Repository navigation
feat(layout): add the ListPack exec node - #10418
joseph-isaacs wants to merge 1 commit into
Conversation
`ListPackNode` runs in two phases, each a readiness rule. It is `Ready::Closed(offsets)` until the offsets are whole, then computes the element range they reference, rebases nonzero first offsets, and spawns the elements. It is then `Ready::AllClosed`, and assembles the lists once. Holes in the parent selection are filtered after assembly, so parent and element row coordinates never mix. Signed-off-by: Joe Isaacs <joe.isaacs@live.co.uk> Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NWYmfHTFtaEdhdyd3u5hv3
Merging this PR will degrade performance by 36.21%
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Simulation | bench_compare_varbinview[(50000, 4)] |
145.1 µs | 400.4 µs | -63.76% |
| ❌ | Simulation | bench_compare_varbinview[(50000, 2)] |
145.3 µs | 400.3 µs | -63.72% |
| ❌ | Simulation | bench_compare_varbinview[(50000, 8)] |
146.1 µs | 400.7 µs | -63.53% |
| ❌ | Simulation | bench_compare_varbinview[(50000, 32)] |
147 µs | 401.5 µs | -63.39% |
| ❌ | Simulation | bench_compare_varbinview[(50000, 128)] |
149 µs | 404.6 µs | -63.17% |
| ❌ | Simulation | bench_compare_primitive[(50000, 4)] |
162.9 µs | 419 µs | -61.11% |
| ❌ | Simulation | bench_compare_primitive[(50000, 2)] |
163 µs | 418.5 µs | -61.04% |
| ❌ | Simulation | bench_compare_primitive[(50000, 8)] |
164 µs | 418.9 µs | -60.85% |
| ❌ | Simulation | bench_compare_primitive[(50000, 32)] |
164.7 µs | 420 µs | -60.8% |
| ❌ | Simulation | bench_compare_primitive[(50000, 128)] |
165.9 µs | 422 µs | -60.68% |
| ❌ | Simulation | bench_compare_varbin[(50000, 2)] |
169.1 µs | 424.3 µs | -60.14% |
| ❌ | Simulation | bench_compare_varbin[(50000, 4)] |
169.5 µs | 424.6 µs | -60.07% |
| ❌ | Simulation | bench_compare_varbin[(50000, 8)] |
170.2 µs | 424.7 µs | -59.93% |
| ❌ | Simulation | bench_compare_varbin[(50000, 32)] |
171.1 µs | 425.6 µs | -59.79% |
| ❌ | Simulation | bench_compare_varbin[(50000, 128)] |
173.4 µs | 428.6 µs | -59.55% |
| ❌ | Simulation | bench_compare_primitive[(50000, 2048)] |
216.4 µs | 501.9 µs | -56.89% |
| ❌ | Simulation | bench_compare_varbinview[(50000, 512)] |
198.7 µs | 455.1 µs | -56.33% |
| ❌ | Simulation | bench_compare_varbinview[(50000, 2048)] |
220.9 µs | 502.2 µs | -56.02% |
| ❌ | Simulation | bench_compare_primitive[(50000, 512)] |
209.6 µs | 472.5 µs | -55.64% |
| ❌ | Simulation | bench_compare_varbin[(50000, 2048)] |
226.6 µs | 506.1 µs | -55.23% |
| ... | ... | ... | ... | ... | ... |
ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing ji/exec-5-list-pack-node (152edd4) with ji/exec-4-eval-row-idx-take-nodes (b21ac66)
Footnotes
-
381 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩
Summary
Part of the exec plan stack extracted from #10349. Stacked on #10393.
Changes
ListPackexec node: it reads list offsets first, then the element range they reference, assembles the lists, and keeps the selected ones.🤖 Generated with Claude Code
https://claude.ai/code/session_01NWYmfHTFtaEdhdyd3u5hv3
Generated by Claude Code