Theodore Omtzigt

Added arbitrary posit conversions to get a feel for the promotion. I was worried about the geometric region, but all is good.

with respect to add_exact/sub_exact. The regions between [0, minpos] and [maxpos, NaR] do not have a representation for `r`. ``` Posit TwoSum validation for posit geometric rounding region a :...

I am experimenting in the https://github.com/stillwater-sc/hpr-blas repo as we doing a lot of math experiments for application requirements, so I already had Knuth's twoSum algorithm implemented there. The test is...

When you have a posit with es > 0, you end up having three distinct 'regions' of discretization: 1. the regime, which are blocks of 2^2^es size 2. the exponent,...

so for adding minpos + minpos, for posit configurations that have exponent bits, the sum is minpos and you would not have a value that is smaller than minpos. In...

Would you be able to give me a sorting algorithm that you want to work with this add_exact, so that I can use it as a target for the development...

Implemented a direct method for sqrt. It consists of three direct iterates with some exponent manipulation and range reduction. However, it requires high precision posits to generate the same rounding...

iterative algorithms can use the quire to calculate a high precision residual without rounding. This residual can be used to inform the next iterate. It is unclear if leveraging a...

@mmoelle1 release v3.34.1 added a math library for the fixpnt number system, including a complex.hpp with real(), imag(), and conj() functions. https://github.com/stillwater-sc/universal/releases/tag/v3.34.1 Let me know if this solves your use...

@mmoelle1 has this issue been resolved for application?