overflow with signed and unsigned values

[Zaki]

Hello,

I want to design an ALU with signed and unsigned addition.
The problem is the overflow:

SIGNAL Sum : STD_LOGIC_VECTOR( 8 DOWNTO 0 );
SIGNAL ALU_output_mux : STD_LOGIC_VECTOR( 7 DOWNTO 0 );
SIGNAL Ainput, Binput : STD_LOGIC_VECTOR( 7 DOWNTO 0 );

Sum <= unsigned('0' & Ainput) + unsigned('0' & Binput);
ALU_output_mux <= Sum(7 DOWNTO 0) ;
Overflow <= Sum(8);

Overflow would be '1' when Sum is bigger than 255.
This works for unsigned addition well but I need that overflow for signed
values.
Why I cant write:

Overflow <= '1' when (Sum >= 128 or Sum < 127) else
'0';

Can someone help me?

Thx

[Nicolas Matringe]

Zaki a écrit:

> Why I cant write:
>
> Overflow <= '1' when (Sum >= 128 or Sum < 127) else
> '0';
>
> Can someone help me?

Your condition looks like it is always true. Maybe you should write "Sum
< -127" ?

--
____ _ __ ___
| _ \_)/ _|/ _ \ Adresse de retour invalide: retirez le -
| | | | | (_| |_| | Invalid return address: remove the -
|_| |_|_|\__|\___/

[Egbert Molenkamp]

"Sum" is declared as std_logic_vector.
I don't think the "<=" is overloaded for left operand of this type and right
operand of type (un)signed.
Change type of SUM in SIGNED(8 downto 0); (Since you want twoscomplement)
Then a type conversion is needed for the ALU_output_mux:
ALU_output_mux <= std_logic_vector( Sum(7 DOWNTO 0) );
Finally, as mentioned by Nicolas
Sum >= 128 or Sum < 127
is a little bit funny. In twoscomplement the most negative value is -128 and
most positive 127
so this should be something like:
Sum >= 128 or Sum <= -129

An alternative solution (without extension of the operands). Look at the
sign bits:
overflow <= '1' when (Ainput(7)=Binput(7)) and (Ainput(7)/=Sum(7)) ELSE
'0';

Egbert Molenkamp

"Zaki" <Klej...@web.de> wrote in message
news:2kfj3iF...@uni-berlin.de...