autotune~ uses float instead of t_sample
Looks like [autotune~]
hard-coded float
in some places instead of t_sample
.
Fix this and test to make sure the algo itself doesn't depend on any single-precision float bitmath or anything else like that.