Trying to design a simple MAC Unit in VHDL
library ieee;
use ieee.std_logic_1164.all;
use ieee.numeric_std.all;
entity MAC is
port (
DIN: IN signed(15 downto 0);
WEIGHT: IN signed(15 downto 0);
CLK: IN std_logic;
RST: IN std_logic;
RESULT: OUT signed(15 downto 0);
);
end entity MAC;
architecture base of MAC is
begin
process(CLK)
variable acc_value : signed(15 downto 0);
begin
IF rising_edge(CLK) then
IF RST='1' THEN
acc_value := (OTHERS => '0'); -- reset accumulated value to 0
ELSE
acc_value := (acc_value + WEIGHT* DIN);
RESULT <= acc_value;
END IF;
END IF;
END process;
end base;
Seems like there is a problem with acc_value := (acc_value + WEIGHT* DIN); .
There is an error in the simulation "ERROR: Array sizes do not match, left array has 16 elements, right array has 32 elements" with confuses me, because for me both arrays have 16 bits (or?)
It's because of overflowing, you need to take care of that.
acc_value := (acc_value + WEIGHT* DIN);
In the right hand side, WEIGHT * DIN is 16 bit * 16 bit = 32 bits. Adding the 16-bit acc_value to that (acc_value + WEIGHT*DIN) gives you a 32-bit value, which cannot be assigned into the 16-bit variable signed acc_value.
Related
So I've been working on some homework for my VHDL course and I can't seem to understand this problem.
The point here is to create the adder/subtractor of an ALU that works both on 2's complement and unsigned 32-bit buses, which is why I have a condition called sub_mode ( A - B = A + !B + 1 ) which will also be the carry-in when activated.
The rest of the different inputs and outputs are pretty self-explanatory.
My problem is with the testbenching of such component where, even though carry_temp and r_temp have been initialized in declaration section of the architecture, end up showing up undefined. I have guessed that it is due to the for loop within the process screwing everything up. Would that be an accurate guess? And if yes, is it possible to proceed to add two bit buses together without having to fully create an n-bit adder made from n 1-bit adder components?
library ieee;
use ieee.std_logic_1164.all;
use ieee.numeric_std.all;
entity add_sub is
port(
a : in std_logic_vector(31 downto 0);
b : in std_logic_vector(31 downto 0);
sub_mode : in std_logic;
carry : out std_logic;
zero : out std_logic;
r : out std_logic_vector(31 downto 0)
);
end add_sub;
architecture synth of add_sub is
signal cond_inv : std_logic_vector(31 downto 0);
signal carry_temp : std_logic_vector(32 downto 0) := (others => '0');
signal r_temp : std_logic_vector(31 downto 0) := (others => '0');
begin
behave : process(a,b,sub_mode)
begin
if sub_mode = '1' then
cond_inv <= b xor x"ffffffff";
else
cond_inv <= b;
end if;
carry_temp(0) <= sub_mode;
for i in 0 to 31 loop
r_temp(i) <= a(i) xor cond_inv(i) xor carry_temp(i);
carry_temp(i+1) <=
(a(i) and cond_inv(i)) or
(a(i) and carry_temp(i)) or
(cond_inv(i)and carry_temp(i));
end loop;
if r_temp = x"00000000" then
zero <= '1';
else
zero <= '0';
end if;
r <= r_temp;
carry <= carry_temp(32);
end process behave;
end synth;
For my project I need to reduce a noise of an ADC output and implemented a simple moving average filter in VHDL.
Although it works in simulation (see the picture):
it has some strange behavior if I display it on the chipscope when the system is running in FPGA (see the picture):
The VHDL code I use for the moving average is as follows:
library ieee;
use ieee.std_logic_1164.all;
use ieee.math_real.all;
use ieee.numeric_std.all;
entity moving_avg is
generic(
SAMPLES_COUNT : integer := 32
);
port (
clk_i : in std_logic;
rst_n_i : in std_logic;
sample_i : in std_logic_vector(11 downto 0);
avg_o : out std_logic_vector(11 downto 0)
);
end;
architecture rtl of moving_avg is
type sample_buff_t is array (1 to SAMPLES_COUNT) of std_logic_vector(11 downto 0);
signal sample_buffer : sample_buff_t;
signal sum : std_logic_vector(31 downto 0);
constant wid_shift : integer := integer(ceil(log2(real(SAMPLES_COUNT))));
signal avg_interm_s : std_logic_vector(31 downto 0);
begin
process (clk_i, rst_n_i) begin
if rst_n_i='1' then
sample_buffer <= (others => sample_i);
sum <= std_logic_vector(unsigned(resize(unsigned(sample_i), sum'length)) sll wid_shift) ;
elsif rising_edge(clk_i) then
sample_buffer <= sample_i & sample_buffer(1 to SAMPLES_COUNT-1);
sum <= std_logic_vector(unsigned(sum) + unsigned(sample_i) - unsigned(sample_buffer(SAMPLES_COUNT)));
end if;
end process;
avg_interm_s <= std_logic_vector((unsigned(sum) srl wid_shift));
avg_o <= avg_interm_s(11 downto 0);
end;
I use Xilinx Vivado tool 2015.2 running on Ubuntu 14.04 x64.
Could you please help me to identify the problem, such
that results in simulation correspond to results after synthesis?
I have a 32 bit barrel shifter using behavior architecture. Now I need to convert it to an n-bit shifter. The problem that I'm facing is that there is some kind of restriction to the for loop that I have to put a constant as sentinel value.
Following is my Code
library IEEE;
use IEEE.std_logic_1164.all;
Entity bshift is -- barrel shifter
port (left : in std_logic; -- '1' for left, '0' for right
logical : in std_logic; -- '1' for logical, '0' for arithmetic
shift : in std_logic_vector(4 downto 0); -- shift count
input : in std_logic_vector (31 downto 0);
output : out std_logic_vector (31 downto 0) );
end entity bshift;
architecture behavior of bshift is
function to_integer(sig : std_logic_vector) return integer is
variable num : integer := 0; -- descending sig as integer
begin
for i in sig'range loop
if sig(i)='1' then
num := num*2+1;
else
num := num*2;
end if;
end loop; -- i
return num;
end function to_integer;
begin -- behavior
shft32: process(left, logical, input, shift)
variable shft : integer;
variable out_right_arithmetic : std_logic_vector(31 downto 0);
variable out_right_logical : std_logic_vector(31 downto 0);
variable out_left_logical : std_logic_vector(31 downto 0);
begin
shft := to_integer(shift);
if logical = '0' then
out_right_arithmetic := (31 downto 32-shft => input(31)) &
input(31 downto shft);
output <= out_right_arithmetic after 250 ps;
else
if left = '1' then
out_left_logical := input(31-shft downto 0) &
(shft-1 downto 0 => '0');
output <= out_left_logical after 250 ps;
else
out_right_logical := (31 downto 32-shft => '0') &
input(31 downto shft);
output <= out_right_logical after 250 ps;
end if;
end if;
end process shft32;
end architecture behavior; -- of bshift
any help will be appreciated
Your code is not a barrel shifter implementation, because a barrel shift is a mux-tree.
If you have a 32 bit BarrelShifter module, you will need a 5 bit Shift input, wherein every bit position i enables a 2^i shift operation.
So for example shift = 5d -> 00101b enables a mux in stage 1 to shift for 1 bit and a mux in stage 3 to shift 4 bits. All other mux stages are set to pass through (shift(i) = 0).
I also would not advice to mix up basic shifting with shift modes (arithmetic, logic, rotate) and directions (left, right).
arithmetic and logic is only different in the shift-in value
shift right can be done by a conversion => shiftright = reverse(shiftleft(reverse(input), n)
An open source implementation can be found here:
https://github.com/VLSI-EDA/PoC/blob/master/src/arith/arith_shifter_barrel.vhdl
I try to write programm on vhdl in ise 14.4 for crc16 calculation but dont understand why get "parse error, unexpected FOR" in it. Tried to put it into process but it dont works too.
library IEEE;
use IEEE.STD_LOGIC_1164.ALL;
entity crc16 is port(
clk : in STD_LOGIC:='0');
end crc16;
architecture Behavioral of crc16 is
signal data:std_logic_vector(15 downto 0):="1010101010101010";
signal ext_data:std_logic_vector(31 downto 0);
signal crc16_original:std_logic_vector(15 downto 0):="1100000000000010";
signal crc16:std_logic_vector(15 downto 0);
signal position:std_logic_vector(5 downto 0);
signal crc_out:std_logic_vector(14 downto 0);
signal i:std_logic_vector(5 downto 0);
begin
for i in 1 to 15 loop
ext_data(i+16)<=data(i);
end loop;
for i in 1 to 15 loop
ext_data(i)<='0';
end loop;
while ext_data > "111111111111111" loop
for i in 0 to 31 loop
if ext_data(i)="1" position=i;
end loop;
crc16<= crc16_original srl 31-position;
ext_data<=ext_data xor crc16;
end loop;
for i in 0 to 14 loop
crc_out(i)<=ext_data(i);
end loop;
end Behavioral;
There are several issues to point out:
The for-loop must be in a process, so that is likely to cause the “parse error, unexpected FOR” that you see.
The relation compare with > may give unexpected result for std_logic_vector, so you may take a look at the numeric_std package for casting as for example unsigned(std_logic_vector) before comparison is made.
Compare ext_data(i) = "1" is illegal, since "1" is taken as std_logic_vector, where as ext_data(i) is std_logic; instead ext_data(i) = '1' will compile.
Illegal construction around if ext_data(i) = "1" position=i;, since no then etc.
There is an signal with identifier i, which i is also used as loop variable, with the result that position <= i is taken as an integer assign to std_logic_vector; use different names for signals and loop variables.
Assign to signal is not position = i but position <= i, like elsewhere.
Expression 31-position mixes integer and std_logic_vector, which can't be done with the selected packages. Use casting with unsigned.
The ext_data<=ext_data xor crc16 uses different size arguments, since ext_data is 32 bits and crc16 is 16 bits; this does probably not yield the expected result.
srl is not defined for std_logic_vector (VHDL-2002), so consider casting with unsigned for well-defined behavior.
Assuming that that your code is "sandbox" code, since it has no outputs.
Based on the above, you may consider doing some initial experiments with smaller designs, in order to get familiar with the different VHDL constructions, and learn how this simulates and maps to hardware; remember VHDL is a "Hardware Description Language" and not a programming language.
Below is some code that compiles in ModelSim, but is unlikely to give the expected result:
library ieee;
use ieee.std_logic_1164.all;
entity crc16 is port(
clk : in std_logic := '0');
end crc16;
library ieee;
use ieee.numeric_std.all;
architecture Behavioral of crc16 is
signal data : std_logic_vector(15 downto 0) := "1010101010101010";
signal ext_data : std_logic_vector(31 downto 0);
signal crc16_original : std_logic_vector(15 downto 0) := "1100000000000010";
signal crc16 : std_logic_vector(15 downto 0);
signal position : std_logic_vector(5 downto 0);
signal crc_out : std_logic_vector(14 downto 0);
signal i_sig : std_logic_vector(5 downto 0);
begin
process (clk) is
begin
if rising_edge(clk) then
for i in 1 to 15 loop
ext_data(i+16) <= data(i);
end loop;
for i in 1 to 15 loop
ext_data(i) <= '0';
end loop;
while ext_data > "111111111111111" loop
for i in 0 to 31 loop
if ext_data(i) = '1' then
position <= i_sig; -- TBD[Probably not right code, but compiles]
end if;
end loop;
crc16 <= std_logic_vector(unsigned(crc16_original) srl (31 - to_integer(unsigned(position))));
ext_data <= ext_data xor crc16;
end loop;
for i in 0 to 14 loop
crc_out(i) <= ext_data(i);
end loop;
end if;
end process;
end Behavioral;
I have an input signal from ADC convertor that is 8 bits (std_logic_vector(7 downto 0)). I have to convert them to a 16 bits signal (std_logic_vector(15 downto 0)) for 16 bits signal processing to the 16 bits system.
If the 8 bit value is interpreted as signed (2's complement), then the general and standard VHDL conversion method is to use the IEEE numeric_std library:
library ieee;
use ieee.numeric_std.all;
architecture sim of tb is
signal slv_8 : std_logic_vector( 8 - 1 downto 0);
signal slv_16 : std_logic_vector(16 - 1 downto 0);
begin
slv_16 <= std_logic_vector(resize(signed(slv_8), slv_16'length));
end architecture;
So first the std_logic_vector is converted to a signed value, then the resize is applied, which will sign extend the signed value, and the result is finally converted back to std_logic_vector.
The conversion is rather lengthy, but has the advantage that it is general and works even if the target length is changed later on.
The attribute 'length simply returns the length of the slv_16 std_logic_vector, thus 16.
For unsigned representation instead of signed, it can be done using unsigned instead of signed, thus with this code:
slv_16 <= std_logic_vector(resize(unsigned(slv_8), slv_16'length));
architecture RTL of test is
signal s8: std_logic_vector(7 downto 0);
signal s16: std_logic_vector(15 downto 0);
begin
s16 <= X"00" & s8;
end;
This handles the conversion without having to edit the widths of the zeroes if either std_logic_vector changes:
architecture RTL of test is
signal s8: std_logic_vector(7 downto 0);
signal s16: std_logic_vector(15 downto 0) := (others => '0');
begin
s16(s8'range) <= s8;
end;
For completeness, yet another way which is occasionally useful:
-- Clear all the slv_16 bits first and then copy in the bits you need.
process (slv_8)
begin
slv_16 <= (others => '0');
slv_16(7 downto 0) <= slv_8;
end process;
I've not had to do this for vectors that I can recall, but I have had need of this under more complex circumstances: copying just a few relevant signals into a bigger, more complex, record was one time.
With the newly released VHDL-2019 standard you can do
larger_vec <= extend(shorter_vec);
where extend is a function defined as follows
function extend(vec : std_logic_vector) return target_vec of std_logic_vector is
variable result : std_logic_vector(target_vec'length - 1 downto 0) := (others => '0');
begin
assert vec'length <= target_vec'length report "Cannot extend to shorter vector";
result(vec'length - 1 downto 0) := vec;
return result;
end function;
Tool support is still a bit limited but at least one simulator supports this (Riviera-PRO).