Synthesis support for 3D HDL arrays

Status
Not open for further replies.
S

shaiko

Advanced Member level 5
Joined
Aug 20, 2011
Messages
2,644
Helped
303
Reputation
608
Reaction score
297
Trophy points
1,363
Activity points
18,302
Hello,

From your experience,
Do modern FPGA synthesis tools ( Vivado / Quartus Prime ) support 3D (or higher dimension) arrays ?
 

This is what I want to do:

Code: 
-- All array elements are intended to be DFFs

----- Entity generics -----
PIXEL_WIDTH 	:	positive := 8 ;
PIXELS_PER_ROW  :	positive := 16 ;
ROWS_PER_FRAME  :	positive := 9 ;
------------------------------

----- The custom types -----
type generic_pixel is std_logic_vector ( PIXEL_WIDTH - 1 downto 0 ) ;
type generic_row is array ( 0 to PIXELS_PER_ROW - 1 ) of generic_pixel ;
type generic_frame is array ( 0 to ROWS_PER_FRAME - 1 ) of generic_row ;
------------------------------

----- Signals declartion of custom type -----
signal pixel : 	generic_pixel ;
signal row :	generic_row ;
signal frame :	generic_frame ;
------------------------------
 

Do modern FPGA synthesis tools ( Vivado / Quartus Prime ) support 3D (or higher dimension) arrays ?
Click to expand...
Yes. According to the respective HDL capabilities.
 

Have you ever tried and succeeded synthesizing something similar to what I wrote in #2 ?
 

Whats stopping you from running a small design through synthesis to get a result?
I was using a 2D (real 2d, not 1D of 1D like you example) array of sfixed type in quartus 7 years ago and it compiled and inferred multipliers just fine. I would hope this capability still remains.
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
Whats stopping you from running a small design through synthesis to get a result?
Click to expand...
My next step.

I was using a 2D (real 2d, not 1D of 1D like you example)
Click to expand...
Can you please post a short example?
 

code using this type synthesised just fine:

Code: 
type filter_array_t is  array(-1 to 1, -1 to 1) of sfixed(9 downto -8);


I suggest trying your own code example.
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
TrickyDicky,

So if I want to define a 3D array - the syntax is:
Code: 
type generic_3d_array is  array( 0 to 15 , 0 to 31 ) of x ( 7 downto 0 );

And if I want to assign a value to an element I write for example :
Code: 
generic_3d_array ( 5 ) ( 3 ) <= "01110000" ;
?
 

That is not a 3D array - it is a 2D array of a 1D array type.
N d arrays are indexed using a comma (,).

so in your case:

generic_3d_array_signal(5,3) <= "01110000";

Just so that you know - slicing Nd arrays is illegal
 
So how would you define a true 3d array?
Will it be like this?
type generic_3d_array is array( 0 to 15 , 0 to 31 , 7 downto 0)
 

shaiko said:
So how would you define a true 3d array?
Will it be like this?
type generic_3d_array is array( 0 to 15 , 0 to 31 , 7 downto 0)
Click to expand...

Yes

Kevin
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
shaiko said:
So how would you define a true 3d array?
Will it be like this?
type generic_3d_array is array( 0 to 15 , 0 to 31 , 7 downto 0)
Click to expand...

Kevin, Isn't this a 3D array of some unspecified type? Along with a really strange 7 downto 0 index for one of the array dimensions.

I'll admit I get somewhat confused when people start describing things like:
Code: 
type filter_array_t is  array(-1 to 1, -1 to 1) of sfixed(9 downto -8);
as a 3D array, to me it's a 2D array of sfixed types.

So as an example for a Rubik's cube...
Code: 
type colors is (red, green, blue, yellow, white, orange);
type rubiks is array(1 to 3, 1 to 3, 1 to 3) of colors;
Not sure if this is correct or not (and admittedly I'm too lazy to try this out).
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
"IN_FRAME" is an entity input. defined as follows:
Code: 
IN_FRAME : generic_3d_array ( 0 to ROWS_PER_FRAME ) - 1 , 0 to PIXELS_PER_ROW - 1 , BITS_PER_PIXEL - 1 downto 0 ) ;

"row" is a signal that's defined as follows:
Code: 
signal row : generic_2d_array ( 0 to PIXELS_PER_ROW - 1 , BITS_PER_PIXEL - 1 downto 0 ) ;

"pixel" is a signal that's defined as follows:
Code: 
signal pixel : std_logic_vector ( BITS_PER_PIXEL - 1 downto 0 ) ;
I want to do 2 things:

1. Assign a row element from "IN_FRAME" into signal "row" - Is this the correct syntax to do so?
row <= IN_FRAME ( some_row_number ) ;
2. Assign a pixel element from "IN_FRAME" into signal "pixel" - Is this the correct syntax?
pixel <= IN_FRAME ( some_row_number , some_pixel_number ) ;
 

Code: 
signal row : generic_2d_array ( 0 to PIXELS_PER_ROW - 1 , BITS_PER_PIXEL - 1 downto 0 ) ;
see this confuses me to no end...

if you access this row array you end up with

Code: 
row(1,4)

which in my mind says you are looking at bit-4 of pixel 1 in the row, if that is what you are after, then I agree that this is a 2d array, but if you are trying to get the pixel value (the BITS_PER_PIXEL -1 downto 0 stuff) then I consider this a 1d array with each array element being a pixel value of BITS_PER_PIXEL width.

Tricky help I'm so confused!
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating

Neither assignment is correct because you have left off the second and/or third index for IN_FRAME. You need to write some functions like 'Get_Row' and 'Get_Pixel' to do what you intend. Both functions will take as input a generic_3d_array parameter as well as the 'some_row_number' and 'some_pixel_number' parameters that you list. So the assignments would look like this:

Code: 
row <= Get_Row(IN_FRAME, some_row_number) ; 
pixel <= Get_Pixel(IN_FRAME, some_row_number ,  some_pixel_number ) ;

Kevin Jennings
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
see this confuses me to no end...
Click to expand...
ads-ee,

I might be wrong calling it a 2d array...
Sorry

- - - Updated - - -

Neither assignment is correct because you have left off the second and/or third index for IN_FRAME.
Click to expand...

That's what Trickydicky meant by:
Just so that you know - slicing Nd arrays is illegal
Click to expand...
?
 

with a 2d array, you cannot slice so the following is illegal:

Code: 
IN_FRAME : generic_3d_array ( 0 to ROWS_PER_FRAME  - 1 , 0 to PIXELS_PER_ROW - 1 , BITS_PER_PIXEL - 1 downto 0 ) ;

SOME_SIGNAL <= IN_FRAME(0 to 3, 0 to 1, 7 downto 0);  -- illegal

For what you want to do, I suggest arrays of arrays.
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
K-J,
In #15 you wrote:
Neither assignment is correct because you have left off the second and/or third index for IN_FRAME.
Click to expand...
Assume the following code:

Code: 
signal x is array ( 0 to 2 ) of std_logic_vector ( 7 downto 0 ) ;
signal y : std_logic_vector ( 7 downto 0 ) ;  
y <= x ( 1 ) ;

Despite the fact that I left off the second boundary of x - the above will be legal.

How is it any different than:
Code: 
row <= IN_FRAME ( some_row_number ) ;
?
 


The first code returns a known type - a std logic vector.
The second code would return an indeterminant type - you're basically slicing the array. 2+D arrays can only be indexed to individual elements.
 
Reactions: shaiko

    S

    shaiko

    Points: 2
    Helpful Answer Positive Rating
So to achieve what I want - I can either:
1. use an array of array.
2. use a custom function like K-J suggested.

Correct?
 

Status
Not open for further replies.

Similar threads

Menu
Log in
Register
Cookies are required to use this site. You must accept them to continue using the site. Learn more…