Skip to content

Optimize arrays top down as execution reaches them, and retry a parent when its children change - #11090

Merged
robert3005 merged 6 commits into
developfrom
rk/optimize-execute-loop
Nov 4, 2026
Merged

robert3005 merged 6 commits into
developfrom
rk/optimize-execute-loop

Optimize each node top down and try it again when a child changes

9d6e6d6
Select commit
Loading
Failed to load commit list.
CodSpeed / CodSpeed Performance Analysis failed Nov 4, 2026 in 0s

Performance Regression: -17.68%

1 warning

⚠️ Walltime results may be unreliable

Some benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.

Learn more about Macro Runners

See affected benchmarks

⚡ 1 improved benchmark
❌ 203 regressed benchmarks
✅ 1561 untouched benchmarks
⏩ 130 skipped benchmarks1
🗄️ 47 archived benchmarks run2

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

Mode Benchmark BASE HEAD Efficiency
❌ CPU Simulation canonicalize_sparse_fixed_size_list[(1024, 17, 8)] 424.1 µs 971.8 µs -56.36%
❌ CPU Simulation canonicalize_sparse_fixed_size_list[(512, 7, 4)] 457.4 µs 820.6 µs -44.26%
❌ CPU Simulation canonicalize[256, 32] 36.3 µs 60.8 µs -40.35%
❌ Walltime decode_chunked_avx2[i64, 524288] 13.5 µs 22.2 µs -39.3%
❌ Walltime decode_chunked_avx512[i64, 524288] 13.5 µs 21.8 µs -38.17%
❌ CPU Simulation filter_random_by_random[10000] 6.8 µs 10.9 µs -37.48%
❌ CPU Simulation take_filter_primitive_slice_mask_sequential_indices[16384, 1000] 47.9 µs 76.4 µs -37.29%
❌ Walltime take_neon[value_shape/shuffled/fsl8/nonnull/chunks=16/indices=100] 32.1 µs 50.5 µs -36.48%
❌ CPU Simulation canonicalize[16, 8] 25.6 µs 40.3 µs -36.36%
❌ CPU Simulation canonicalize[256, 8] 25.7 µs 40.3 µs -36.31%
❌ CPU Simulation take_filter_primitive_slice_mask_sequential_indices[4096, 1000] 50.2 µs 78.6 µs -36.09%
❌ CPU Simulation canonicalize[256, 2] 22.3 µs 34.6 µs -35.46%
❌ CPU Simulation fsl_i32[16, 0.01] 19.4 µs 29.6 µs -34.38%
❌ CPU Simulation canonicalize_sparse_list[(512, 7, 4)] 426.9 µs 649.6 µs -34.28%
❌ CPU Simulation canonicalize_sparse_list[(1024, 17, 8)] 368.4 µs 554.1 µs -33.52%
❌ CPU Simulation canonicalize[16, 32] 50.4 µs 74.9 µs -32.7%
❌ CPU Simulation decode[msp_i32_p18_nullable] 18.6 µs 27.4 µs -32.02%
❌ CPU Simulation take_filter_primitive_nullable_slice_mask_random_indices[4096, 1000] 94.5 µs 134.9 µs -29.92%
❌ CPU Simulation take_filter_primitive_nullable_slice_mask_random_indices[16384, 1000] 112.5 µs 155.2 µs -27.54%
❌ CPU Simulation take_filter_primitive_small_uncached_random_mask_random_indices[4096, 10] 21.1 µs 29 µs -27.27%
... ... ... ... ... ...

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 rk/optimize-execute-loop (9d6e6d6) with develop (d5271f9)

Open in CodSpeed

Footnotes

  1. 130 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. ↩

  2. 47 benchmarks were run, but are now archived. If they were deleted in another branch, consider rebasing to remove them from the report. Instead if they were added back, click here to restore them. ↩