Intrinsics & Builtins

• Mapping of math and SIMD intrinsics to backend equivalents
• CUDA-specific intrinsics availability
• Fallback behavior on non-supporting platforms
• Guidelines to write portable high-performance code