Contents

• 1. cuObjClient API Specification
  • 1.1. Overview
  • 1.2. Architecture
  • 1.3. Core Types and Enumerations
    • 1.3.1. Error Types
    • 1.3.2. Protocol Types
    • 1.3.3. Operation Types
    • 1.3.4. Memory Types
  • 1.4. cuObjClient Class API
    • 1.4.1. Class Declaration
    • 1.4.2. Constructor
  • 1.5. Memory Management APIs
    • 1.5.1. cuMemObjGetDescriptor
    • 1.5.2. cuMemObjPutDescriptor
    • 1.5.3. Registered and Unregistered Memory
    • 1.5.4. cuMemObjGetMaxRequestCallbackSize
    • 1.5.5. cuMemObjGetRDMAToken
    • 1.5.6. cuMemObjPutRDMAToken
  • 1.6. I/O Operations
    • 1.6.1. cuObjGet
    • 1.6.2. cuObjPut
  • 1.7. Connection Management
    • 1.7.1. isConnected
  • 1.8. Static Utility Methods
    • 1.8.1. getCtx
    • 1.8.2. getMemoryType
  • 1.9. Telemetry Management
    • 1.9.1. setupTelemetry
    • 1.9.2. shutdownTelemetry
    • 1.9.3. setTelemFlags
  • 1.10. Callback Interface
    • 1.10.1. CUObjIOOps Structure
    • 1.10.2. GET Callback
    • 1.10.3. PUT Callback
  • 1.11. Error Handling
    • 1.11.1. Error Codes
    • 1.11.2. Return Value Conventions
    • 1.11.3. Best Practices
  • 1.12. Client IPv6 Endpoint Support
  • 1.13. Usage Patterns
    • 1.13.1. Basic Client Usage With Explicit Descriptor Acquisition
    • 1.13.2. Callback Implementation Example
    • 1.13.3. CUDA Device Memory Usage
    • 1.13.4. Manual RDMA Token Management
    • 1.13.5. Telemetry Configuration
  • 1.14. Logging Configuration
  • 1.15. RDMA Peer Memory Configuration
  • 1.16. Constants and Limits
    • 1.16.1. Memory Limits
    • 1.16.2. Protocol Constants
    • 1.16.3. Operation Limits
  • 1.17. RDMA Multipathing
    • 1.17.1. How RDMA Multipathing Is Used
    • 1.17.2. Callback Return Values That Can Trigger Retry
  • 1.18. Thread Safety
    • 1.18.1. Client Thread Safety
  • 1.19. API Usage Notes