Hardware Squaring in VHDL

[Carl]

Hi

I have got an input value (31 downto 0) and I want to square this value. In
hardware this is easy I just have to
expand this value to an array which is (63 downto 0) bits long and I have to
add between every 2 coeffecients a zero. My problem is now how can I
implement this in VHDL effecient.

sure i would be possible that I say: result(1) <= inp(0);
result(3) <= inp(1);

and so on. This would be a lot of work to write this stuff. Is there another
possibilty how I can address this problem?

Thanks a lot for yout help

C

 

[rickman]

Carl wrote:

Hi

I have got an input value (31 downto 0) and I want to square this value. In
hardware this is easy I just have to
expand this value to an array which is (63 downto 0) bits long and I have to
add between every 2 coeffecients a zero. My problem is now how can I
implement this in VHDL effecient.

sure i would be possible that I say: result(1) <= inp(0);
result(3) <= inp(1);

and so on. This would be a lot of work to write this stuff. Is there another
possibilty how I can address this problem?

As Pieter posted, a loop is ideal for your algorithm. But are you sure
this works? Here is an example that does not go the way you
described...

"00011" x "00011" = "0000001001" vs. "0000001010"
"00010" x "00010" = "0000000100" vs. "0000001000"
"00001" x "00001" = "0000000001" vs. "0000000010"

I guess this is not exactly what you are doing.

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design URL http://www.arius.com
4 King Ave 301-682-7772 Voice
Frederick, MD 21701-3110 301-682-7666 FAX

 

[Pieter Hulshoff]

Carl wrote:

I have got an input value (31 downto 0) and I want to square this value.
In hardware this is easy I just have to
expand this value to an array which is (63 downto 0) bits long and I have
to add between every 2 coeffecients a zero. My problem is now how can I
implement this in VHDL effecient.

sure i would be possible that I say: result(1) <= inp(0);
result(3) <= inp(1);

and so on. This would be a lot of work to write this stuff. Is there
another possibilty how I can address this problem?

PROCESS
BEGIN

WAIT UNTIL clk = '1';

-- Default all bits to '0'
--
result <= (OTHERS => '0');

-- Overwrite odd bits
--
FOR i IN 0 TO 31 LOOP
result(i*2+1) <= inp(i);
END LOOP;

-- Reset condition: set result to 0
--
IF reset = '1' THEN
result <= (OTHERS => '0');
END IF;

END PROCESS;

Regards,

Pieter Hulshoff

 

"Carl" <tocotronic9@gmx.at> writes:

I have got an input value (31 downto 0) and I want to square this value. In
hardware this is easy I just have to
expand this value to an array which is (63 downto 0) bits long and I have to
add between every 2 coeffecients a zero. My problem is now how can I
implement this in VHDL effecient.

Perhaps I have misunderstood your algorithm: it would seem that this only very
roughly approximates x^2 since you do not compute any of the partial products
"between" the bits. Consider for example the number 7 in binary: 111. Your
method turns it into 42 = 101010, rather than 49 = 110001. Other numbers
produce even larger errors, especially powers of two which are off by a factor
of 2.

Here a table of the first few:

N Computed Result Correct result
7 => 101010 = 42 (49)
8 => 10000000 = 128 (64)
9 => 10000010 = 130 (81)
10 => 10001000 = 136 (100)
11 => 10001010 = 138 (121)
12 => 10100000 = 160 (144)
13 => 10100010 = 162 (169)
14 => 10101000 = 168 (196)
15 => 10101010 = 170 (225)
16 => 1000000000 = 512 (256)
17 => 1000000010 = 514 (289)
18 => 1000001000 = 520 (324)
19 => 1000001010 = 522 (361)
20 => 1000100000 = 544 (400)