Repository navigation
Compare primitive values through the shared RowFn executor - #10676
connortsui20 wants to merge 2 commits into
1 benchmark regressed
2 warnings
⚠️ Walltime results may be unreliableSome benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.
⚠️ Base and head ran on different runner typesSome benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.
⚡ 28 improved benchmarks
❌ 1 regressed benchmark
✅ 2275 untouched benchmarks
Warning
Please fix the performance issues or acknowledge them on CodSpeed.
Performance Changes
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ❌ | Walltime | mul_u64_nonnull_neon |
20.3 µs | 39 µs | -47.93% |
| ⚡ | CPU Simulation | filter_powerlaw_by_random[10000] |
590.3 ns | 318.7 ns | +85.2% |
| ⚡ | Walltime | compare_u64_avx2 |
3.6 µs | 2.8 µs | +28.36% |
| ⚡ | Walltime | compare_float_avx2 |
4.8 µs | 3.7 µs | +27.51% |
| ⚡ | CPU Simulation | missing_stats[16] |
85.4 µs | 67.4 µs | +26.75% |
| ⚡ | CPU Simulation | bench_compare_varbin[(50000, 2)] |
162.9 µs | 133.7 µs | +21.82% |
| ⚡ | CPU Simulation | bench_compare_varbin[(50000, 64)] |
164.3 µs | 135.3 µs | +21.5% |
| ⚡ | Walltime | lt_i64_nullable_avx2 |
8.3 µs | 7 µs | +18.05% |
| ⚡ | CPU Simulation | random_i8[0.01] |
113 µs | 97.5 µs | +15.92% |
| ⚡ | CPU Simulation | test_repeated_scan_filter[int] |
2.4 ms | 2.1 ms | +15.86% |
| ⚡ | CPU Simulation | fsl_i32[16, 0.01] |
13.4 µs | 11.6 µs | +15.55% |
| ⚡ | CPU Simulation | bench_compare_sliced_dict_varbinview[(7500, 10000)] |
228.6 µs | 200.3 µs | +14.13% |
| ⚡ | CPU Simulation | compare_in_range[widen] |
161.6 µs | 142 µs | +13.79% |
| ⚡ | CPU Simulation | bench_compare_sliced_dict_varbinview[(10000, 10000)] |
234.8 µs | 206.5 µs | +13.75% |
| ⚡ | CPU Simulation | bench_compare_sliced_dict_varbinview[(1000, 10000)] |
90.4 µs | 79.5 µs | +13.66% |
| ⚡ | CPU Simulation | test_scan_filter[int] |
2.9 ms | 2.5 ms | +13.45% |
| ⚡ | CPU Simulation | test_scan_filter_to_arrow_array[int] |
2.8 ms | 2.4 ms | +13.39% |
| ⚡ | Walltime | compare_f32_avx2 |
2.8 µs | 2.5 µs | +13.28% |
| ⚡ | Walltime | add_constant_shapes_neon[(32768, PerRowConstant)] |
15.8 µs | 14.1 µs | +12.25% |
| ⚡ | Walltime | compare_int_eq_avx2 |
3.2 µs | 2.8 µs | +11.76% |
| ... | ... | ... | ... | ... | ... |
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 ct/row-fn-cleanup/primitive-compare (adcf673) with ct/row-fn-cleanup/fused-execution (37c7a3e)
