OMP + CUDA

• Hybrid deployments mixing CPU (OpenMP) and GPU (CUDA)
• Strategy: fallback paths, workload partitioning, environment detection
• Build configuration for dual backends
• Testing parity across CPU/GPU results