Capability of using float16/float32 in waveforms

[kazewong]

Tensor core in Nvidia GPU supposedly does wonder, but they need either float16 or float32 to work with. See if this help boost the performance

[tedwards2412]

I've removed the default float64 requirement within ripple so we can start trialing this in Jim