vhdl error: integer literal cannot have negative exponent

Question

The Quartus compiler reports the said error message at this line

constant LIMIT : float (5 downto -27) := 1E-12;

the constant is float type, why is it treated as an integer?

This constant is used only in if-statements and is compared to another float type variable.

Answer 1

The literal 1E-12 is an integer, since the mantissa (1 before 'E') is an integer; for a real type literal type write 1.0E-12.

But that returns a VHDL real type, and not a float as used in the VHDL float_generic_pkg package. For a float type constant use the to_float functions in the float_generic_pkg package.