Sync: corpus/clubb_solve switches its scalar loop the way CLUBB does - #17
Merged
Merged
Conversation
…b_solve switches its scalar loop the way CLUBB does The anchor's held loop bounds (_dolo_sclr, _dohi_sclr) start as jnp.int32(0) at the top of the kernel and are carried by every cond around their loop, but a store of one in a branch the trace resolves -- CLUBB's variance solve sets `_dohi_sclr = sclr_dim` under the PDF switch's else, after an error return -- went through no cast: the literal-strengthening rules exempt a name that bounds a range, and the tree marks a local set from statics a trace-time value. Under the return flag's cond the arms disagreed, int64 against int32, and the whole step would not trace on gabls2, the one case with passive scalars, while every case without them compiled and passed the gate; the 103-job re-gate did not include the scalar recordings, and the closed-loop run found it. lower_assign now wraps every store to a `_do(lo|hi|st)_*` name in jnp.int32. tests/test_jax_transform.py holds the rule with the extents and statics set the way CLUBB's kernel had them. corpus/clubb_solve's xp2_solve takes CLUBB's shape: an error return after the first solve, a static PDF selector (ipdf_type, a module constant to compare with), the per-scalar assembly with its own returns under any other PDF and the held-bound loop under ADG1 -- on the engine before this change its port fails with the same TypeError; both summaries re-recorded from the isolated environment CI uses. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Signed-off-by: lewisychen <yueqichen.0x0@gmail.com> Carried-From: 19607c9e3cee34b662ad1b6010b7d8d7ce5f1779
chenyueqi
added
corpus
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.
Carried by tools/export sync; the commit names its origin in a Carried-From trailer.
corpus/clubb_solve's xp2_solve now takes CLUBB's shape around its scalar loop: an error return after the first solve, a static PDF selector (ipdf_type, compared with a module constant), the per-scalar assembly with its own returns under any other PDF and the held-bound loop under ADG1. On the engine before the accompanying fix, the port of this tree failed to trace (a lax.cond whose arms disagreed, int64 against int32) -- the same failure the whole CLUBB step showed on the one recorded case with passive scalars while every case without them passed. Both summaries re-recorded from the isolated environment CI uses.
🤖 Generated with Claude Code