Error 10327: Can't determine definition of operator ""+""

Hello,

addtion isn't defined for std_logic_vector. You can use explicite numeric types SIGNED or UNSIGNED. Alternatively you can use a library that treats all std_logic_vector as SIGNED or UNSIGNED type: IEEE.STD_LOGIC_SIGNED respectively IEEE.STD_LOGIC_UNSIGNED.

Regards,

Frank

B.T.W: I see this message every time, when I start writing a new component. It's cause VHDL is so strictly typified.

To add to what Frank said, if you are doing arithmetic, it is quicker (simulation-wise) to use INTEGER types, rather than vectors of bits. Other things that you get from this are errors if you use restricted ranges of integer and it goes outside of the range. This does mean that if you want modulo-n rollover in a counter or similar, you have to code it explicitly:

if (counter + 1 > t_counter'high) then

counter := 0;

else

counter := counter +1;

end if;

If t_counter'high is a power of 2, then the "counter+1>" comparison will usually optimise out in synthesis.

Cheers,

Martin

If you use integer types, declare each object with a specific range if possible. Integer types require 32-bits in Quartus II. If you simply add two integer objects, you'll get a 32-bit adder initially, which Quartus II may or may not be able to optimize fully. For sure, it won't always find the solution you can get by simply defining the appropriate dynamic range for each object. It will certainly improve the compile-time/memory use in Quartus II!

I generally use signed/unsigned from ieee.numeric_std, using integer types only for very limited purposes, such as array indices (it's a pain to convert from signed/unsigned to integer all the time :)

Does Quartus not optimise based on reachability analysis? If I code a counter that wraps around at 255, does it not infer an 8-bit counter? I know Max Plus II never did, but surely Quartus can do this...

Also, why do you say use signed/unsigned all the time? Most of the time, you only *need* vectors for the top level port pins. Just pass integers around your design hierarchy the rest of the time. Typing to_integer() isn't that much grief anyway is it? :-)? All IMHO of course!

Cheers,

Martin

Though I expect most synthesis tools to do a decent job optimizing my arithmetic, I don't expect miracles. I almost always specify a precise dynamic range if I use integer types. If you use INTEGER all over the place, I see two problems:

First, most tools extract single hierarchies, which are later merged and optimized. Initially, the tool extracts 32-bits for every INTEGER object. If you know something only stores 8-bits of data, that's a lot of wasted bits that consume runtime and memory.

Second, if you use Incremental Compilation flow, you will find the all-INTEGER approach unworkable. If a partition boundary contains a 32-bit INTEGER port, then it will take 32-bits in the final netlist - the tool can't use the dynamic range of the input to optimize the arithmetic.

I prefer unsigned/signed because they force me to think about the size of my data path, and it's easy to access specific bits in the value.

PS: I think my_ram(addr) looks much better than my_ram(to_integer(addr)) :)

Hello,

I'm not sure about all implications of using INTEGER type for bit vectors that represent numbers. It should be stated clearly, that it's possible, including entity ports, also at the top level. Compile time behaviour could be different, as HDL Guru mentioned. But if only an insignificant effect, this shouldn't determine the design methodology.

The other thing is functionality when designing with parameterized reusable modules. Numerical parameters then always need associated generics defining ranges exept for a few cases, where the range is fixed by design function, e. g. a MAC address having always 32 bit. There is no basic difference with SIGNED/UNSIGNED or INTEGER. With INTEGER, one could try to have parameters, that are either be SIGNED or UNSIGNED, depending on range generics.

Cause INTEGER type is only an implicite bit vector, without explicitely defining the bit positions, it could be, that bit operations I'm used to with SIGNED/UNSIGNED could be more complicated or even require type conversion. But I didn't yet try, cause my usage of INTEGER is mainly limited to indexes.

It's been said (I think, in another discussion), that INTEGER usage is shown in most VHDL text books. That is correct so far, as the type and the RANGE syntax are shown, but equivalence to bit vectors isn't discussed necessarily. I was surprized at first look, when I found Uwe Meyer-Baese using INTEGER in many examples in digital signal processing with fpga. Many people with some knowledge of VHDL think, INTEGER can't be synthesized at all.

As a last remark, VHDL always need type conversion in some places, cause it is so typified. At least when instantiating Altera LPM or IP cores, they are necessary. Some logic functions can't be inferred from HDL, e. g. dividers or dual port RAM with different port widths.

Regards,

Frank

--- Quote Start ---

Hello,

I'm not sure about all implications of using INTEGER type for bit vectors that represent numbers. It should be stated clearly, that it's possible, including entity ports, also at the top level.

--- Quote End ---

It should also be noted that if you *do* use integers at the top level you'll need an extra wrapper around any post-synth or post-fit netlists you want to simulate, as they'll end up as logic vectors in the end.

--- Quote Start ---

Cause INTEGER type is only an implicite bit vector, without explicitely defining the bit positions, it could be, that bit operations I'm used to with SIGNED/UNSIGNED could be more complicated or even require type conversion. But I didn't yet try, cause my usage of INTEGER is mainly limited to indexes.

--- Quote End ---

Also agreed - if you want to do bitwise things, it makes sense to use a vector representation from the start.

--- Quote Start ---

It's been said (I think, in another discussion), that INTEGER usage is shown in most VHDL text books. That is correct so far, as the type and the RANGE syntax are shown, but equivalence to bit vectors isn't discussed necessarily. I was surprized at first look, when I found Uwe Meyer-Baese using INTEGER in many examples in digital signal processing with fpga. Many people with some knowledge of VHDL think, INTEGER can't be synthesized at all.

--- Quote End ---

Not helped by the amount of example code which start with

use ieee.std_logic_unsigned.all;

and the use of slv's as unsigned vectors :-)

<snip>

Cheers,

Martin

I wite a code for multiplier and I have a problem with this erro " can't determine definition of operator ""+"". I have used almost library as you see and tried many ways that I know but it's still bad. Help me plz. can you help me re-write this code or show me a good way to solve. Thanks alot. I use Quatus II , Web edition

library IEEE;

use IEEE.STD_LOGIC_1164.ALL;

use IEEE.STD_LOGIC_ARITH.ALL;

use ieee.numeric_std.all;

use ieee.std_logic_signed.all;

--------------------------

entity mul is

port(a: in bit_vector(3 downto 0);

b: in bit_vector(3 downto 0);

kq: out bit_vector( 7 downto 0));

end mul;

--------------------------

architecture mul of mul is

type arr is array( 0 to 3) of bit_vector(7 downto 0);

begin

process(a,b)

variable x: arr;

variable t: bit_vector(7 downto 0);

begin

for j in 0 to 3 loop

if b(j)='1' then

x(j):=("0000"& a);

x(j):=(x(j) sll j);

else

x(j):=(others =>'0');

end if;

t:=(t+x(j));

end loop;

kq<=t;

end process;

end mul;

I have the same problem.. Error (10327): can't determine definition of operator ""+"" -- found 0 possible definitions

I write this program:

LIBRARY ieee;

USE ieee.std_logic_1164.ALL;

use ieee.numeric_std.all;

use ieee.std_logic_signed.all;

ENTITY filter_iir IS

GENERIC ( L: INTEGER :=3;

W1: INTEGER :=16;

W2: INTEGER :=32;

W3: INTEGER :=33;

W4: INTEGER :=14

);

PORT ( --x_in: IN signed (15 downto 0);; --INPUT

--y_out: OUT BITS14; --RESULT

x_in : in BITS16;

y_out : out signed (31 downto 0);

clk : IN STD_LOGIC);

END filter_iir;

ARCHITECTURE myiir of FILTER_IIR IS

SIGNAL x : ARRAY16TO3;

signal x_num : ARRAY32TO3;

constant b0 : signed(W1-1 downto 0) :=x"0119";

constant b1 : signed(W1-1 downto 0) :=x"0232";

constant b2 : signed(W1-1 downto 0) :=x"0119";

constant a1 : signed(W1-1 downto 0) :=x"FFB5";

constant a2 : signed(W1-1 downto 0) :=x"00B0";

shared variable acc : signed (W2-1 downto 0):= (others => '0');

shared variable y : signed (W2-1 downto 0):= (others => '0'); -- 28 bit

shared variable y_den : signed (W2-1 downto 0):= (others => '0'); -- 27 bit (11+16)

BEGIN

sop: PROCESS

BEGIN

wait until (clk='1') ;

x(0)<=x_in;

x_num(0)<=b0*x(0); -- b0x(k)

debugx<=x_num(0);

x_num(1)<=b1*x(1); -- b1x(k-1)

debugx1<=x_num(1);

x_num(2)<=b2*x(2); -- b2x(k-2)

debugx2<=x_num(2);

for i in 0 to L-2 loop

x(i+1)<=x(i);

end loop;

for j in 0 to L-3 loop

y(j+1):=y(j);

end loop;

y(0):=(x_num(0)+x_num(1)+x_num(2)+y_den(0)+y_den(1));

acc:=y(0);

y_den(0):=a1*y(0)(15 downto 0); -- a1y(k-1)

y_den(1):=a2*y(1)(15 downto 0); -- a2y(k-2)

end process;

y_out <= y(0)(31 downto 0);

end myiir;

when it sum these parts Y(0):=(x_num(0)+x_num(1)+x_num(2)+y_den(0)+y_den(1)) the compiler give me an error.. why?

i use ieee.std_numeric.all library and in this library there is the definition of "+"....?!?!?:confused::confused::confused:

Boh.. i search the error but i don't understad..

so initially i think that this function want the adder at the same lengths.. but i read the source code.. and seem that it isn't a problem...

x_num is array of signed with length 16 and y_den is array of signed with length 32.. and y(0) is 32bit signed register..

so i don't know where is the problem?

can somebody help me..?