Re: Array Indexing possible in VHDL?

Altera_Forum · ‎08-17-2011

Hi everyone, i have an enquiry regarding VHDL. let's say i have a 128 bits key which is split into 4 parts:

--divide the key into 4 parts    
        k0<=key(127 downto 96);
        k1<=key(95 downto 64);
        k2<=key(63 downto 32);
        k3<=key(31 downto 0);

is it possible to do something like in c language in VHDL:

k[0]=k0;

k[1]=k1;

k[2]=k2;

k[3]=k3;

where k[] is a 4 element array...i am aware that the bus width is uneven issue but is there anyway to implement this in VHDL? if not any ideas how i can implement this in VHDL? thanx in advance!

Altera_Forum · ‎08-17-2011

You can create an indexed type

type key_t is array (0 to 3) of std_logic_vector(31 downto 0);

signal k : key_t;

and then in your code, you can take advantage of loops too

process(key)

begin

for i in 0 to 3 loop

k(i) <= key(32*(i+1)-1 downto 32*i);

end loop;

end process;

Cheers,

Dave

Altera_Forum · ‎08-17-2011

hi dave, thank you so much for the informative reply! actually i'm trying to implement XTEA:http://en.wikipedia.org/wiki/xtea

The following is the VHDL code i'm working on:


Library IEEE;
USE IEEE.std_logic_1164.all;
use ieee.numeric_std.all;
entity XTEA is    
Port     (
    clk:in std_logic;
    input: in unsigned(63 downto 0);
    key : in unsigned(127 downto 0);
    Encoder_output,Decoder_output: out unsigned(63 downto 0)
        );
end entity;
architecture rtl of XTEA is
signal k0,k1,k2,k3: unsigned(31 downto 0);
signal     n:integer:=0;
signal key_array: unsigned (3 downto 0);
type key_t is array (0 to 3) of unsigned(31 downto 0);    
signal k : key_t;
begin
process(clk,input)
variable z,z_up,z_mid,z_down,z_total,y,y_up,y_mid,y_down,y_total: unsigned (31 downto 0);
variable sum: unsigned (31 downto 0) := x"00000000";
variable delta: unsigned (31 downto 0) := x"9e3779b9";
variable test,test1,test2,test3 : unsigned (31 downto 0);
variable natvar,natvar1,natvar2,natvar3 : integer;
begin
        --divide the key into 4 parts    
        k(3)<=key(127 downto 96);
        k(2)<=key(95 downto 64);
        k(1)<=key(63 downto 32);
        k(0)<=key(31 downto 0);
    
                    
if(rising_edge(clk)) then
    if (n=0) then
        --divide input into y and z
        y:= input(63 downto 32);
        z:= input (31 downto 0);
        --sum:=x"00000000";         -- try take this out later
        elsif (n=32) then
        --for decode:delta is delta*2^5 for 32=(2^5) round 
        sum:=delta(26 downto 0)&"00000";
        end if;
--encode        
if (n <32) then
test:=sum and x"00000003";
test1:="00000000000" & sum(31 downto 11);    --shift by 11 bits to the right
natvar:=to_integer(test);
natvar1:=to_integer(test1);
z_total:=(((z(27 downto 0)&"0000") XOR ("00000"&z(31 downto 5)))+z) XOR(   sum + k( natvar  downto natvar )   );
y:=y+z_total;
            sum:=sum+delta;
y_total:=(((y(27 downto 0)&"0000") XOR ("00000"&y(31 downto 5)))+y) XOR (sum+ k(natvar1 downto natvar1  ) );
z:=z+y_total;
            
            Encoder_output<=y&z;    
            
--decode            
elsif(n<64) then
        if(n=32) then
        sum:=delta*32;
        end if; 
test2:=sum and x"00000003";
test3:="00000000000" & sum(31 downto 11);    --shift by 11 bits to the right
natvar2:=to_integer(test2);
natvar3:=to_integer(test3);    
    
test1:=sum and x"00000003";
natvar1:=to_integer(test1);
y_total:=(((y(27 downto 0)&"0000") XOR ("00000"&y(31 downto 5)))+y) XOR (sum+ k(natvar3 downto natvar3));
z:=z-y_total;
            sum:=sum-delta;
z_total:=(((z(27 downto 0)&"0000") XOR ("00000"&z(31 downto 5)))+z) XOR(sum + k(natvar2 downto natvar2));
y:=y+z_total;
            Decoder_output<=y&z;
        end if;
            n<=n+1; --counter
    end if;
    
end process;
end rtl;

Using the method you suggested, i encountered some problems:

Error (10327): VHDL error at XTEA.vhd(59): can't determine definition of operator ""+"" -- found 0 possible definitions

The problem lines are bolded...apparently my array (named k(0 to 3)) has the type of key_t but other operations are done in unsigned type. i tried converting k() to unsigned using "unsigned(k())" but to no avail. is there anyway to go about this? Than a lot for the great help again! have a nice day ahead!:)

Altera_Forum · ‎08-17-2011

The cause of the error is different from what you assume. This expression can't work, because it results in a subarray rather than a single array element k( natvar downto natvar ). Do you mean k( natvar)?

In addition, an expression length error is involved with sum:=delta*32;

Altera_Forum · ‎08-17-2011

On another note - you're indexing into K (that has 4 elements) with a 32 bit integer. What happens when natvar2 or 3 are greater than 3?

Altera_Forum · ‎08-18-2011

Thank you FvM for pointing out my mistakes. yes i meant k(natvar)...also the delta*32 is not possible so i replaced it with delta(26 downto 5) & "00000". and the code works! thank you again for your help! on a side note, i am experiencing some weird simulation waveform where by after 64 counts, the cycle repeats itself again which is not suppose to happen (circled in red)

http://img811.imageshack.us/img811/8240/problemwy.jpg (http://imageshack.us/photo/my-images/811/problemwy.jpg/)

...any idea what happen?

Thanx tricky for bringing that natvar issue up...i'll try and figure it out...sorry for my mistakes...i'm a total newbie trying to learn haha

Altera_Forum · ‎08-18-2011

Apparently the integer counter is optimized to 0 to 63 range and overflows. If you want to stop operation after one pass, you should do explicitely.