News
- 2009-09-26 - This page created.
Examples of what you might use it for
- Linear algebra on arbitrary fields
- Plane geometry with fixed point arithmetic
- Automatically differentiated FFT output
By templating numerical algorithms, many important auxilliary types of computations become possible.
What this is not
This code is not meant to be the fastest. The use of templates necessitates certain tradeoffs in how numerical algorithms are coded in order to accomodate the most types.
Practical performance and convergence considerations also necessitate algorithmic modifications that prevent a fully generic type specialization. It is hoped that the balance achieved is simply generally useful.
Unique features
- Single file, header-only FFT implementation
- Toeplitz matrix inversion in O(n^2) time
- Templated BLAS
- Automatically differentiated matrix/vector classes