feat(fixtures): survey FlashInfer as an authored-HIR corpus - #125
Draft
zhen8838 wants to merge 1 commit into
Draft
feat(fixtures): survey FlashInfer as an authored-HIR corpus#125zhen8838 wants to merge 1 commit into
zhen8838 wants to merge 1 commit into
Conversation
WIP. Collects 89 FlashInfer kernels rewritten as authored HIR, plus the two scripts that produced and triaged them. 35 parse and analyze; 54 carry a `.blocked.py` suffix and record in their docstring the refusal they hit, so the corpus states what the surface cannot express yet rather than hiding it. The suffix is what keeps a blocked fixture out of packaging, and the source loader now resolves a sibling by its base stem so the rest still import. A fixture docstring is a parsed key/value record rather than prose, so the corpus joins the directories the documentation lint leaves alone.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
WIP. Opened to park the corpus and its tooling; not ready to merge.
Why
express was known only from the tests written alongside each feature, which
is the sample most likely to agree with it.
What
tests/fixtures/flashinfer/.35 parse and analyze; 54 carry a
.blocked.pysuffix and record the refusalthey hit in their docstring, so a gap is a file rather than a memory.
scripts/survey_flashinfer.pyproduces them from an upstream revision;scripts/summarize_blocked.pygroups the blocked ones by root cause.tests/fixtures/test_corpus.pyruns the unblocked ones through the authoredcommand boundary and reads each blocked file's recorded reason.
cli/source.pyresolves a sibling module by its base stem, so.blocked.pynames load and packaging leaves them out.
Contract
tests/fixtures/flashinfer/joins thedirectories the documentation lint leaves alone, because a fixture docstring
is a parsed key/value record rather than prose.
Risk
probe, 11 need an
nvfp4dtype, 10 name ops the surface does not have, 3declare no Module. None of them is triaged into a decision yet.
from the revision its docstring names.
instead, which needs the record format to survive wrapping.