undersieved" and "too oversieved" is THAT narrow.

[jasonp]

I think oversieving to the point of failure is pretty much impossible for all but the smallest problems. When the input is small and hugely oversieved, it's possible to delete almost all of the relations during clique removal. Hey, you want a sparse matrix, and you can't get more sparse than an empty matrix :)

[Jeff Gilchrist]

Originally Posted by Andi47

With 17.5M unique relations I was still tooo low ("filtering wants 1M more relations"), so I don't think that the gap between "undersieved" and "too oversieved" is THAT narrow.

No, I was just pointing out that I have also seen that message when things were vastly oversieved. So for some people that might be their problem if they saw it.