Provide a high-level and cross-platform way to use GPU resources for computing and rendering. Minimize the amount of low-level and platform-specific coding required to implement useful features.