I want to port a program from C to RTL that includes a 283*696 array of floats. Unfortunately, this is too big for the BRAM to handle. I realize that this would introduce a loss of precision, but would it be reasonable to use a declared fixed data type in order to save space? Note that the precision of the floats is never past the 10,000ths in decimal, and never greater than 7 in decimal as well. Is this a feasible implementation if I'm currently about 150% above my alloted BRAM for my device?