AI Architecture & Systems

Distributed Compilation

3.12.1SPMD, Sharding Annotation, and Device Mesh#

3.12.2Automatic Parallelization Strategies and Graph Partitioning#

3.12.3Collective Insertion and Resharding#

3.12.4Compile-Time Scheduling of Communication–Computation Overlap#

3.12.5Pipeline Partitioning, Cross-Device Layout, and Memory Constraints#

3.12.6OpenXLA, GSPMD, and Shardy#