AMD Releases Orochi 1.0 For Dynamic Runtime Switching Between Radeon HIP & NVIDIA CUDA
Earlier this year AMD's GPUOpen team announced the Orochi project for dynamic HIP/CUDA run-time handling. Orochi makes it easier for application developers to ship AMD HIP and NVIDIA CUDA support within a single code-base / binary that is then selected at run-time based on the GPU in use.
Today marks the Orochi 1.0 stable release for this helper to allow applications to run gracefully on either AMD or NVIDIA GPUs where as prior HIP/CUDA efforts have been limited to making support decisions at compile-time. Orochi in turn dynamically loads the necessary HIP/CUDA shared libraries. Orochi though does require adapting the code to make use of the Orochi API, even if it's currently written for HIP there are basic changes to the API.
Since being announced and going public earlier this year, Orochi has seen a number of fixes, error reporting improvements, API additions, and a variety of other changes made to this library.
Orochi 1.0 can be downloaded from GitHub.
Today marks the Orochi 1.0 stable release for this helper to allow applications to run gracefully on either AMD or NVIDIA GPUs where as prior HIP/CUDA efforts have been limited to making support decisions at compile-time. Orochi in turn dynamically loads the necessary HIP/CUDA shared libraries. Orochi though does require adapting the code to make use of the Orochi API, even if it's currently written for HIP there are basic changes to the API.
AMD GPUOpen Orochi
Since being announced and going public earlier this year, Orochi has seen a number of fixes, error reporting improvements, API additions, and a variety of other changes made to this library.
Orochi 1.0 can be downloaded from GitHub.
12 Comments