Skip to content

CORRECTNESS_ONLY_STAGES is empty — skip-speedup behavior not exposed #10

@sandlbn

Description

@sandlbn

The CORRECTNESS_ONLY_STAGES set in optimize_stage is currently hardcoded to {}, disabling the skip-speedup-check behavior for all stages.
Originally intended to allow stages like DTYPE_FIX and ALGORITHMIC to produce kernels that are temporarily slower (correctness fix first, performance stages recover later). This makes sense when casting from fp16→fp32 or fp16→bf16 — the cast itself may be slightly slower but is required for correctness.
Proposal: Either remove the dead code entirely, or expose it as CORRECTNESS_ONLY_STAGES = {OptimizationStage.DTYPE_FIX} with a comment explaining why, and add a --skip-speedup-stages CLI option.

Metadata

Metadata

Assignees

No one assigned

    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