Closed Dawoodoz closed 1 year ago
The backends are now partially vectorized in the experimental branch. Fully vectorizing them would require extending simd.h with signed 16-bit integers or adding 16-bit packed memory write functions.
Creating a vectorized sound engine would require knowing the padded buffer lengths, which requires changing the raw data pointer into SafePointer
Merged into master.
Aligning the sound buffers and vectorizing the conversions would allow creating SIMD vectorized sound engines on top of the framework. A simple sound engine where sounds must be pre-scaled when changing playback speed to avoid runtime interpolation, might be fast enough even on a weak ARM processor as long as it has NEON.