Fails ubsan because memcpy declares args as non-null

No longer needed now that PLC is trained with PyTorch stack

Starting with compute_linear()

Gets us about 2x speedup on x86

And references to nnet_data.h

Remove old pitch estimator and retrain all models

Totally untested -- most likely doesn't work

Untested

saves a few cycles

And missing prototypes

Should be able to handle all previous GRU variants and more.

Marcus Asteborg authored 1 year ago and Jean-Marc Valin committed 1 year ago

Signed-off-by: Jean-Marc Valin <jmvalin@amazon.com>

first attempt of C implementation of fec encoder (not tested yet due to NEON/DOT_PROD not being separable)