Tracking: Parallel Sparse Trie Optimisations

[yongkangc]

Motivation

Parallel sparse trie (PST) still underperforms the serial trie path on large blocks. State-root task execution spends a large fraction of time coordinating subtrie shards, sending data back and forth through channels, and processing redundant proof updates produced by the prewarming/multiproof pipeline.

Impact

• Speeding up PST for large blocks
• Reduces redundant proof work feeding PST.
• Lowers peak CPU and memory usage during state-root calculation by removing per-update spawns and channel churn.

[0xKarl98]

Can i try it ?