https://github.com/kis-balazs/cuda-research

CUDA Research & Code. Course-style structured. Inspiration from @Infatoshi.
https://github.com/kis-balazs/cuda-research

cuda

Last synced: 16 days ago
JSON representation

CUDA Research & Code. Course-style structured. Inspiration from @Infatoshi.

• Host: GitHub
• URL: https://github.com/kis-balazs/cuda-research
• Owner: kis-balazs
• License: bsd-3-clause
• Created: 2024-11-16T14:40:33.000Z (about 2 months ago)
• Default Branch: main
• Last Pushed: 2024-12-22T11:33:44.000Z (19 days ago)
• Last Synced: 2024-12-22T12:26:25.312Z (19 days ago)
• Topics: cuda
• Language: Cuda
• Homepage:
• Size: 4.08 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# CUDA-Research

Research & Code for CUDA components developed/experimented by me.

## Resources Outline

- [cuda-course](https://github.com/Infatoshi/cuda-course)

- [cuda-mnist](https://github.com/Infatoshi/cuda-course)

- [cuda-opencv-examples](https://github.com/evlasblom/cuda-opencv-examples/tree/master)

## Related

- [kis-balazs/CUDA-Containers-Infrastructure-Repository](https://github.com/kis-balazs/cuda-containers-infra)

## Important Notice - GPU-specific nvcc params

When running `nvcc` commands, the `-arch` command is not always synced to the correct physical GPU. This can be fixed by specifying this.

Steps:

- find compute version: find the GPU compute version [here](https://developer.nvidia.com/cuda-gpus)

- specify when compiling: `nvcc -o exec code.cu -arch=compute_XX`

- optionally, `code` can be specified as well: `-code=sm_XX,compute_XX`