search

morousg / cvGPUSpeedup

A faster implementation of OpenCV-CUDA that uses OpenCV objects, and more!
Apache License 2.0
34 stars 5 forks source link

Organize the code in 3 levels: FusedKernel (Core), FK_Algorithms, cvGPUSpeedup #75

Closed morousg closed 9 months ago

morousg commented 9 months ago

We want to have the code organized in 3 levels:

  1. Fused Kernel or Core: here we will have all the basic components to be able to create an launch Fused CUDA Kernels. That is the GridPatterns, DeviceFunctions, and Operations.
  2. Fused Kernel Algorithms: here we want to create template interfaces to specific functionalities that we want to expose, in a way that is more understandable, that having to compose the kernels by combining DeviceFunctions with Operations.
  3. cvGPUSpeedup: the OpenCV wrapper that allows to use all the rest, from OpenCV, with OpenCV like function names and OpenCV types and enums.