Closed achille-roussel closed 2 years ago
Follow up to #289, this PR optimizes the decoding of DELTA_LENGTH_BYTE_ARRAY pages.
We get ~2x better throughput with this change compared to the parent branch:
name old time/op new time/op delta Decode/DELTA_LENGTH_BYTE_ARRAY/byte_array 24.0µs ± 0% 12.0µs ± 0% -50.07% (p=0.000 n=10+10) name old speed new speed delta Decode/DELTA_LENGTH_BYTE_ARRAY/byte_array 4.53GB/s ± 0% 9.07GB/s ± 0% +100.28% (p=0.000 n=10+10) name old value/s new value/s delta Decode/DELTA_LENGTH_BYTE_ARRAY/byte_array 416M ± 0% 834M ± 0% +100.29% (p=0.000 n=10+10)
Follow up to #289, this PR optimizes the decoding of DELTA_LENGTH_BYTE_ARRAY pages.
We get ~2x better throughput with this change compared to the parent branch: