Generate Loop in Verilog 2001

[appu1985]

Pls give me an idea if there is a module which i need to instantiate in a generate for loop such as outputs of the first instant shall go as an input to the next instant.If u could show me by the help of a small example its shall be good.

 

[echo47]

verilog generate loop

I assume you are talking about Verilog 2001.

Here's a simple 8-bit shift register created by generating eight "dflop" modules.
The nine "path" wires connect the chain of dflops, from first input through last output.
"count" simply generates a square-wave test signal, like a built-in test bench.

module top (clk, out);
  input         clk;
  reg     [3:0] count = 0;
  wire    [8:0] path;
  output        out;

  always @ (posedge clk)
    count <= count + 1;

  assign path[0] = count[3];
  assign out = path[8];

  genvar n;
  generate
    for (n=0; n<=7; n=n+1) begin : flops
      dflop u (.clk(clk), .in(path[n]), .out(path[n+1]));
    end
  endgenerate
endmodule


module dflop (clk, in, out);
  input         clk, in;
  output reg    out;

  always @ (posedge clk)
    out <= in;
endmodule

 

[appu1985]

system verilog generate

And is this generate keyword synthesizeable

 

[echo47]

generate loop verilog

Xilinx ISE 9.1i does synthesize my example code.

 

[appu1985]

verilog 2001 generate

thanx for ur help

 

[vfdff]

verilog for loop assign

output reg out;

I don't think it is a good style

 

[rberek]

verilog generate for loop

output reg out;

I don't think it is a good style

That is standard Verilog 2001 syntax. You no longer need to have separate input/output and reg/wire definitions for the same signal.

 

[boardlanguage]

systemverilog generate statement

That is standard Verilog 2001 syntax. You no longer need to have separate input/output and reg/wire definitions for the same signal.

I've noticed Verilog-2001 module headers (inline port declaration + name) still causes problems with some tools. It's 6 years after Verilog-2001 was formally standardized, and a lot of low-end simulator/synthesis tools have impose limitations on Verilog-2001 syntax.

For example, Cadence Ambit/PKS5 doesn't like downward-ordered (Verilog-2001) generate loops. Here's a (silly) example:

  // bad loop - won't synthesize properly in Xilinx XST 9.1i, or Cadence Buildgates
  genvar g;
  generate for ( x=7; x > 0; x=x-1 ) begin : bad_loop
     if ( x == 7 )
      assign my_signal[x] = my_input;
    else
      assign my_signal[x] = ^my_signal[7:x+1]; // composite-XOR
  end

 

[rberek]

verilog for loop generate

True. I've been lucky so far in that the tools I've used have been happy with Verilog 2001. I'm glad, because it does save some typing and makes things more readable.

r.b.

 

[echo47]

generate systemverilog

Hi boardlanguage, Your "generate" statement degenerates into almost nothing. Maybe there's a typo somewhere?

assign my_signal[7] = my_input;           // = my_input
assign my_signal[6] = ^my_signal[7:7];    // = my_input
assign my_signal[5] = ^my_signal[7:6];    // = zero
assign my_signal[4] = ^my_signal[7:5];    // = zero
assign my_signal[3] = ^my_signal[7:4];    // = zero
assign my_signal[2] = ^my_signal[7:3];    // = zero
assign my_signal[1] = ^my_signal[7:2];    // = zero

Also, "genvar g" should be "genvar x"

My test module:

module top (my_input, my_signal);
  input         my_input;
  output  [7:1] my_signal;

  genvar x;
  generate
    for ( x=7; x > 0; x=x-1 ) begin : bad_loop
      if ( x == 7 )
        assign my_signal[x] = my_input;
      else
        assign my_signal[x] = ^my_signal[7:x+1]; // composite-XOR
    end
  endgenerate
endmodule

It synthesizes fine in ISE 9.1.03i to Spartan-3 (I checked the post-route simulation), but there's almost nothing to synthesize.

I use "generate" frequently in my Xilinx projects, so I'm interested in any XST bugs.

 

[boardlanguage]

system verilog genvar

Opps! Yes, I typed that example on the fly, so it wasn't a practical (*real*) example.

The true-testcase I had instantiated some instances inside the body of the generate/for loop. XST 9.1i and Cadence/PKS5 did NOT like the downward ordered loop. But when I changed the loop-order, the same file synthesized correctly.

Anyway, I've also had A\ltera's Qua\rtus-II 7.1 (web-edition) crash on me, when I declare signals inside a Systemverilog generate-block:

typedef enum int { 
  TARGET_A = 0,
  TARGET_B = 1,
  TARGET_C = 2
  } et_target;

module haha #(
  parameter et_target target = TARGET_A
  ) (
  input logic [7:0] x, 
  input logic [7:0] y,
  output logic [7:0] z
  );

// yet ... another ... silly ... example ...
 
generate
if ( target == TARGET_A ) begin : gen_target_a

    logic [7:0] x1, y1;  // <-- sometimes causes 
                    // A ltera Quartus-II 7.1 to crash!
    assign x1 = ~x;
   assign y1 = {y,y}>>1; // rotate-right 1-bit

  assign z = x1 + y1;
end // : gen_target_a

if ( target == TARGET_B ) begin : gen_target_b

  assign z = x + y;
end // : gen_target_a

endgenerate

endmodule // : haha

Now Quartus-II 7.1 doesn't crash consistently -- it really seems to depend on what else is in the design-hierarchy. But I did find out, moving the 'logic [7:0] x1, y1' declaration to *OUTSIDE* the generate-block always fixed the crashes.

 

[echo47]

generate genvar usage

What language is that? I've never seen "typedef" or "logic" in Verilog. SystemVerilog or VPI perhaps?

 

[boardlanguage]

verilog generate genvar

What language is that? I've never seen "typedef" or "logic" in Verilog. SystemVerilog or VPI perhaps?

Sorry again. Yes, that's Systemverilog. Altera Quartus-II 7.1 supports Verilog, Systemverilog, and VHDL synthesis. The Systemverilog support is primitive (lots of constructs aren't suppored yet), but I like using Quartus-II 7.1 web-edition as an excuse to practice Systemverilog.

 

[kumarpals]

Re: Generate Loop

Hi all,
I have a doubt in echo 47 code ..
in your code is generate/endgenerate needed? .. wont the for loop do the same job of calling the module mant times with different parameters..