This project's visualization is not in the form of writing a "hard to see" .ppm file,
instead, it's based on my previous project in visualime,
which is a simple 2D visualization rendering engine, I borrow its capability of directly popping up an
opengl window given a CUDA&OpenGL Interoperability memory(which is visualime::opengl_wrapper::cuda_opengl_wrapper).
So, that's pretty neat, if you haven't checked out visualime, you should definitely check it out.
You can find my previous project on raytracing in C++ in this repository.
Work progress:
- visualization based on visualime
- basic rendering complete (naive raytracing)
- accumulative ray color
TODOs:
- speed up by Importance Sampling
- ...
Recently I just found out that Nvidia Nsight Compute is a perfect tool for analyzing CUDA code.
So in this project, I just simply use it to analyze the instruction level fp32/fp64 arithmetic call.
Also, I tried out the System Trace functionality, though I currently didn't work it out very nicely.
