generic vector

[valentin tihomirov]

I'm trying to do the following:

entity LFSR is
generic (
SIZE : Integer := 8;
SEED: std_logic_vector(SIZE-1 downto 0) := (othres => '1')
);
port (

The seed should be loaded into FFs on reset. The problem is that SIZE
generic cannot be used in SEED declaration. Am I doing something wrong?

 

[MK]

I'm trying to do the following:

entity LFSR is
generic (
SIZE : Integer := 8;
SEED: std_logic_vector(SIZE-1 downto 0) := (othres => '1')
);
port (

The seed should be loaded into FFs on reset. The problem is that SIZE
generic cannot be used in SEED declaration. Am I doing something wrong?

"generic ( ... )" statement create declaration region - you cannot use
any generic declared inside until region is closed (you can use generics in
port ( ... ) region).

regards,
MK.

 

[Jonathan Bromley]

in message
hi Valentin,

I thought we had already talked about some solutions to this?

I'm trying to do the following:

entity LFSR is
generic (
SIZE : Integer := 8;
SEED: std_logic_vector(SIZE-1 downto 0) := (othres => '1')
);
port (

The seed should be loaded into FFs on reset. The problem is that SIZE
generic cannot be used in SEED declaration. Am I doing something wrong?

As MK pointed out, you can't use a generic within the same generic
declaration.

However, there is a much simpler, much nicer solution to your
problem.

entity LFSR is
generic (
SEED: std_logic_vector := "11111111"
);
port (

Now, within the design you can use SEED'LENGTH wherever you would
have used your unnecessary generic SIZE. Or even...

architecture xxx of LFSR is
constant SIZE: positive := SEED'LENGTH;
signal LFSR_Register: std_logic_vector(SEED'RANGE);
begin
...

Now you can configure each instance of LFSR with a single
generic map - the number of bits in the generic sets SIZE,
and its value sets SEED.

This leaves you with a different problem: In most cases
I guess you want to configure SIZE through a generic, but
you want the seed automatically set to (others=>'1').
This is easily done by adding a wrapper around LFSR:

entity LFSR_All_Ones_Seed is
generic (
SIZE: positive := 8
);
port (
clk: in std_logic;
...
);
end;
architecture XXX of LFSR_All_Ones_Seed is
constant SEED: std_logic_vector(SIZE-1 downto 0) := (others => '1');
component LFSR
generic (
SEED: std_logic_vector
);
port (
...
end component;
begin
-- Nothing here except an instance of the standard LFSR:
LFSR_Instance: LFSR
generic map (SEED => SEED)
port map (
clk => clk,
...
end;

Hope this helps
--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL * Verilog * SystemC * Perl * Tcl/Tk * Verification * Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, Hampshire, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail: (e-mail address removed)
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.

 

[valentin tihomirov]

"generic ( ... )" statement create declaration region - you cannot use

any generic declared inside until region is closed (you can use generics in
port ( ... ) region).

Why does VHDL prohibit doing this? Why does VHDL prohibit negative-index
loops while most general-purpose languages support both? Is it because VHDL
is not matured language and things will change soon or there are any
fundamental barriers? Many thanks.

 

[Jonathan Bromley]

Why does VHDL prohibit negative-index loops [...] ?

Can you provide an example of this restriction?
I am not aware of it. Do the following not
meet your needs?

for i in -3 to -1 loop ...

for i in 10 downto -20 loop ...

Perhaps you are concerned that std_logic_vector can
take only non-negative subscripts. This is simply
because it is defined as "array (NATURAL range <>) of...".
You are quite entitled to define an array indexed by
integer if you so wish, and such an array may have
negative subscripts. C-like languages lack this
flexibility.
--
Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL * Verilog * SystemC * Perl * Tcl/Tk * Verification * Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, Hampshire, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail: (e-mail address removed)
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.

 

[valentin tihomirov]

Here is an example of negative index I'm talking about

@CLK:
Buf := Rx & Buf(WIDTH-2 downto 1);

I'm trying to shift buffer to right. But this neat 1-line code should be
modified and becomes cumbersome because the generic WIDTH parameter can be
any integer > 0. This means if WIDTH = 1 I'm getting (-1 downto 1) loop and
when WIDTH = 2 I'm getting (0 downto 1 loop). Both these cases should be
handeled specially in VHDL (using if clause). In Pascal/C and any other
general-purpose language negative loops like

for I := WIDTH-2 downto 1 do
...

are ignored for width < 2 cases. VHDL compiler could ignore as well, but it
complains "negative index" instead.

 

[Jonathan Bromley]

Here is an example of negative index I'm talking about

@CLK:
Buf := Rx & Buf(WIDTH-2 downto 1);

THAT IS NOT A LOOP!

I'm trying to shift buffer to right. But this neat 1-line code should be
modified and becomes cumbersome because the generic WIDTH parameter can be
any integer > 0. This means if WIDTH = 1 I'm getting (-1 downto 1) loop and
when WIDTH = 2 I'm getting (0 downto 1 loop). Both these cases should be
handeled specially in VHDL (using if clause). In Pascal/C and any other
general-purpose language negative loops like

for I := WIDTH-2 downto 1 do
...

are ignored for width < 2 cases. VHDL compiler could ignore as well, but it
complains "negative index" instead.

No, no, no. VHDL also ignores procedural loops with no trips,
like that. You can't take slices of an array in C anyway!

Please now, tell me what you want to happen in the example...

I guess you are trying to shift the single bit Rx into the multi-bit
register Buf. If that's true, your code is wrong, I think.

Presumably Buf is declared
Buf: std_logic_vector(WIDTH-1 downto 0);

If this is right, and Rx is a single bit, then

Buf := Rx & Buf(WIDTH-2 downto 1)

is erroneous because the right-hand side is one bit shorter than
the left!

However, let's suppose that your Rx is in fact 2 bits. In that
case it would be ridiculous to have WIDTH < 2, because you could
never assign Rx into Buf!

I guess your shift operation should be correctly coded:

Buf := Rx & Buf(WIDTH-1 downto 1);

This will work correctly for any WIDTH >= 2.

However, if WIDTH=1 you have a problem because the shift register
is not a shift register at all, but merely a simple register.
Strictly speaking VHDL should handle this, because Buf(0 downto 1)
should be a null range, but some tools complain about it. This
is the one and only case where your argument might have some force.
If you really must do such a thing, then simply USE a loop!

This code assumes that Buf was declared with a DOWNTO range - it is
possible, but a little harder, to write code that works correctly
with both TO and DOWNTO ranges, but I don't think that is necessary.

for i in Buf'RANGE loop
if i = Buf'LEFT then
Buf(i) := Rx;
else
Buf(i) := Buf(i+1);
end if;
end loop;

I agree that this is not so pretty as the single-line slice, but
it's easy to understand, fairly efficient in simulation, and
for synthesis it is just as good as the other form.

--

Jonathan Bromley, Consultant

DOULOS - Developing Design Know-how
VHDL * Verilog * SystemC * Perl * Tcl/Tk * Verification * Project Services

Doulos Ltd. Church Hatch, 22 Market Place, Ringwood, Hampshire, BH24 1AW, UK
Tel: +44 (0)1425 471223 mail: (e-mail address removed)
Fax: +44 (0)1425 471573 Web: http://www.doulos.com

The contents of this message may contain personal views which
are not the views of Doulos Ltd., unless specifically stated.

 

[valentin tihomirov]

Please, foregive me. I should be more specific. RXix is one-bit signal and
BUF is a bit-vector.