rocm-examples/HIP-Basic/hello_world_cuda

HIP-Basic Hello World on the CUDA platform Example

Description

This example showcases a simple HIP program that is compiled on the CUDA platform using CMake.

Application flow

1. A kernel is launched: the function hello_world_kernel is executed on the device. The kernel is executed on a single thread, and prints "Hello, World!" to the console.
2. A launch error check is performed using hipGetLastError.
3. Synchronization is performed: the host program execution halts until the kernel on the device has finished executing.

Key APIs and Concepts

• For introduction to the programming concepts in this example, refer to the general hello world example.
• This example showcases setting up a HIP program to be compiled to the CUDA platform using CMake.
  • Since CMake (as of version 3.21) does not support compiling to CUDA in HIP language mode, CUDA language mode has to be used. Thereby the project language is specified as CUDA.
  • Additionally, we must "teach" CMake to compile the source file main.hip in CUDA language mode, because it cannot guess that from the file extension. This is done by set_source_files_properties(main.hip PROPERTIES LANGUAGE CUDA).
  • The HIP "runtime" on the CUDA platform is header only. Thereby there is no need to link to a library, but the HIP include directory have to be added to the search paths. This is performed by target_include_directories(${example_name} PRIVATE "${ROCM_ROOT}/include".

Demonstrated API Calls

HIP Runtime

• hipGetLastError
• hipDeviceSynchronize
• __global__

Supported Platforms

This example is only supported on the CUDA platform.