Repository navigation
SPF: ordered bucket queue, reusable workspace, fixed-routing reuse (0.11.0) - #11
Merged
Merged
Conversation
….11.0) Dijkstra in shortest_paths / shortest_paths_to runs on a monotone bucket queue whenever max_cost / gcd(costs) + 1 <= 65536, with the active bucket ordered by the heap's (-bottleneck, node) key so every output is bit-identical; wider costs keep the binary heap. Scratch state is a thread-local workspace reset per search (touched nodes only when few were touched). Settled neighbours are skipped before their parallel edges are scanned, the predecessor output is sized from live per-node counts, a frontier of one node is held inline, and bucket storage retention is bounded. calc_max_flow / place_max_flow with require_capacity=false compute their cost-only search once per call; FlowPolicy and the spf bindings move the DAG instead of copying it. StrictMultiDiGraph gains max_cost() / cost_gcd(); the C++ search functions take a trailing SpfQueue for tests and benchmarks, and NGRAPH_CORE_SPF_QUEUE=heap|bucket overrides the choice process-wide. Verified by an old-versus-new bit-gate (identical SHA-256 over SPF, max-flow, batch, sensitivity, FlowPolicy and KSP outputs), 1,351,168 bit-exact harness comparisons on both routes, the C++ differential tests, lint/types, 177 C++ and 466 Python tests, ASan/UBSan, and NetGraph's integration suite (1,278 tests). Measured old/new/old: 1.4-1.8x on weighted graphs, 1.2-1.35x on unit-cost fabrics, 2.7-3x on near-destination searches at the Core level; 1.06-1.3x through NetGraph. Plan, harness, bit-gate, benchmark, recorded results and the independent review live under dev/perf/spf_queue; the preceding Metal GPU research under dev/research. dev/ is excluded from the sdist. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
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.
Summary
Dijkstra in
shortest_paths/shortest_paths_tonow runs on a monotone bucket queue whenever the graph's edge costs fit 65,536 buckets (max_cost / gcd(costs) + 1), which covers unit, uniform and small-integer cost graphs including zero-cost pseudo attachments; wider or int64-offset costs keep the binary heap. The active bucket is ordered by the same(-bottleneck, node)key as the heap, so the pop order and every output (distances, predecessor order, single-path capacity ties, early exit) are bit-identical. Scratch arrays live in a thread-local workspace reset per search (only the touched nodes when few were touched), removing the O(N) initialisation that dominated small destination-limited searches on large graphs.Adopted from an independent review of the first version: settled neighbours are skipped before their parallel edges are scanned, the predecessor output is sized from live per-node counts, a frontier of one node is held inline, bucket storage retention is bounded, an exception-path ordering bug was fixed,
calc_max_flow/place_max_flowwithrequire_capacity=falsecompute their cost-only search once per call, and two redundant DAG copies were removed.Also:
StrictMultiDiGraph::max_cost()/cost_gcd(), a trailingSpfQueueparameter on the C++ search functions for tests and benchmarks,NGRAPH_CORE_SPF_QUEUE=heap|bucket, and the developer tooling behind the change underdev/perf/spf_queue/(plan, parity harness, old-versus-new bit-gate, NetGraph-level benchmark, recorded A/B/A results, review record) plus the preceding Metal GPU research underdev/research/.dev/is excluded from the sdist. Version 0.11.0.Measured (old / new / old, fresh processes, Apple M4 Max shared with other sessions)
Every NetGraph workload digest was identical across old, new and the forced heap route. Details, raw samples and the independent review are in
dev/perf/spf_queue/PLAN.md.Verification
calc_max_flowin every placement mode,batch_max_flow,sensitivity_analysis,FlowPolicysequences and KSP: identical SHA-256 (6,226 output lines), also with the new build forced onto the heap.bash .superset/workspace.sh check(lint, types, 177 C++ tests, 466 Python tests),make sanitize-test(ASan/UBSan), NetGraphdev/check_core_integration.shagainst this tree (1,278 tests).🤖 Generated with Claude Code