GPU computing with Vulkan

Goal

Understand Vulkan’s cost and explicit control model when running general-purpose GPU computing, including complete device and compute pipeline initialization.

Experiment

The program generates a 1000 by 1000 pixel fractal through two equivalent implementations. In the run documented in the repository, the CPU took 285 ms and the GPU took 4 ms. This result is not presented as a universal benchmark, but as a practical demonstration of the parallelism available for this workload.

What I learned

The project follows a Vulkan GPU computing course and documents initialization, memory selection, buffer creation, descriptors, and synchronization. It is presented as a learning experiment and kept distinct from the portfolio’s original projects.