4 ms·
Seems like BFloat32 would have the same exponent range as Float64 (double precision), and a correspondingly shorter significand.
by ScottBurson 7y ago
Seems like BFloat32 would have the same exponent range as Float64 (double precision), and a correspondingly shorter significand.