Skip to content

CountMinSketchWithHeap still costs at the flat stub — cms-heap wrapper exists but isn't wired up #651

Description

@milindsrivastava1997

sketch-bench's `cms-heap` wrapper landed (sketch-bench PR #104), but ASAPQuery's `sketch_bench_key()` (atomic_costs.rs) has no case for `CountMinSketchWithHeap`, so it still always resolves to `AtomicCosts::default()` regardless of params.

Two things needed:

  1. `export_atomic_costs.sh` doesn't sweep `cms-heap` yet.
  2. Param mismatch: sketch-bench's `CmsHeapParams` is `{rows, cols}` only (`top_k` is compile-time, Frequency/TopK is two registry variants not a field), while ASAPQuery's grid sweeps `{depth, width, heapsize, count_events}`. `heapsize` has no home on the sketch-bench side — stays analytic (`heap_size · avg_key_size`, per Optimizer: param-aware atomic costs from sketch-bench #524) even once CPU costs are wired up. Also unresolved which of the 4 sketch-bench variants (FastPath/RegularPath × Frequency/TopK) matches what ASAPQuery's accumulator actually deploys.

Example: `candidate-gen-dump --atomic-costs ...` shows every `CountMinSketchWithHeap` params row with identical `[stub] mem=1024B insert=1.000e-7s ...`, regardless of depth/width/heapsize.

Part of #524.

Activity

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

Metadata

Metadata

Labels

No labels
No labels

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions