The results of benchmarking gov2.sorted:
Code:
| Algorithm |Encoded Size |Ratio % |Encoding Speed|Decoding Speed|
|:-------------------|----------------:|--------:|-------------:|-------------:|
| VTEnc | 2,889,599,350| 12.08| 77.69 M/s| 69.08 M/s|
| Delta+FastPFor128 | 3,849,161,656| 16.09| 855.60 M/s| 850.43 M/s|
| Delta+FastPFor256 | 3,899,341,376| 16.30| 915.43 M/s| 916.01 M/s|
| Delta+BinaryPacking| 4,329,919,808| 18.10| 3.03 G/s| 2.94 G/s|
| Delta+VariableByte | 6,572,084,696| 27.48| 2.05 G/s| 2.09 G/s|
| Delta+VarIntGB | 7,923,819,720| 33.13| 2.54 G/s| 3.73 G/s|
| Copy | 23,918,861,764| 100.0| 6.27 G/s| - |
As I expected, VTEnc's encoding and decoding speed are poor. The encoding ratio is the best among the tested algorithms though.
For now, I'm going to stop doing more benchmarking and going to focus on improving the encoding and decoding speed.