v1.2.2

DoubleFloats v1.2.2

Diff since v1.2.1

Merges the LinearAlgebra.floatmin2 fixes from David Gleich (dgleich). Replaces expm1 implementation with Quadmath's implementation which is much more stable for small values. This rectifies the most notable surprise in the code base.

Closed issues:

• plane rotations or givens rotations are slow for Double64 due to Int64 conversion using BigFloats (#128)
• Method definition string(..) overwritten (#146)

Merged pull requests:

• Fixes #128 (#147) (@dgleich)