Closed nilfm99 closed 3 months ago
Benchmarks on 10-bit input on a Macbook M1 show no significant difference. Benchmarks on 10-bit input on an ubuntu machine with AMD cpu show 1% performance degradation.
Both cases running just CAMBI, single-threaded.
LGTM for now. Might need to revisit if we need to go after CAMBI speedups.
Replaces #1352. Benchmarks on my laptop show no observable overhead wrt master but need to check further.
(my benchmarks were on a 10-bit input so it's not just because of the early return in 8-bit and 16-bit cases)