HPC-X current version provides the following changes and new features:
|
Category |
Change Description |
|---|---|
|
HPC-X Content |
Updated HPC-X Content section to reflect the communication libraries versions embedded in this HPC-X release:
|
|
RMA & Transport Enhancements |
Added large-message Remote Direct Memory Access (RMA) rendezvous protocols for GET/PUT operations, alongside improved multi-lane rendezvous selection logic. |
|
[Partial support] Fault Tolerance |
Introduced infrastructure for endpoint fault tolerance and recovery. |
|
CUDA & GPU Memory Improvements |
Added same-node CUDA Virtual Memory Management (VMM) Inter-Process Communication (IPC) using POSIX file-descriptor handles, CUDA 13.1 Multi-Process Service (MPS) MLOPart topology support, and enhanced CUDA IPC and memory-type handling. |
|
Plugin & Transport Support |
Added support for the GDP plugin, including the D2P transport layer. |
|
Platform Defaults |
Enabled four rails by default and relaxed memory ordering on Vera platforms. |
|
Bug Fixes |
See Bug Fixes. |
Last updated: