On Tue, 2017-08-01 at 09:21 +0200, Sebastien Binet wrote: > the other place I had (very recently) felt they could have been > useful: > - having a mat.Dense of math/big.Rat (in lieu of a mat.Dense of > float64)
I think that would be very hard. The reasons behind this claim are that there is lot of non-generic code behind the mat.Dense user-facing implementation; we would need generic BLAS and generic LAPACK, with appropriate testing. Having said that, there is no reason not to implement a big.Rat-based implementation of mat.Matrix for the methods that you need. It would return reduced precision float64 values from the At method, but have a RatAt method returning the exact value. -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
