CompactDNA is not effective for small length sequences #2
Description
| Sequence Length | Bases/µs | Sequences/µs |
|---|---|---|
| 5 | 818.21 | 163.64 |
| 10 | 1,593.39 | 159.34 |
| 15 | 2,454.58 | 163.64 |
| 20 | 2,418.44 | 120.92 |
| 25 | 4,066.13 | 162.65 |
| 30 | 4,908.97 | 163.63 |
| 35 | 5,761.40 | 164.61 |
| 40 | 6,560.00 | 164.00 |
The reason smaller sequences perform worse may be because of how I extract score from u64.
Currently I will pack as many whole sequences as I can into u64.
I use 3bit per base, so 21 bases fit into one u64.
This means if we pack a word with multiple words, we have to take extra steps per u64 in order to extract. It would likely be better to use a different usize than to pack many words into one u64. There may be more effective ways to sum are scores.
It may be better to just pack 1 word into most effective usize.