Skip to content

WIP: plan overlap-aware expert prefetch - #180

Draft
drunkcoding wants to merge 1 commit into
mainfrom
plan/overlap-aware-expert-prefetch
Draft

WIP: plan overlap-aware expert prefetch#180
drunkcoding wants to merge 1 commit into
mainfrom
plan/overlap-aware-expert-prefetch

Conversation

@drunkcoding

Copy link
Copy Markdown
Contributor

Description

Plan-only design for budgeting expert prefetch by measured transfer bandwidth and available compute overlap.

Motivation

Current speculative prefetch is not constrained by overlap capacity, queue debt, or exact expert bytes.

Scope

  • Deterministic overlap-budget controller
  • Native queue accounting, cancellation, and timing
  • Coverage/waste/late-byte telemetry
  • Native router remains authoritative

Dependencies / limitations

Independently implementable on eager routing. First-release co-enablement with GPU-only routing is rejected until route-ready ordering is reconciled.

Type of Change

  • Documentation update
  • Production implementation

Performance / support evidence

  • Not applicable — benchmark methodology only; no speedup claim.

Checklist

  • Source-aligned plan
  • Default-off fallback specified
  • No production files changed

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
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.

1 participant