I made my own GPU graphics API
Hello everyone, in the last few weeks i decided to start a new project to learn and experiment with CUDA. So I decided to create my own graphics api for fun since i have not really seen a project like this before.
It's nothing particulary impressive but it could be interesting to some.
The performance is very bad but hey at least it works!
API Interface
The API interface is exposed with a normal C header but the implementation is in CUDA.
The interface exposes:
- stage buffers
- textures
- vertex and index buffers
- command buffers
- render passes
- line, quad, and triangle draws
- swapchain
Of course I didnt implement shaders or custom vertex attributes since I wanted to keep this simple and that would have been way too much work to do.
Current Performance Bottleneck
Right now the main bottleneck is Windows presentation through GDI (SetDIBitsToDevice).
C3D renders into CUDA-managed GPU memory, but CUDA cannot present that memory directly to a window surface. Instead, it has to copy the rendered image back to CPU-visible memory and hand it off to GDI for presentation. That extra GPU-to-CPU transfer plus the GDI blit is currently the slowest part of the frame path.
Here is a Tracy capture from the demo showing the presentation path under inspection:
Github
Heres the github repo if you want to check the code for yourself: https://github.com/luppichristian/C3D
Note: I used AI to write some of the code, my objective was to learn.