What's wrong in this VHDL subtraction?

hi,

I have this code (pseudo code, all inputs and signal should be unsigned
integers):

input first_input, second_input: std_logic_vector (4 downto 0);

Library IEEE;
use IEEE.std_logic_1164.all;
use IEEE.std_logic_arith.all;
use IEEE.std_logic_unsigned.all;

signal total : std_logic_vector (5 downto 0);
get_total : process (my_inputs)
begin -- process
total <= '0' & (("11110" - first_input) + ("11110" -
second_input));
end process;

that is, total = (30 - first input) + (30 -second_input).

When first_input and second_input are both zero total comes equal to 28
decimal, not 60,
with modelsim. What am i doing wrong? Bad libraries, bad signal types,
what? I have experience
with Verilog, but not vhdl....

Thanks,
Mon

 

[David R Brooks]

mweb@inwind.it wrote:

hi,

I have this code (pseudo code, all inputs and signal should be unsigned
integers):

input first_input, second_input: std_logic_vector (4 downto 0);

Library IEEE;
use IEEE.std_logic_1164.all;
use IEEE.std_logic_arith.all;
use IEEE.std_logic_unsigned.all;

signal total : std_logic_vector (5 downto 0);
get_total : process (my_inputs)
begin -- process
total <= '0' & (("11110" - first_input) + ("11110" -
second_input));
end process;

that is, total = (30 - first input) + (30 -second_input).

When first_input and second_input are both zero total comes equal to 28
decimal, not 60,
with modelsim. What am i doing wrong? Bad libraries, bad signal types,
what? I have experience
with Verilog, but not vhdl....

Thanks,
Mon

Consider, 28 = 60 - 32

You have lost the top bit. Reason: your intermediate results are still
of type std_logic_vector(4 downto 0), ie 5 bits. So 60 (111100 binary)
gets truncated to fit.

Try this:
input first_input, second_input: std_logic_vector (4 downto 0);

Library IEEE;
use IEEE.std_logic_1164.all;
use IEEE.std_logic_arith.all;
use IEEE.std_logic_unsigned.all;

signal total : std_logic_vector (5 downto 0);
get_total : process (my_inputs)
variable a, b : std_logic_vector (5 downto 0);
begin -- process
a := '0' & first_input; -- prepend zero to each
b := '0' & second_input;
-- again, prepending zero to the consts.
total <= ("011110" - a) + ("011110" - b);
end process;

 

David,
you're right of course. Everything works now.

I had already realized that 28 = 60 -32, with all it implies, by
myself, ten minutes
after posting. I'm embarassed I missed something so elementary. Ah,
well, probably
useful as a proof to my boss (who knows I *can* code HDL circuits)
that he's stressing
me too much

Mon

 

[Eric Smith]

mweb@inwind.it writes:

Library IEEE;
use IEEE.std_logic_1164.all;
use IEEE.std_logic_arith.all;
use IEEE.std_logic_unsigned.all;

Someone else pointed out the problem with the subtraction.

I'd suggest using IEEE.numeric_std rather than std_logic_arith or
std_logic_unsigned, as neither of the latter are actually IEEE standard
packages (the library naming notwithstanding).