6. How Things Work

This section explains how some of the functions inside PMIx perform their job. If you have questions about, for example, how device distances are computed, then you will find information on that subject here.

• 6.1. Session Directories
  • 6.1.1. Client Session Directories
  • 6.1.2. Tool and Server Session Directories
• 6.2. Distance Computation
• 6.3. Scheduler Integration
  • 6.3.1. Overview
• 6.4. Process Sets and Groups
  • 6.4.1. Overview
  • 6.4.2. Process Sets
  • 6.4.3. Group Construction
• 6.5. Allocation Inheritance
  • 6.5.1. Overview
• 6.6. Resolving Peers and Nodes
• 6.7. How to Add a PMIx Datatype
  • 6.7.1. Step 1: Definitions
  • 6.7.2. Step 2: Buffer Operations
  • 6.7.3. Step 3: Random Things
• 6.8. PMIx_Log Support
  • 6.8.1. Public API
  • 6.8.2. Channel Constants
  • 6.8.3. Call Flow
  • 6.8.4. The MCA plog Framework
  • 6.8.5. plog Components
  • 6.8.6. pmix_show_help Aggregation
  • 6.8.7. Gateway Servers vs. Non-Gateway Servers
  • 6.8.8. The pmix_log_host_only MCA Parameter
  • 6.8.9. The PMIX_OPERATION_SUCCEEDED convention
  • 6.8.10. Thread Safety
  • 6.8.11. Key Source Files
• 6.9. Regular Expression (Node/Proc Map) Support
  • 6.9.1. The Problem
  • 6.9.2. Public APIs
  • 6.9.3. Producer / Consumer Flow
  • 6.9.4. The MCA preg Framework
  • 6.9.5. preg Components
  • 6.9.6. Wire Format
  • 6.9.7. Thread Safety
  • 6.9.8. Key Source Files
• 6.10. Transport Layer (Client/Server/Tool Connections)
  • 6.10.1. The Problem
  • 6.10.2. An Unusual Framework Shape
  • 6.10.3. Publishing a Server: the Listener
  • 6.10.4. Finding and Connecting to a Server
  • 6.10.5. The Handshake
  • 6.10.6. Messages in Steady State
  • 6.10.7. Threading Model
  • 6.10.8. Configuration
• 6.11. Process & Resource Statistics (Monitoring)
  • 6.11.1. The Problem
  • 6.11.2. Public API
  • 6.11.3. Request Attributes
  • 6.11.4. From API to Framework
  • 6.11.5. Inside the Framework
  • 6.11.6. Threading
  • 6.11.7. Summary