Theodore Omtzigt

@davidmallasen got some new info. As you know the `posit` arithmetic tests exhaustively trace 4B combinations. To make these regressions complete in a reasonable time I use randoms just to...

@davidmallasen oh, this is so embarrassing! I went through all the history of this issue that @RaulMurillo started. Raul offered to do the implementation as the posit is now the...

@davidmallasen 1. Dang, thanks for finding that bug. I vaguely remember that I was trying to implement `posit` and `posit` as our application profiling showed these posits are good replacements...

@davidmallasen the specialized posits are following the softposit algorithms with the proper C++ skeleton around it to create a plug-in type.

@davidmallasen This commit is a new baseline to work from: https://github.com/stillwater-sc/universal/commit/cac65dca3dd535dcbe9573ecf88b4e1d507ef05c I discovered that the fast `posit` has regressed and is failing as well. Need to RCA why that has...

In this commit: https://github.com/stillwater-sc/universal/commit/025e51f4961e47bb6103906b751840a073b252d3 I have added a `posito`, short for posit oracle that we can use to RCA the specialized posit failures. you can bring in this type via:...

@anubane Thank you for your contribution. We have been looking for a Numpy integration for half a decade, so this work is much appreciated. We have been looking at cppyy...

@RaulMurillo the intent of the specialized bfloat16 is to have a fast bfloat16 that is emulated through the native FP32. The cfloat parameterization contains all the special formats used in...

@RaulMurillo use *bfloat16* in any performance-sensitive code. The storage is a fixed _uint16_t_ and the arithmetic operators will use the floating-point hardware. use *bfloat_t* in comparison codes that are evaluating...

TBD: we need to expand this to the big cfloats through a native implementation.