aiur: retire the virtual-gas meter's API surface - #611
Merged
Conversation
samuelburnham
force-pushed
the
sb/aggregate-first-drop-meter
branch
2 times, most recently
from
September 3, 2026 14:10
58c93f0 to
10a82bb
Compare
The recalibration loop never reads the meter: splitting, the wave audit, refinement, and the budget-gated proves all run on `peak_prove_bytes` and the suggested part count, measured per record — and cost-weighted splitting was benchmarked against block-count cuts on a synthetic fixture, init, and lean, and lost on every axis (vspan tracks time where the RAM model tracks padded rows). Its consumers were `ix check --profile` and the batch weight rows, which nothing downstream reads. This removes that surface: the `--profile` flag and `dumpProfile`, the `QueryRecordHandle` accessors and `ExecuteResult`'s record slot, the `Full` execution wrappers, `const_weights` and `ShardResult.weights`, and the unused merge estimator (`raw_heights`/`peak_of_raws` — the refinement path reports a sum of measured peaks instead). The meter RUNTIME stays, dormant: querymap, execute, the runners, `Codegen.lean`, and the three generated kernels are untouched so the generated code keeps compiling and `ix codegen check` keeps passing, with every FFI entry hardwiring `profile = false`. Excising the runtime means regenerating ~20k lines of kernels, so it rides with the regeneration this branch already owes at merge time, as one commit. Verified: the wave audit, `ix shard refine`, and the budget-gated prove reproduce their pre-strip behavior (4 leaves from 1, refined manifests emitted, all proofs bind-verifying against them, reseeded runs splitting nothing); aiur's 9 and kernel's 693 tests pass; workspace clippy at CI flags and rustfmt are clean.
johnchandlerburnham
previously approved these changes
Sep 3, 2026
samuelburnham
dismissed
johnchandlerburnham’s stale review
September 3, 2026 19:59
The merge-base changed after approval.
samuelburnham
force-pushed
the
sb/aggregate-first-drop-meter
branch
from
September 3, 2026 19:59
10a82bb to
f714d98
Compare
samuelburnham
enabled auto-merge
September 3, 2026 20:57
Remove the dormant profile, weight, and virtual-span state from query maps and execution entrypoints while retaining witness multiplicity accounting. Update the Lean emitter and regenerate all three production kernels.
arthurpaulino
approved these changes
Sep 3, 2026
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
3 participants
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.
The recalibration loop never reads the meter: splitting, the wave audit, refinement, and the budget-gated proves all run on
peak_prove_bytesand the suggested part count, measured per record — and cost-weighted splitting was benchmarked against block-count cuts on a synthetic fixture, init, and lean, and lost on every axis (vspan tracks time where the RAM model tracks padded rows). Its consumers wereix check --profileand the batch weight rows, which nothing downstream reads.This removes that surface: the
--profileflag anddumpProfile, theQueryRecordHandleaccessors andExecuteResult's record slot, theFullexecution wrappers,const_weightsandShardResult.weights, and the unused merge estimator (raw_heights/peak_of_raws— the refinement path reports a sum of measured peaks instead).Verified: the wave audit,
ix shard refine, and the budget-gated prove reproduce their pre-strip behavior (4 leaves from 1, refined manifests emitted, all proofs bind-verifying against them, reseeded runs splitting nothing); aiur's 9 and kernel's 693 tests pass; workspace clippy at CI flags and rustfmt are clean.