Std_logic_vector assignment with variable length

[t***@gmail.com]

Hello good people of google,

I am trying to assign a certain part of a SLV(std_logic_vector) to another shorter SLV, and forcing the remaining to be '0'.

Ex, lets assign the middle 5 bits of big_word to small word.
--

Signal big_word :std_logic_vector(10 - 1 downto 0) := (others => '0');
Signal small_word :std_logic_vector(5 - 1 downto 0) := (others => '0');

...
Begin

-- I can of course use this, but this is not very elegant for large SLVs
Big_word <= "000" & small_word & "00";

--I would like to do something more like:
Big_word <= (7 -1 downto 2 => small_word, others => '0');
--The tool I am using (vivado 2018.3) complains about small_word not being of type std_ulogic. Well the word is larger than 1 bit, so that is an issue.

--
My constraints are:
VHDL '87
Using ieee.std_logic_1164.all
Using ieee.numeric_std
Make this assignment in 1 line

I really appreciate the help
CT

[Rick C]

Post by t***@gmail.com
Hello good people of google,
I am trying to assign a certain part of a SLV(std_logic_vector) to another shorter SLV, and forcing the remaining to be '0'.
Ex, lets assign the middle 5 bits of big_word to small word.
--
Signal big_word :std_logic_vector(10 - 1 downto 0) := (others => '0');
Signal small_word :std_logic_vector(5 - 1 downto 0) := (others => '0');
...
Begin
-- I can of course use this, but this is not very elegant for large SLVs
Big_word <= "000" & small_word & "00";
Big_word <= (7 -1 downto 2 => small_word, others => '0');
--The tool I am using (vivado 2018.3) complains about small_word not being of type std_ulogic. Well the word is larger than 1 bit, so that is an issue.

If this is being done in a process you can always do it with two assignments.

Big_word <= (others => '0');
Big_word (7 -1 downto 2) <= small_word;

[CT]

Rick,

Thanks for the input.

I have done this in a process before with the method you mentioned.

Do you know of a 1 or 2 lines solution when NOT in a process?

CT

[Rick C]

Post by CT
Rick,
Thanks for the input.
I have done this in a process before with the method you mentioned.
Do you know of a 1 or 2 lines solution when NOT in a process?
CT

I would have expected your original solution to have worked. I believe this was only supported once VHDL-2008 came out. Is your tool set for 2008? Often this has to be enabled since by default tools frequently don't support these "modern extensions". Or it is possible it is simply not supported.

Oh, wait! I see in your original post you are limited to VHDL '87. So there's the rub!

Doulos is a great source of info on issues like this. Here is a page talking about the 2008 extensions.

https://www.doulos.com/knowhow/vhdl_designers_guide/vhdl_2008/vhdl_200x_ease/

Search down the page for "Vectors in aggregates". They explain this was only added in 2008. What tool are you using? I would be hard pressed to use a tool that is so out of sync with the real world.

[CT]

Rick,

I am using vivado 2018.3. I believe it is defaulted to VHDL 87. I will look into upgrading to 08. I'm still green here, and wanted to write my design in a way that could be used in as many places as possible. VHDL 08 is supported most everywhere you say? I'll have to do some reading up on this.

That link is a great resource, I will keep that in mind.

Thanks again

CT

[Rick C]

Post by CT
Rick,
I am using vivado 2018.3. I believe it is defaulted to VHDL 87. I will look into upgrading to 08. I'm still green here, and wanted to write my design in a way that could be used in as many places as possible. VHDL 08 is supported most everywhere you say? I'll have to do some reading up on this.
That link is a great resource, I will keep that in mind.
Thanks again
CT

https://www.xilinx.com/support/documentation/sw_manuals/xilinx2015_3/ug901-vivado-synthesis.pdf

Chapter 5, page 177.

https://www.xilinx.com/support/answers/62005.html

[Rick C]

Post by CT
Rick,
I am using vivado 2018.3. I believe it is defaulted to VHDL 87. I will look into upgrading to 08. I'm still green here, and wanted to write my design in a way that could be used in as many places as possible. VHDL 08 is supported most everywhere you say? I'll have to do some reading up on this.
That link is a great resource, I will keep that in mind.
Thanks again
CT

Do us a favor and report back once you have found a solution to your problem. It's nice to learn what you've found out.

[CT]

Unfortunately even with vhdl 08 the solution:

Big_word <= (7 -1 downto 2 => small_word, others => '0');

Does not work. When this is used, the whole big_word signal is assigned to '0'.

I will keep trying with vhdl 08 and report back once ive found a solution

[Rob Gaddi]

Post by t***@gmail.com
Big_word <= (7 -1 downto 2 => small_word, others => '0');
Does not work. When this is used, the whole big_word signal is assigned to '0'.
I will keep trying with vhdl 08 and report back once ive found a solution

Slices like that only operate with single bit values.

You can write it as a function or procedure though. That way you can do
sequential assignment.

function pad_right(val:std_logic_vector, len:integer) return std_logic_vector is
variable x : std_logic_vector(len-1 downto 0) := (others => '0');
begin
x(len-1 downto len-(val'length)) := val;
return x;
end function pad_right;

big_word <= pad_right(small_word, big_word'length);

[Rick C]

Post by Rob Gaddi

Post by t***@gmail.com
Big_word <= (7 -1 downto 2 => small_word, others => '0');
Does not work. When this is used, the whole big_word signal is assigned to '0'.
I will keep trying with vhdl 08 and report back once ive found a solution

Slices like that only operate with single bit values.

Can you explain what you mean by "slices like that"??? Like what?

Here is a hard to find example of exactly the slice aggregate operation the OP is talking about.

https://books.google.com/books?id=ETxLguPMEY0C&pg=PA166&lpg=PA166&dq=VHDL+slice+in+an+array+aggregate&source=bl&ots=q9wZSq_TB_&sig=ACfU3U3hcaEpAlXpDSw93ilUOjLnmPUIHg&hl=en&sa=X&ved=2ahUKEwiMm9DE75LoAhX1mXIEHXgcDKEQ6AEwBnoECAgQAQ#v=onepage&q=VHDL%20slice%20in%20an%20array%20aggregate&f=false

Scroll down to p167.

Post by Rob Gaddi
You can write it as a function or procedure though. That way you can do
sequential assignment.
function pad_right(val:std_logic_vector, len:integer) return std_logic_vector is
variable x : std_logic_vector(len-1 downto 0) := (others => '0');
begin
x(len-1 downto len-(val'length)) := val;
return x;
end function pad_right;
big_word <= pad_right(small_word, big_word'length);

The function suggestion might be the best to get past the issue and get the work done.

[Rob Gaddi]

Post by Rick C

Post by Rob Gaddi

Post by t***@gmail.com
Big_word <= (7 -1 downto 2 => small_word, others => '0');
Does not work. When this is used, the whole big_word signal is assigned to '0'.
I will keep trying with vhdl 08 and report back once ive found a solution

Slices like that only operate with single bit values.

Can you explain what you mean by "slices like that"??? Like what?
Here is a hard to find example of exactly the slice aggregate operation the OP is talking about.
https://books.google.com/books?id=ETxLguPMEY0C&pg=PA166&lpg=PA166&dq=VHDL+slice+in+an+array+aggregate&source=bl&ots=q9wZSq_TB_&sig=ACfU3U3hcaEpAlXpDSw93ilUOjLnmPUIHg&hl=en&sa=X&ved=2ahUKEwiMm9DE75LoAhX1mXIEHXgcDKEQ6AEwBnoECAgQAQ#v=onepage&q=VHDL%20slice%20in%20an%20array%20aggregate&f=false
Scroll down to p167.

Post by Rob Gaddi
You can write it as a function or procedure though. That way you can do
sequential assignment.
function pad_right(val:std_logic_vector, len:integer) return std_logic_vector is
variable x : std_logic_vector(len-1 downto 0) := (others => '0');
begin
x(len-1 downto len-(val'length)) := val;
return x;
end function pad_right;
big_word <= pad_right(small_word, big_word'length);

The function suggestion might be the best to get past the issue and get the work done.

You're right. Per 9.3.3.3 of the 2008 LRM:

---
For an element association with a choice that is a discrete range and an
expression of the element type of the aggregate, the value of the expression is
the element at each index value in the range.

For an element association with a choice that is a discrete range and an
expression of the type of the aggregate, each element of the value of the
expression is the value of the element of the aggregate at the matching index
value in the range.
---

I don't think I've ever trusted that a tool was going to implement this properly.

[CT]

Thanks guys,

Youve helped me find a solution that fits my very specific requirements. Seeing as that you guys are definitely more experienced than I, and it dosen't seem like you use the method I am trying to do very frequently, what methods do you prefer for this task (assigning a part of a slv to another slv, the rest '0')?

Using the function Rob used? Or always making this sort of assignment in a combinational process?

CT

[Rick C]

Post by CT
Thanks guys,
Youve helped me find a solution that fits my very specific requirements. Seeing as that you guys are definitely more experienced than I, and it dosen't seem like you use the method I am trying to do very frequently, what methods do you prefer for this task (assigning a part of a slv to another slv, the rest '0')?
Using the function Rob used? Or always making this sort of assignment in a combinational process?

Unless it were part of a conditional, what would be the value of assigning part of an array permanently to zero? In a conditional this could be used since the '0' parts might have other values for other input combinations. But in a simple assignment the '0' parts would always be zero. In that case I likely would have three concurrent assignment statements, one each for the left and right zeros and one for the variable part.

I don't use functions very often unless they are part of a library with potential reuse. Likely I just never developed the habit. It's not like the modularization isn't useful.