This is a project to have a minimum working example of compute workloads across APIs. At present the app loads a big array of ints into the GPU, multiplies them, and reads them back.
This is also used as a testbed for looking at profiling tools and what they can offer to compute only applications
Currently implemented APIs:
- Vulkan
- DX12 Compute
- Cuda
- OpenCL
With helper code to work with the following Profiling tools
- RenderDoc
- PIX
- GPUView (via windows EventWriteString)
Work currently ongoing to add
- Radeon GPU Profiler (RGP)
- Radeon Compute PRofiler (RCP)
- Intel Tools
- Nvidia NSight
- anything else I find
I've had to write this code fast and on a deadline, it's horrible, don't hate me. Also See CONTRIBUTING.md