high-performance-computing

Heston model has accurate density approximations for European option prices, which are of interest.

The module implementing this method should live under tf_quant_finance/volatility/heston_approximation.py. It should support both European option puts and calls approximations. Tests should be in heston_approximation_test.py in the same folder.

Almost all of Kokkos should be noexcept, but let's break it up into pieces that can be tackled in reasonable-sized issues. This one is the most urgent.

Just an FYI whilst I was trawling through the ROCm GitHub page:

https://rocmdocs.amd.com/en/latest/Programming_Guides/Programming-Guides.html#

Please describe the problem you are trying to solve.
The current implementation of the barycentric coordinates are overly complicated and explicitly require the normals of triangles or edges. There are simpler ways to implement this.

Describe the solution you propose.

• Simplify math/barycenter.[ch]pp
• Remove Edge::computeNormal() as it is unnecessary.

**Describe alternativ