Reverse bit order on VHDL

Question

I\'m having trouble doing something like

b(0 to 7) <= a(7 downto 0)

when I compile it with ghdl, I have an order error. The only way I have

Answer 1

OUT_PUT(7 DOWNTO 0) <= IN_PUT(0 DOWNTO 7)

Answer 2

Suggestions?

Because your example specifies fixed length:

architecture rtl of reverser is 
    -- signal b: std_logic_vector (7 downto 0);

begin

    -- b(7) <= a(0);
    -- b(6) <= a(1);
    -- b(5) <= a(2);
    -- b(4) <= a(3);
    -- b(3) <= a(4);
    -- b(2) <= a(5);
    -- b(1) <= a(6);
    -- b(0) <= a(7);

    -- y <= b when rev = '1' else a;

    y <= a(0)&a(1)&a(2)&a(3)&a(4)&a(5)&a(6)&a(7) when rev = '1' else a;

end rtl;

The theory being that this should be less overhead than a function call or loop statement.

Answer 3

Really inverting:

for i in 0 to intermediate_data'left loop

  inverted_vector(i) <= intermediate_data(intermediate_data'left - i);

end loop;

Answer 4

There are several solutions for this problem. One possibility is the following:

gen: for i in 0 to 7 generate
  y(i) <= a(i) when rev='0' else a(7-i);
end generate;

Answer 5

That's not allowed - VHDL is so strongly typed that if you want to reverse bit orders, you have to do it explicitly.

The standard solution is to use a function (I didn't write this - Jonathan Bromley did):

function reverse_any_vector (a: in std_logic_vector)
return std_logic_vector is
  variable result: std_logic_vector(a'RANGE);
  alias aa: std_logic_vector(a'REVERSE_RANGE) is a;
begin
  for i in aa'RANGE loop
    result(i) := aa(i);
  end loop;
  return result;
end; -- function reverse_any_vector

Answer 6

The question asks how to deal specifically with b(0 to 7) <= a(7 down 0). I don't know of the reasons, but sometimes this assignment works for me (assigns things from left to right regardless of the slicing) and sometimes this assignment throws compiler errors (mismatched slicing or something).

Fortunately you do not need to use a function to handle mismatched slicing. If you are getting compiler errors for this specific problem you can use a generate loop to assign a to b.

for i in a'range generate
   b(i) <= a(i)  
   --when i is 0, you assign a's right-most bit to b's left-most bit
end generate;

It does basically the same unrolled assignment as in your example, just tight and scale-able.

I've also used this pattern when I have mismatched slicing on the right side of the assignment. For example:

signal a : std_logic_vector(0 to 7);
signal b : std_logic_vector(7 downto 0);
signal c : std_logic_vector(0 to 7);

...

for i in a'range generate
   c(i) <= a(i) xor b(i);
end generate;

Which is equivalent to:

c(0) <= a(0) xor b(0);
c(1) <= a(1) xor b(1);
c(2) <= a(2) xor b(2);
c(3) <= a(3) xor b(3);
c(4) <= a(4) xor b(4);
c(5) <= a(5) xor b(5);
c(6) <= a(6) xor b(6);
c(7) <= a(7) xor b(7);