Write a C program called bitflip to run on ocelot which will flip bits in an unsigned integer using the binary representation of the number. The user should input the original integer value between 1 ...

FP-NN (digital CIM, INT8) : 8 bits per weight, positional encoding value = sum_{b=0..7} bit_b * 2^b MSB carries 50% of the dynamic range. A single MSB flip changes the weight by ~50% of full scale.