FLT

Flt is a small header-only library designed to abstract vectors of floats, doubles, complex floats, and complex doubles so they can be used with a common type-erased interface.

Useful typedefs:

• flt::cfloat - Maps to std::complex<float>
• flt::cdouble - Maps to std::complex<double>

Important types:

• flt::vector - Similar to std::vector<float>, std::vector<double>, std::vector<flt::cfloat>, and std::vector<flt::cdouble>. This class manages its own memory, but doesn't provide most of the std::vector interface.
• flt::vector_ref - Type erased wrapper for a floating point vector. Changes made to the wrapper affect the underlying std::vector, as they share the same data in memory. Think of this as a std::vector<T>&.

Example Usage

#include<vector>
#include"flt/flt.h"
std::vector<double> v1(16, 3.0);
std::vector<float> v2(16, 2.0f);
flt::vector_ref v1Ref(v1);
flt::vector_ref v2Ref(v2);
v1Ref[0] = v1Ref[0] * 5.0f + v2Ref[0];
std::cout << v1Ref[0].as<double>() << std::endl;

Building

The library is header only. Simply include the entire /include directory to get started. It can also be 'built' as a CMake interface library for inclusion with other projects.

The /tests directory contains several function and performance tests that ensure correctness and speed.

Why?

When used correctly, flt's primitives add no additional overhead to std::vectors, but they can be used without templates, which is desirable in certain situations. That being said, when the underlying type of the container is difficult or impossible to determine at compile time, the generated assembly can be very suboptimal, leading to poor performance. In particular, writing functions that take flt::vector or flt::vector_ref objects (either by value or reference) will likely produce poor results, though the actual assembly should be examined if possible.

Ultimately, this project was a fun thought experiment and may not provide much real world value, but the techniques used for implementation may be helpful for some other use case. :-)