refactor(breakdown): delete the dead and repeated code in two over-gate kernel-event and attribution-renderer functions by fengshaoyi-amd · Pull Request #2074 · AMD-AGI/Hyperloom

fengshaoyi-amd · 2026-10-21T03:11:31Z

What this does

Two session-breakdown functions were over the code-metrics gate: assemble_kernel_ext, which builds a kernel event's ext from its recorded rows, and the attribution renderer's render. This PR brings both under every limit they broke, mostly by deleting: eight hand-written read-and-sort blocks (now one table), a second pass that patched e2e results onto attempts after they were built, a helper whose tuple result was taken apart and rebuilt into the rebench block, nine guards on input every caller already fills, a fallback for a field its only writer always sets, and, in the renderer, a second read of the ledger state and three one-use accumulators.

Production code: +152 / −177 (net −25) across 2 files. No test is added or changed. The code-metrics baseline loses 3 entries, lowers 1 module-length entry and gains none.

This continues the complexity cleanup series that started with #2024; #2052, an earlier breakdown PR in it, shares no source file with this one. The 3 commits read one at a time; each commit message lists what it deleted and why that is safe.

Commit	Module	Production lines	Baseline file
`1fed096` read the attribution section's ledger state once	`reporters/_renderers/attribution.py`	+16 / −22
`756dade` assemble the kernel event from one ordered read of its sections	`recorder/kernel_event.py`	+136 / −155
`a642b0f` retire the code-metrics baseline entries the rework cleared	baseline only		3 removed, 1 lowered

Both production files are under src/hyperloom/inference_optimizer/breakdown/.

Gate numbers per function

Measured with the gate's own collectors (ruff C901, complexipy, function lines), origin/main → this branch. Limits are 20, 30 and 80; bold values were over the limit at base.

Module	Function	Cyclomatic	Cognitive	Lines
`recorder/kernel_event.py`	`assemble_kernel_ext`	11 → 3	50 → 11	206 → 77
`reporters/_renderers/attribution.py`	`render`	11 → 8	28 → 24	85 → 79

Also changed, never over at base: _lane_attempt 1 → 1 / 0 → 1 / 35 → 39.

New functions in kernel_event.py, cyclomatic / cognitive / lines: _forge 3 / 6 / 29, _geak 3 / 12 / 35, _geak_rebench 2 / 8 / 25, _outcome 1 / 2 / 18.

Deleted function: _geak_settlement (3 / 7 / 20), replaced by _geak_rebench.

Across the 2 touched production files: C901 total 141 → 136, cognitive total 287 → 266, functions 51 → 54. Functions over a gate limit: cognitive 3 → 2, length 4 → 2. What is left is outside this work, all in kernel_event.py: _merge_acceptances (cognitive 31), record_geak_attempts (cognitive 31, 102 lines; #2048 changes it) and record_integrate_verdict (89 lines).

Module lengths: kernel_event.py 2,168 → 2,149, attribution.py 102 → 96.

What changes in each module

Kernel event assembly (`recorder/kernel_event.py`)

New names:

_ROW_ORDER: the fields each of the eight row sections is ordered by. One comprehension reads and orders all of them.
_geak_rebench: builds GEAK's wire rebench block (V6KernelGeakRebench), including the attempt GEAK's acceptances settle against and any conflicting decisions.
_forge: builds the forge block (V6KernelForge), or None when forge did not run.
_geak: builds the geak block (V6KernelGeak), or None when GEAK did not run.
_outcome: builds the outcome block (V6KernelOutcome).
_lane_attempt now takes the integrate rows by id. A lane that states no e2e result takes its gate's result while the lane is built.

The ext literal and its field comments stay in assemble_kernel_ext, because test_sbd_v6_ext_schema_contract.py reads the wire keys from that function's body (:31, :40-62).

Removed:

Eight read-and-sort blocks, one per row section.
The second pass over the attempts that patched an e2e block onto lanes after they were built.
_geak_settlement and its (ref, conflicts) tuple, and the second derivation of "a GEAK ledger exists or a rebench was stated" in the GEAK engagement test, which now reads rebench["required"].
Nine parts.get(section) or [] guards.
The entry.route fallback for outcome.route.

Attribution renderer (`reporters/_renderers/attribution.py`)

Removed from render: the second read of attribution.available (the "no per-source split" fact is now the elif of the "no ledger" fact), the notes list built up front, the loop that accumulated decisions (now a comprehension over the same rows and filter), and the one-use md local.

Why the deletions are safe

Line numbers are on this branch unless marked "on main".

Row order table. Each row of _ROW_ORDER (kernel_event.py:2037-2046) holds the sort fields of the block it replaces (kernel_event.py:1948-1989 on main). The sections main passed through wire_rows still go through it where they are read: trace runs and discovered kernels in _forge (:1966-1967), discovery runs in _geak (:2009), GEAK kernel rows before _geak_attempt (:2065). wire_rows drops only rows that are not mappings (event_rows.py:88), and rows_for_event already kept only mappings (event_rows.py:35), so the GEAK engagement test, which now reads the unwired kernel rows (:1995), sees the same rows.
E2E join in _lane_attempt. On main, the second pass (kernel_event.py:1999-2003 on main) filled e2e only for an attempt whose e2e was empty and whose integrate_ref was set. Only a lane sets integrate_ref: _geak_attempt (:833-872) and _acceptance_attempt (:875-895) keep the template's "" (:705), and acceptance attempts were appended after the pass anyway (:2004 on main). _lane_attempt now does the same join while it builds the lane (:802-803, :828). A lane without a reference calls _integrate_e2e([]), which returns None (:659-660), as before.
GEAK rebench block. _geak_rebench (:1921-1945) derives the settled attempt and the conflicting decisions from the same settled rows _geak_settlement read, and keeps the block's key order. assemble_kernel_ext reads rebench["settled_against"] for _merge_acceptances (:2066) and "conflicting_decisions" in rebench for _settle (:2073); that key is present exactly when the old tuple's conflict list was non-empty (:1943-1944). rebench["required"] is bool(ledger) or bool(stated) (:1935), the two facts the old engagement test read separately (kernel_event.py:2044-2046 on main).
Sections read without a guard. Every production caller builds parts with event_parts, which puts a list under every section it is asked for (assembler.py:297-300), and all three ask for EVENT_SECTIONS (sections.py:141-152). That includes KERNEL_EVENT_SECTIONS (sections.py:58-68), which is exactly the event section read at kernel_event.py:2055 plus the eight _ROW_ORDER keys. The callers:
- the recorder's close (kernel_event.py:1601-1604);
- _republish_closed_event (kernel_event.py:641-642);
- the finalize spec (event_finalize.py:62-63, with parts from event_parts(spec.sections) at :147, called at :162).
Tests call it with kernel_event_parts() (assembler.py:333-335), which is event_parts over KERNEL_EVENT_SECTIONS: test_sbd_v6_kernel_timeline.py:112, :984, :1076 and test_coordinator_gemm_promote_units.py:1578. Nothing in scripts/ calls it.
outcome.route. The only code that writes route or entry.route into the kernel event section is KernelEventRecorder.__init__ (kernel_event.py:1098-1110), which writes both from self._route in the same record; every version of the file since it was created (5085b93) does the same. The entry fragment begin writes later (:1131-1149) carries no route, and the route that begin hands to open_event (:1154) goes into the timeline envelope, not into the section's rows (event_timeline.py:122-135). The recorder is built only by make_kernel_recorder (kernel_event.py:2128, constructor call at :2144), called from orchestrator/phases/kernel.py:480.
Attribution facts. attribution is a plain dict (attribution.py:35), so reading available once and branching with elif (:62-64) adds the same facts as main's two separate tests. The breakdown's validation.notes is written by _validation_notes (collectors/v6.py:560, defined at :588), which returns strings; for a string, and for any JSON value read back from disk, f"{note}" and str(note) give the same text.

Observable effect

Nothing an operator sees changes. For every input the repo's own callers produce, kernel event ext blocks (contents and key order), the breakdown JSON and the attribution section stay the same. A seeded differential of assemble_kernel_ext before and after this change, over 6,000 generated events, matched byte for byte apart from the 292 events that hit the first case below.

The differences below are reachable only by input that no caller in the repo produces:

A kernel event header with a blank route but a non-blank entry.route. outcome.route used to fall back to entry.route; it is now "". Only KernelEventRecorder.__init__ writes either field, both from the same value in one record (see Why the deletions are safe).
assemble_kernel_ext given a parts mapping that lacks one of the nine kernel sections, or holds None for one. Such a section used to read as empty; now a missing section raises KeyError and None raises TypeError. Every caller passes event_parts output, which holds a list for each of them.
attribution.render: no difference. The facts, their order, the decisions and the table are unchanged. Notes are formatted by an f-string instead of str(), which gives the same text for every JSON value.

Checklist

Description: above.
Linked issue(s): none.
Tests:
- No test was added or changed, so no contract or failure-mode assertion moved. The existing tests keep pinning what these functions export: test_sbd_v6_kernel_timeline.py and test_coordinator_gemm_promote_units.py call assemble_kernel_ext directly, test_sbd_v6_ext_schema_contract.py pins its top-level wire keys, and test_reporters_smoke.py covers the attribution section's table, shares, notes and decisions.
- Locally on Windows, before the rebase onto main, the affected tests gave the same result on the base (b4e296a) and on these commits: 715 passed on both sides, with the same two pre-existing problems. One failure is in the framework recorder, which this change does not touch (test_sbd_v6_framework_wiring.py::test_source_attempt_records_its_pair_gate_and_lifecycle_step, a CRLF mismatch in the patch content on Windows); one collection error is in test_tracelens_csv.py, because TraceLens is not installed. None of the test files that reach these two modules changed on main since. Linux CI on this PR is the authority.
Size/complexity triggers crossed:
- 329 changed production lines, under the 400-line review trigger.
- No added or changed function crosses a gate limit, and none is over the cyclomatic-10 trigger.
- Two functions sit over the ~60-line trigger and under the 80-line gate. render (79), 15 lines of which are its docstring, is one pass that builds the section's facts, decisions and table. assemble_kernel_ext (77) has 28 comment lines, 15 of them on the ext literal, which stays in this function because the schema contract test reads its keys there.
- kernel_event.py (2,149) stays over the 1,200-line module limit it was already over, and stays in the module-length baseline, 19 lines shorter.
If this simplifies or refactors:
- Mechanisms removed: listed under What changes in each module.
- Contract preserved: every in-repo caller and all persisted data.
- Tests: as above.
Observable effect: none; see the section above for why, and for the out-of-contract inputs that now behave differently.
Breaking changes: no. assemble_kernel_ext keeps its signature and now needs every kernel section in parts, which every caller already passes; _lane_attempt and _geak_settlement are private.
PR addresses single concern: yes, bringing the two over-gate session-breakdown functions under the gate.
Root cause is upstream: no.

Known merge interactions

Trial merges used git merge-tree --write-tree against the current head of every open PR that shares a file with this branch, plus the five whose file list GitHub cuts off at 100 files. A conflict counts as this branch's only when the same merge onto main (931af8a) does not have it.

fix(breakdown): drop never-populated 6.1 keys and report unmeasured figures as null #2047 (lishuoshuo-amd, "drop never-populated 6.1 keys and report unmeasured figures as null") and fix(breakdown): record the GEAK candidate at close and kernel e2e gain from the rebench #2048 (lishuoshuo-amd, "record the GEAK candidate at close and kernel e2e gain from the rebench") share kernel_event.py, and both merge their kernel_event.py changes cleanly with this branch.
- fix(breakdown): drop never-populated 6.1 keys and report unmeasured figures as null #2047 removes recorder fields this branch does not touch.
- fix(breakdown): record the GEAK candidate at close and kernel e2e gain from the rebench #2048 changes record_geak_attempts, reject_geak_attempts, the body of _integrate_e2e and one call inside _settle. This branch calls _integrate_e2e and _settle with the same inputs as before, and _integrate_e2e([]) is still None there.
- Each conflicts in the baseline file on one line, the module-lines entry of kernel_event.py, which all three PRs lower: to 2,161 in fix(breakdown): drop never-populated 6.1 keys and report unmeasured figures as null #2047, 2,167 in fix(breakdown): record the GEAK candidate at close and kernel e2e gain from the rebench #2048 and 2,149 here. Whichever lands second records the merged file's length. That conflict does not exist against main alone; fix(breakdown): record the GEAK candidate at close and kernel e2e gain from the rebench #2048's two other conflicts (orchestrator/phases/close.py, orchestrator/state/_shared_state/render.py) do.
refactor(kernelforge): delete the dead and repeated code in twelve over-gate functions #2026 and refactor(breakdown, kernelforge): delete what eleven over-gate functions re-checked #2052, earlier PRs in this series, share only the baseline file and conflict there on neighbouring lines: refactor(kernelforge): delete the dead and repeated code in twelve over-gate functions #2026 deletes the two rewrite_by_flydsl function-lines entries just above the kernel_event.py module-lines entry lowered here, and refactor(breakdown, kernelforge): delete what eleven over-gate functions re-checked #2052 deletes the three entries just above the assemble_kernel_ext function-lines entry deleted here. Keep both edits. Neither conflict exists against main alone.
fix(report): render fields on the axis and scope they were recorded on #2043 and fix(breakdown): rename throughput_tok_s_per_gpu to output_throughput (schema 6.1) and record per-GPU output on baseline and validation rows #2023 (lishuoshuo-amd) change other renderers (final.py, baseline.py and more), not attribution.py; both trial merges are clean.
Add RDNA4 GPU support (RX 9070 XT, RX 9070, RX 9060 XT, RX 9000 XT) f… #1032, Add xDiT kernel source mapping support #1154 and Set --pids-limit in docker-mode launch templates (fixes silent hangs + OpenBLAS crashes under Ray + TP>1) #1276 are stale: each has hundreds of conflicts with main, attribution.py among them, and exactly the same conflicts with this branch.
Of the other open PRs that share only the baseline file, 37 merge cleanly and 11 conflict there (feat(explore): opt-in pre-filters that drop EXPLORE variants through InferaSim before they are benchmarked #1494, xDiT profiling and TraceLens analysis support #1762, fix(agentx): serve the replay with its prefix cache, and stop reading an unmeasured round as a result #1867, fix: judge explore variants and warm recipes by what they change about the running launch #1908, AgentX: Move profiler setup and patching to common path #2025, fix(prelude): keep a skip_to_close hint from ending PRELUDE once a baseline has landed #2033, feat(kernelforge): spend the first round on a knowledge base batch with --wiki-guided #2036, feat(kernelforge): collect KernelForge runs into the Experience KB #2037, Build a specialist's agent-CLI launch and trace through one method each #2065, Run specialists on the DeepSeek Harness as dsh CLI subprocesses #2066, fix(specialist): keep specialists out of shared trees and dependencies — prompt guardrails plus post-run checks #2068) with the same number of conflict hunks against main alone, none of them on an entry this PR changes.

Gate

The code-metrics gate (scripts/code_metrics.py --base-ref origin/main, run locally) passes on the 3 touched files: 0 new, 0 worse and 0 out-of-date entries in every dimension, and 0 findings from the comment, English-only, test-import, repeated-literal and A1–A12 checks. Baseline changes:

removed cognitive-complexity …/kernel_event.py::assemble_kernel_ext 50;
removed function-lines …/kernel_event.py::assemble_kernel_ext 206;
removed function-lines …/_renderers/attribution.py::render 85;
lowered module-lines …/kernel_event.py from 2168 to 2149.

ruff check and ruff format --check are clean on the touched files.

Delete the second read of attribution.available: the "no per-source split" fact is now the elif of the "no ledger" fact, which is what the old `not has_any_split and available` test spelled out. Delete the `notes` list that was built up front only to be looped at the end; the notes are appended where they always landed, as the last facts. The kept decisions become a comprehension over the same rows and filter, and the markdown table is passed straight into the section instead of through a one-use local. Behaviour is identical: the facts, their order, the decisions and the table are unchanged, nothing between the old and new position of the notes read can raise, and an f-string formats every JSON value exactly as str() did. render: cyclomatic 11 -> 8, cognitive 28 -> 24, lines 85 -> 79. Co-authored-by: Cursor <cursoragent@cursor.com>

…of its sections Delete the eight hand-written read-and-sort blocks in assemble_kernel_ext: each section's sort fields are now one row of _ROW_ORDER, read by a single comprehension. Delete the second pass that patched an e2e block onto attempts after they were built: _lane_attempt now joins its integrate gate rows while it builds the row, which is the only place the join could fire (a GEAK or acceptance attempt never carries an integrate_ref). Delete _geak_settlement and its (ref, conflicts) tuple: _geak_rebench builds the wire rebench block directly, the attempts read settled_against and conflicting_decisions off it, and the GEAK engagement test reads its `required` field instead of re-deriving the same ledger-or-stated test. Two guards go as dead. `parts.get(section) or []`: every caller builds parts with event_parts(), which fills every requested section with a list (the recorder's own close, _republish_closed_event, event_finalize's kernel spec over EVENT_SECTIONS, and assembler.kernel_event_parts). The entry.route fallback for outcome.route: the header's only writer, KernelEventRecorder.__init__, has put `route` and `entry.route` in the same record with the same value since kernel_event.py was created (5085b93). What stays per wire block moves into _forge, _geak and _outcome, each returning its schema block, so the ext literal the schema contract test pins stays in assemble_kernel_ext. Block contents, key order and the order of the roofline and baseline sub-assemblies are unchanged; a seeded differential over 6000 random events is byte-identical except for a header whose route is blank while entry.route is set, which no writer produces. assemble_kernel_ext: cyclomatic 11 -> 3, cognitive 50 -> 11, lines 206 -> 77. Co-authored-by: Cursor <cursoragent@cursor.com>

…ork cleared Delete the three entries for functions now under the gate: kernel_event.py::assemble_kernel_ext (cognitive 50, lines 206) and attribution.py::render (lines 85). Lower kernel_event.py's module-lines entry from 2168 to its new length, 2149. No entry is raised or added. Co-authored-by: Cursor <cursoragent@cursor.com>

github-actions · 2026-10-21T03:12:47Z

Code metrics gate: PASSED

Judged: the 3 files this change touches. Files it does not touch never fail it; anything found there is listed under Outside this change for information.

Dimension	Limit	Baseline entries	vs base
Cyclomatic complexity	fails > 20	14	+0
Cognitive complexity	fails > 30	89	-1
Function length (lines)	fails > 80	221	-2
Nested block depth	fails > 5	8	+0
Parameter count	fails > 5	87	+0
Module length (lines)	warns > 800, fails > 1200	47	+0
Duplicated lines per file	fails > 0	95	+0
Dead code (vulture)	fails > 0	2	+0

Check	Findings	Effect
Comments	0
English only	0
Production code imports test code	0
Repeated literals	0
A1 Private name of another package	0
A8 Flag argument	0
A9 Tuple return	0
A10 Raw vocabulary comparison	0
A11 Hidden global write	0
A12 Pure forwarder	0
A12 Single-implementation abstraction	0

61 findings in files this change does not touch

Outside this change (informational): 61

Unit	Dimension	Baseline	Now
`src/hyperloom/orchestrator/state/shared_state.py` `resolve_graded_comparison`	Cognitive complexity	absent	34
`src/hyperloom_kb/collect/mapping.py` `_experience`	Cognitive complexity	absent	31
`src/kernelforge/cli/forge_loop_command.py` `forge_loop`	Cognitive complexity	absent	31
`src/kernelforge/loop/runner.py` `IterationLoop._collect_session_candidate`	Cognitive complexity	absent	44
`src/kernelforge/loop/runner.py` `IterationLoop._rule_on_attempt`	Cognitive complexity	absent	38
`src/hyperloom/inference_optimizer/cli/__init__.py` `_reexport_session_environment`	Function length (lines)	absent	107
`src/hyperloom/inference_optimizer/cli/__init__.py` `_resolve_agent_lineup`	Function length (lines)	absent	89
`src/hyperloom/inference_optimizer/cli/__init__.py` `_resolve_run_topology`	Function length (lines)	absent	92
`src/hyperloom/inference_optimizer/multi_node/scripts/launch_multinode.py` `_build_parser`	Function length (lines)	absent	84
`src/hyperloom/inference_optimizer/multi_node/scripts/launch_multinode.py` `_spawn_rank_actors`	Function length (lines)	absent	84
`src/hyperloom_kb/sync.py` `GlobalSync.pull`	Function length (lines)	absent	84
`src/kernelforge/loop/runner.py` `IterationLoop._close_iteration`	Function length (lines)	absent	117
`src/kernelforge/loop/runner.py` `IterationLoop._collect_session_candidate`	Function length (lines)	absent	185
`src/kernelforge/loop/runner.py` `IterationLoop._implementer_history`	Function length (lines)	absent	112
`src/kernelforge/loop/runner.py` `IterationLoop._rule_on_attempt`	Function length (lines)	absent	102
`src/kernelforge/loop/runner.py` `IterationLoop._source_attempt`	Function length (lines)	absent	175
`src/hyperloom_kb/http_service.py`	Module length (lines)	absent	1525
`src/kernelforge/cli/forge_loop_command.py`	Module length (lines)	absent	2003
`src/hyperloom/orchestrator/actions/executors/_kv_metrics.py` `port_from_server_log`	Nested block depth	absent	6
`src/hyperloom/orchestrator/specialists/dispatch.py` `SpecialistDispatchCollaborator.warm_specialist_params`	Cognitive complexity	93	94
`src/hyperloom/agents/quantization/driver/retry.py` `quantize_via_prompt`	Function length (lines)	102	103
`src/hyperloom/inference_optimizer/breakdown/exporter.py` `build`	Function length (lines)	132	136
`src/hyperloom/inference_optimizer/cli/__init__.py` `_run_optimize`	Function length (lines)	273	279
`src/hyperloom/inference_optimizer/cli/preflight.py` `_preflight`	Function length (lines)	203	208
`src/hyperloom/inference_optimizer/multi_node/scripts/kernel_patch_multinode.py` `_do_apply`	Function length (lines)	98	99

...and 36 more; the job log lists every row.

Verdict computed by the gate scripts in /home/runner/work/_temp/code-metrics-gate, not by this tree's copy.

_{Thresholds and sources: [tool.hyperloom.code_metrics] in pyproject.toml. Tools: complexipy 8.0.1, jscpd 5.4.1, ruff 0.16.2, vulture 2.16.}

…akdown-2-pr Co-authored-by: Cursor <cursoragent@cursor.com>

ZhengGong-amd

Reviewed: gate numbers re-measured (ruff C901, complexipy, function lines) and match the description; each removed guard / "producer always sets it" claim checked against its producers and call sites; base vs head differential runs show no undisclosed behaviour change; tests covering the touched modules pass.

LGTM.

…akdown-2-pr Co-authored-by: Cursor <cursoragent@cursor.com>

fengshaoyi-amd and others added 3 commits October 21, 2026 10:47

fengshaoyi-amd mentioned this pull request Oct 21, 2026

refactor(enablement, measurement): delete the dead and repeated code in seven over-gate functions #2076

Merged

Merge remote-tracking branch 'origin/main' into lean/shafeng/cc20-bre…

67eb0f8

…akdown-2-pr Co-authored-by: Cursor <cursoragent@cursor.com>

fengshaoyi-amd marked this pull request as ready for review October 21, 2026 06:46

fengshaoyi-amd requested a review from a team as a code owner October 21, 2026 06:46

ZhengGong-amd approved these changes Oct 21, 2026

View reviewed changes

fengshaoyi-amd and others added 2 commits October 21, 2026 17:54

Merge remote-tracking branch 'origin/main' into lean/shafeng/cc20-bre…

8fa7c14

…akdown-2-pr Co-authored-by: Cursor <cursoragent@cursor.com>

Merge remote-tracking branch 'origin/main' into lean/shafeng/cc20-bre…

35dd80d

…akdown-2-pr Co-authored-by: Cursor <cursoragent@cursor.com>

xiaofei-zheng merged commit c76424b into main Oct 22, 2026
40 checks passed

xiaofei-zheng deleted the lean/shafeng/cc20-breakdown-2-pr branch October 22, 2026 08:05

Navigation Menu

Repository navigation

refactor(breakdown): delete the dead and repeated code in two over-gate kernel-event and attribution-renderer functions - #2074

refactor(breakdown): delete the dead and repeated code in two over-gate kernel-event and attribution-renderer functions#2074
xiaofei-zheng merged 6 commits into
mainfrom
lean/shafeng/cc20-breakdown-2-pr

fengshaoyi-amd commented Oct 21, 2026

Uh oh!

github-actions Bot commented Oct 21, 2026 •

edited

Loading

Outside this change (informational): 61

Uh oh!

ZhengGong-amd left a comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Conversation

fengshaoyi-amd commented Oct 21, 2026

What this does

Gate numbers per function

What changes in each module

Kernel event assembly (recorder/kernel_event.py)

Attribution renderer (reporters/_renderers/attribution.py)

Why the deletions are safe

Observable effect

Checklist

Known merge interactions

Gate

Uh oh!

github-actions Bot commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Code metrics gate: PASSED

Outside this change (informational): 61

Uh oh!

ZhengGong-amd left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Kernel event assembly (`recorder/kernel_event.py`)

Attribution renderer (`reporters/_renderers/attribution.py`)

github-actions Bot commented Oct 21, 2026 •

edited

Loading