VoidXH/Cavern

QMFB optimization - round 3

Closed this issue · 2 comments

It's that time of the year again when I try to optimize a CPU-resistant algorithm which is already over 10x real-time anyway.

Let's try creating an FFT cache with the twisted AC-3 coefficient generation method. And have tests for QMF so it's not needed to rely on conversions for testing.

It looks like it's DCT-IV.

Did not work as usual.