langou
langou
@jprhyne just pushed his implementation in #1020 I am sorry for the duplication of work. I did not expect it. It seems like both implementations are suffering from corner cases....
Related, not necessarily urgent, but it would be good to push the change to more subroutines than only ORM2R. For example GEQR2 as the same "guilty" lines: https://github.com/Reference-LAPACK/lapack/blob/dd2e5efd8f1a209bbd4a82abf4d2922c8ba1b7db/SRC/dgeqr2.f#L184-L191 Let us...
> are we expecting any breaking changes in the behavior of 'un/orm2r'? No. Behavior of 'un/orm2r' should be unchanged. 'un/orm2r' will be thread-safe though, so you will be able to...
> are we expecting any breaking changes in the behavior of 'larfgp' as a result? Why are you mentioning 'larfgp'? (1) Are you using it? - I am being curious,...
> the 1l with a "firstv" variable I had not thought about this. Oh my. @jprhyne is making a good point and proposing to introduce a new feature that is...
Issue resolved by #1020 and #1019. Thanks @EduardFedorenkov and @jprhyne
I am not a big fan of sign(0) = 0. I think sign(0) = 1 sounds great to me.
Thanks Thijs. I am worried to have the sign function have a different behavior whether this is +0 or -0. I think a single behavior ( sign(0)=+1 ) for +0...
> If we test the backward error, there should be no conditioning right? Correct.
Good question, @thijssteel. There is the theory and there is the practice. When we work out the theory, we assume that "the internal calculations are done in extended precision and...