Unexpected behaviour upgrading 2.1.0 --> 2.2.0

[Federico G Lopez Armengol]

Hi,

I recently upgraded 2.1.0 --> 2.2.0 but found unexpected behaviours. 

1) Problems that used to converge with DENSE_NORMAL_CHOLESKY are now failing because max_num_consecutive_invalid_steps is reached. They do converge if I use DENSE_QR, though. Do you know if this is expected?

2) Mixed precision is no longer compatible with DENSE_QR?

I have already taken care of the LocalParametrization deprecation.

Thanks! Cheers,

Federico

[Sameer Agarwal]

I am surprised by (1).
I forget if mixed precision ever worked with dense qr or not,  but there should not be a regression.

--
You received this message because you are subscribed to the Google Groups "Ceres Solver" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ceres-solver...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ceres-solver/72e2c85a-d165-4ae5-b7c8-ba7f3c511b65n%40googlegroups.com.