4. CUDA Features

• 4.1. Unified Memory
• 4.2. CUDA Graphs
• 4.3. Stream-Ordered Memory Allocator
• 4.4. Cooperative Groups
• 4.5. Programmatic Dependent Launch and Synchronization
• 4.6. Green Contexts
• 4.7. Lazy Loading
• 4.8. Error Log Management
• 4.9. Asynchronous Barriers
• 4.10. Pipelines
• 4.11. Asynchronous Data Copies
• 4.12. Work Stealing with Cluster Launch Control
• 4.13. L2 Cache Control
• 4.14. Memory Synchronization Domains
• 4.15. Interprocess Communication
• 4.16. Virtual Memory Management
• 4.17. Extended GPU Memory
• 4.18. CUDA Dynamic Parallelism
• 4.19. CUDA Interoperability with APIs
• 4.20. Driver Entry Point Access