VHDL-200X-FT review meeting. 7-Feb-2005 Attendees: Chuck Swart Charlie Guy David Bishop Jim Lewis Depack Pant Ajay Varikat Karl Eisenhofer Next Phone Meeting: Monday Feb 28th, 2005 2-4 PM PST Agenda Review of FT 26 Fixed Point Package Issues Discussion FT 26 looks good In numeric_std, the functions to_unsigned and to_signed currently provide an assertion of level warning if the integer value to be represented is too large for the corresponding unsigned/signed value. In fixed point point package: to_unsigned(ufixed, ...) to_signed(sfixed, ...) Will saturate or do modulo math based on parameters _Currently produces warnings if wraps or saturates_ to_ufixed(...) (integer, real, unsigned, signed, ...) to_sfixed(...) to_float(...) Will saturate or modulo based on parameters, but no warnings when wraps or saturates In numeric_std when a signed value wraps from a negative number to a positive number the result is adjusted so that it will be negative. Similarly when it wraps from a positive number to a negative number, the result is adjusted so that it will be positive. In fixed point package when a sfixed value wraps from a negative number to a positive number the result remains positive. Likewise when a sfixed value wraps from a positive number to a negative number the result remains negative. Functions: to_ufix, to_fix, to_ufix_high, to_ufix_low, ... Are intended as a special interface to some DSP packages and a note will be added to guide a user to the corresponding to_ufixed and to_sfixed functions. Names to be changed to "to_ufix, to_sfix, ufix_high, ufix_low, .... Add functions: ufixed_high, ufixed_low, sfixed_high, sfixed_low Review: from_string, ... Used for providing a literal value for fixed point types Have a version of abs(sfixed) that returns ufixed Size of values shall be such that they retain full precision of the number. Remove the "remove_sign (sfixed)" function as it will be redundant with the ABS(sfixed) added above.