Skip to content

Conversation

@jvarela-jump
Copy link
Contributor

Two architectural optimizations to the snapshot load pipeline under vinyl:

  1. snapin tile is split into two:
  • snapin handles parsing and bstream pair generation.

  • snapwm handles vinyl meta_map and bstream allocation.

    With this change: snapin->snapwm->snapwh->snapwr.

  1. The topology now supports multiple snapwr tiles, and snapwh handles multiple fseq accordingly.
  • the number of snapwr tiles is configurable from the toml.

The measured speedup is around ~1.6x.

@github-actions
Copy link

Performance Measurements ⏳

Suite Baseline New Change
backtest mainnet-368528500-perf per slot 0.050439 s 0.050729 s 0.575%
backtest mainnet-368528500-perf snapshot load 1.627 s 1.649 s 1.352%
backtest mainnet-368528500-perf total elapsed 50.438584 s 50.728887 s 0.576%
firedancer mem usage with mainnet.toml 1005.23 GiB 1005.23 GiB 0.000%

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants