Skip to content

refactor: propagate bound expressions through layouts - #9126

Merged
joseph-isaacs merged 4 commits into
developfrom
ji/bound-expr-stack
Aug 5, 2026
Merged

refactor: propagate bound expressions through layouts#9126
joseph-isaacs merged 4 commits into
developfrom
ji/bound-expr-stack

chore: remove dead immediate access analysis

c98bf5e
Select commit
Loading
Failed to load commit list.
CodSpeed HQ / CodSpeed Performance Analysis failed Aug 5, 2026 in 0s

Performance Regression: -8.47%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚡ 2 improved benchmarks
❌ 8 regressed benchmarks
✅ 1875 untouched benchmarks
⏩ 1 skipped benchmark1

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

Mode Benchmark BASE HEAD Efficiency
Simulation chunked_canonicalize_into[(1000, 50, 8, 4)] 14.6 ms 17.5 ms -16.97%
Simulation chunked_canonicalize_into[(1000, 50, 8, 64)] 17.1 ms 20.3 ms -16%
Simulation chunked_canonicalize_into[(1000, 50, 8, 16)] 15.7 ms 18.6 ms -15.9%
Simulation chunked_canonicalize_into[(1000, 100, 16, 4)] 20.8 ms 24.2 ms -13.78%
Simulation chunked_into_canonical[(1000, 50, 8, 4)] 17.4 ms 19.7 ms -11.8%
Simulation chunked_canonicalize_into[(1000, 100, 16, 16)] 24.6 ms 27.9 ms -11.75%
Simulation chunked_into_canonical[(1000, 50, 8, 16)] 18.5 ms 20.9 ms -11.65%
Simulation chunked_into_canonical[(1000, 50, 8, 64)] 20.1 ms 22.7 ms -11.41%
WallTime cuda/bitpacked_u8/unpack/3bw[100M] 354.3 µs 300.5 µs +17.9%
Simulation chunked_varbinview_opt_into_canonical[(10, 1000)] 6.2 ms 5.5 ms +13.57%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.


Comparing ji/bound-expr-stack (c98bf5e) with develop (465fab3)

Open in CodSpeed

Footnotes

  1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports.