Parquet Net 4.12.0 Performance
This release is about twice as fast comparing to version 4.9.1:
Method | Version | Mean | Error | StdDev | Allocated |
---|---|---|---|---|---|
WriteNullableInts | 4.12.0 | 6.205 ms | 3.359 ms | 0.1841 ms | 15.27 MB |
ReadNullableInts | 4.12.0 | 11.175 ms | 3.762 ms | 0.2062 ms | 6.68 MB |
WriteNullableInts | 4.5.0 | 13.690 ms | 9.804 ms | 0.5374 ms | 21.08 MB |
ReadNullableInts | 4.5.0 | 6.702 ms | 1.810 ms | 0.0992 ms | 15.27 MB |
WriteNullableInts | 4.9.1 | 11.862 ms | 5.538 ms | 0.3035 ms | 18.22 MB |
ReadNullableInts | 4.9.1 | 11.439 ms | 5.792 ms | 0.3175 ms | 6.68 MB |
This is primarily due to optimisations of how RLE/Bitpacking works internally, which is now itself about 14-18 times faster:
Method | Mean | Error | StdDev | Ratio | RatioSD |
---|---|---|---|---|---|
BitShifting | 0.5844 ns | 0.4151 ns | 0.0228 ns | 1.00 | 0.00 |
BitConverter | 8.6110 ns | 2.6350 ns | 0.1444 ns | 14.74 | 0.33 |
To contact me, send an email anytime or leave a comment below.