ROCm-Developer-Tools/HIP
HIP
HIP is a C++ GPU kernel language and cross-platform runtime designed for writing portable high-performance compute applications. It provides a programming interface that allows a single source codebase to execute on both AMD and NVIDIA GPU architectures.
The project functions as a compatibility layer that enables the conversion and migration of existing CUDA source code to run on AMD hardware. This is achieved through a syntax mapping that mirrors CUDA and a source-to-source translation process during compilation.
The toolkit covers the broader surface of cross-platform GPGPU development, including heterogeneous compute optimization and the creation of portable kernels. It utilizes a runtime abstraction to map unified API calls to vendor-specific driver libraries for memory and kernel management.
Features
- Portable Kernel Languages - Provides a C++ GPU kernel language for writing parallel compute kernels that target multiple hardware backends.
- CUDA Source Converters - Converts existing NVIDIA CUDA source code into a compatible format to run on AMD hardware.
- Cross-Vendor Compute Engines - Functions as a cross-platform runtime allowing compute applications to run on different GPU manufacturers.
- GPU Kernel Development - Enables the development of single-source C++ kernels that execute on multiple GPU architectures.
- Syntax Mapping Layers - Implements a language grammar that mirrors NVIDIA CUDA to simplify the porting of GPU kernels.
- GPU Programming Runtimes - Supports writing single-source C++ code that runs on both AMD and NVIDIA GPUs.
- Source-to-Source Compilers - Translates a shared C++ dialect into platform-specific GPU code during the compilation phase.
- CUDA and HIP Code Generators - Facilitates the migration of existing CUDA source code to run on AMD hardware via the HIP interface.
- CUDA Compatibility Layers - Provides a programming interface that allows NVIDIA CUDA source code to be converted and run on AMD hardware.
- Runtime Abstraction Layers - Maps unified API calls to vendor-specific driver libraries for memory and kernel management.
- Kernel Dispatchers - Provides a common interface to schedule GPU workloads across different hardware vendors.
- Build-Time Dialect Mappings - Uses preprocessor macros and conditional headers to resolve hardware-specific types and function signatures at build time.
- Heterogeneous Computing Libraries - Creates efficient GPU-accelerated applications that can be deployed across diverse hardware environments.
- Cross-Platform Development - Provides a framework for developing high-performance compute kernels that remain portable across GPU vendors.
- Parallel Programming Frameworks - C++ runtime API and kernel language for GPU acceleration.
