Use Uint8 for GT field in statpopgen instead of UInt64#8050
Conversation
Signed-off-by: Mikhail Kot <mikhail@spiraldb.com>
54ef201 to
340e387
Compare
Merging this PR will not alter performance
|
| Mode | Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|---|
| ⚡ | Simulation | chunked_varbinview_canonical_into[(100, 100)] |
307.9 µs | 273.2 µs | +12.73% |
| ❌ | Simulation | new_alp_prim_test_between[f32, 16384] |
104.6 µs | 119.1 µs | -12.17% |
| ❌ | Simulation | new_alp_prim_test_between[f32, 32768] |
154 µs | 182.9 µs | -15.79% |
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing myrrc/statpopgen-u8 (340e387) with develop (f852d72)
🚨🚨🚨❌❌❌ SQL BENCHMARK FAILED ❌❌❌🚨🚨🚨Benchmark |
Polar Signals Profiling ResultsLatest Run
Powered by Polar Signals Cloud |
joseph-isaacs
left a comment
There was a problem hiding this comment.
lets audit all the fields
Benchmarks: PolarSignals ProfilingVortex (geomean): 0.980x ➖ datafusion / vortex-file-compressed (0.980x ➖, 0↑ 0↓)
|
File Sizes: PolarSignals ProfilingFile Size Changes (1 files changed, -0.0% overall, 0↑ 1↓)
Totals:
|
Benchmarks: FineWeb NVMeVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.881x ✅, 6↑ 0↓)
datafusion / vortex-compact (0.887x ✅, 7↑ 0↓)
datafusion / parquet (0.867x ✅, 9↑ 0↓)
duckdb / vortex-file-compressed (0.927x ➖, 3↑ 0↓)
duckdb / vortex-compact (0.901x ➖, 6↑ 0↓)
duckdb / parquet (0.877x ✅, 8↑ 0↓)
Full attributed analysis
|
File Sizes: FineWeb NVMeFile Size Changes (2 files changed, -0.0% overall, 0↑ 2↓)
Totals:
|
Benchmarks: TPC-H SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.987x ➖, 0↑ 0↓)
datafusion / vortex-compact (0.976x ➖, 0↑ 0↓)
datafusion / parquet (0.963x ➖, 3↑ 1↓)
datafusion / arrow (0.990x ➖, 0↑ 1↓)
duckdb / vortex-file-compressed (0.996x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.988x ➖, 0↑ 0↓)
duckdb / parquet (0.996x ➖, 0↑ 0↓)
duckdb / duckdb (0.990x ➖, 1↑ 0↓)
Full attributed analysis
|
File Sizes: TPC-H SF=1 on NVMEFile Size Changes (18 files changed, -0.0% overall, 0↑ 18↓)
Totals:
|
Benchmarks: TPC-DS SF=1 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.998x ➖, 1↑ 1↓)
datafusion / vortex-compact (0.994x ➖, 1↑ 0↓)
datafusion / parquet (0.997x ➖, 1↑ 1↓)
duckdb / vortex-file-compressed (1.000x ➖, 1↑ 1↓)
duckdb / vortex-compact (1.007x ➖, 1↑ 2↓)
duckdb / parquet (0.998x ➖, 1↑ 0↓)
duckdb / duckdb (0.983x ➖, 3↑ 1↓)
Full attributed analysis
|
File Sizes: TPC-DS SF=1 on NVMEFile Size Changes (48 files changed, -0.0% overall, 0↑ 48↓)
Totals:
|
Benchmarks: FineWeb S3Verdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.988x ➖, 0↑ 1↓)
datafusion / vortex-compact (0.907x ➖, 1↑ 0↓)
datafusion / parquet (1.058x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (0.838x ➖, 1↑ 0↓)
duckdb / vortex-compact (0.907x ➖, 1↑ 0↓)
duckdb / parquet (0.978x ➖, 0↑ 0↓)
Full attributed analysis
|
Benchmarks: TPC-H SF=10 on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (0.935x ➖, 7↑ 0↓)
datafusion / vortex-compact (0.971x ➖, 5↑ 0↓)
datafusion / parquet (0.948x ➖, 7↑ 0↓)
datafusion / arrow (0.988x ➖, 3↑ 0↓)
duckdb / vortex-file-compressed (0.996x ➖, 0↑ 0↓)
duckdb / vortex-compact (0.971x ➖, 4↑ 0↓)
duckdb / parquet (0.957x ➖, 3↑ 0↓)
duckdb / duckdb (0.992x ➖, 0↑ 0↓)
Full attributed analysis
|
File Sizes: TPC-H SF=10 on NVMEFile Size Changes (48 files changed, -0.0% overall, 0↑ 48↓)
Totals:
|
Benchmarks: Clickbench on NVMEVerdict: No clear signal (low confidence) datafusion / vortex-file-compressed (1.012x ➖, 0↑ 2↓)
datafusion / parquet (1.010x ➖, 0↑ 0↓)
duckdb / vortex-file-compressed (1.004x ➖, 1↑ 3↓)
duckdb / parquet (1.001x ➖, 0↑ 1↓)
duckdb / duckdb (0.994x ➖, 0↑ 0↓)
Full attributed analysis
|
File Sizes: Clickbench on NVMEFile Size Changes (201 files changed, -0.0% overall, 0↑ 201↓)
Totals:
|
Benchmarks: TPC-H SF=1 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (0.881x ➖, 5↑ 0↓)
datafusion / vortex-compact (1.008x ➖, 0↑ 1↓)
datafusion / parquet (0.953x ➖, 3↑ 3↓)
duckdb / vortex-file-compressed (1.010x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.032x ➖, 0↑ 0↓)
duckdb / parquet (1.007x ➖, 0↑ 0↓)
Full attributed analysis
|
Benchmarks: TPC-H SF=10 on S3Verdict: No clear signal (environment too noisy confidence) datafusion / vortex-file-compressed (1.005x ➖, 1↑ 2↓)
datafusion / vortex-compact (1.050x ➖, 2↑ 5↓)
datafusion / parquet (0.939x ➖, 1↑ 2↓)
duckdb / vortex-file-compressed (1.026x ➖, 0↑ 0↓)
duckdb / vortex-compact (1.028x ➖, 0↑ 0↓)
duckdb / parquet (1.121x ➖, 0↑ 2↓)
Full attributed analysis
|
With current U64 we're mostly measuring Duckdb upcast performance.
list_sum(GT) upcasts GT to u128 which takes most CPU time.
Since GT can fit in U8, use this to measure our performance.