https://github.com/pkestene/cuda-proj-tmpl

A minimal cmake based project skeleton for developping a CUDA application
https://github.com/pkestene/cuda-proj-tmpl

cea cmake cuda gpu gpu-computing parallel-computing parallel-programming template

Last synced: 4 days ago
JSON representation

A minimal cmake based project skeleton for developping a CUDA application

• Host: GitHub
• URL: https://github.com/pkestene/cuda-proj-tmpl
• Owner: pkestene
• License: gpl-3.0
• Created: 2019-10-29T13:00:29.000Z (about 5 years ago)
• Default Branch: master
• Last Pushed: 2024-01-20T23:22:21.000Z (11 months ago)
• Last Synced: 2024-01-21T00:28:09.860Z (11 months ago)
• Topics: cea, cmake, cuda, gpu, gpu-computing, parallel-computing, parallel-programming, template
• Language: CMake
• Homepage:
• Size: 41 KB
• Stars: 7
• Watchers: 2
• Forks: 3
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# cuda-proj-tmpl

A minimal cmake based project skeleton for developping a CUDA application

## Download this skeleton

```bash

git clone --recursive [email protected]:pkestene/cuda-proj-tmpl.git

```

## Modern CMake and CUDA

See https://github.com/CLIUtils/modern_cmake

## Requirements

- cmake version >= 3.18 (when using nvcc)

- cmake version >= 3.22 (when using nvc++/nvhpc compiler)

- cuda toolkit

## How to build with nvcc ?

```bash

# set default CUDA flags passed to nvcc (Nvidia compiler wrapper)

# at least one hardware architecture flags

export CUDAFLAGS="-arch=sm_75 --expt-extended-lambda"

mkdir build

cd build

cmake ..

make

# then you can run the application

./src/saxpy_cuda

```

If you use cmake version >= 3.18, here is a slightly updated version using `CMAKE_CUDA_ARCHITECTURE`

to pass information about actual GPU hardware target architecture:

```bash

# set default CUDA flags passed to nvcc (Nvidia compiler wrapper)

# at least one hardware architecture flags

export CUDAFLAGS="--expt-extended-lambda"

mkdir build

cd build

cmake -DCMAKE_CUDA_ARCHITECTURES="75" ..

make

# then you can run the application

./src/saxpy_cuda

```

# How to build with nvc++ (from nvhpc) ?

```bash

# Warning, cmake 3.22.0 required

export CUDAFLAGS="--expt-extended-lambda"

export CXX=nvc++

mkdir build

cd build

cmake -DCMAKE_CUDA_HOST_COMPILER=nvc++ -DCMAKE_CUDA_ARCHITECTURES="75" ..

make

# then you can run the application

./src/saxpy_cuda

```