Interoperation between CUDA and Vulkan
Published:
Blog posts
2024
Vulkan Memory Transfer
Published:
Vulkan Memory Transfer
In an classic graphics pipeline, the first step is always turning a bunch of points into triangles or other primitives. The input data usually consist of a array of vertice and corresponding indices. Specifically, in the Vulkan, the actual first step is passing the vertices and indices into the gpu memory. CUDA, as general purpose compute API, offers a higher-level API for memory management. Vulkan, on the contrast, providesa fine-grained control over memory allocation and usage, which adds complexity but allows for highly optimized GPU memory management.