[SOLVED] object cannot be indexed because it has natural type rather than array type

[rafimiet]

I am getting the error:
Error(13692): VHDL error at generic_multiplier.vhd(66): object cannot be indexed because it has natural type rather than array type
The code is as follows:

[/QUOTE]
library ieee;
use ieee.std_logic_1164.all;
use ieee.numeric_std.all;

library vvc_lib;
use vvc_lib.pkg_TrCoeffMatrix.all;

entity generic_multiplier is
generic (
    g_data_width : natural := 512; -- 64 bytes (32 (16bit) coefficients)
    g_coeff_width : natural := 16;
    g_bl_size       : natural := 4
    );
port (
    i_data_row     : in  std_ulogic_vector(g_data_width-1 downto 0); -- (M--> L)SB <= (|Pix_N|...|Pix_2|Pix1|) in memory
    o_data_row     : out std_ulogic_vector(2*g_bl_size*g_data_width-1 downto 0)
);
end entity generic_multiplier;

architecture rtl of generic_multiplier is
    subtype t_dim1 is signed;
    type t_dim1_vector is array (natural range <>) of t_dim1;
    type t_dim2_vector is array (natural range <>) of t_dim1_vector;
    -- type t_dim2_vector is array (0 to 3, 0 to 7) of std_ulogic_vector(7 downto 0);
    signal s_data_in : t_dim2_vector (0 to g_bl_size-1)(0 to g_bl_size-1)(g_coeff_width-1 downto 0);
    signal s_data_out : signed(2*g_bl_size*g_data_width-1 downto 0) := (others => '0');

begin
    -- Reshape input data -- No need to change upto 4x4 block
    gen00: for i in 0 to g_bl_size-1 generate
        gen01: for j in 0 to g_bl_size-1 generate
            s_data_in(i)(j) <= signed(i_data_row(g_coeff_width*(i*g_bl_size+j+1)-1 downto g_coeff_width*(i*g_bl_size+j)));
        end generate gen01;
    end generate gen00;

    -- Matrix Multiplication
    gen02:  for i in 0 to g_bl_size-1 generate
        gen03: for j in 0 to g_bl_size-1 generate
            gen04: for k in 0 to g_bl_size-1 generate
                s_data_out(2*g_coeff_width((i+j)*g_bl_size+k+1)-1 downto 2*g_coeff_width((i+j)*g_bl_size+k)) <= s_data_in(i)(k) * c_dct2_b4(k)(j);
            end generate gen04;
        end generate gen03;
    end generate gen02;

    -- Output
    o_data_row(g_bl_size*g_bl_size*g_bl_size*2*g_coeff_width-1 downto 0) <= std_ulogic_vector(s_data_out(g_bl_size*g_bl_size*g_bl_size*2*g_coeff_width-1 downto 0));
    o_data_row(s_data_out'length-1 downto g_bl_size*g_bl_size*g_bl_size*2*g_coeff_width) <= (others => '0');
end architecture rtl;
[QUOTE]

I get the error in the line:

s_data_out(2*g_coeff_width((i+j)*g_bl_size+k+1)-1 downto 2*g_coeff_width((i+j)*g_bl_size+k)) <= s_data_in(i)(k) * c_dct2_b4(k)(j);
            end generate gen04;

 

[FvM]

Sounds plausible. I presume the error is referring to the term g_coeff_width(x) which makes no sense for a natural value. Either it's a typo or some misunderstanding.

 

[Akanimo]

Hi,

The parameter g_coeff_width is a constant (generic) with a natural type. By writing g_coeff_width((i+j)*g_bl_size+k+1), you are trying to index it. That cannot be done. I believe this is the problem.