Blocking Vs Non Blocking

We had presented some introductory tutorial on blocking and non blocking assignment. We will now present some real life issues, solution and best practices for blocking and non blocking assignment statements
Consider the 4 bit ring counter example. A 4 bit ring counter will count as 

0001
0010
0100
1000
0001
..
.

and so on
The following is one implemmentation of the 4 bit ring counter 

// referencedesigner.com 
// 4 bit ring counter example 
module four_bit_ring_counter (
      input clock,
      input reset,
  output [3:0] q
    );
 
  reg[3:0] a;
 
    always @(posedge clock)
      if (reset)
        a = 4'b0001;
 
      else
        begin
        a <=  a<<1; // Notice the non blocking assignment
        a[0] <=a[3];
        end
 
    assign q = a;
 
  endmodule


At the reset, the initial value of the counter is 4'b1000. The ring counter is basically an implemmentation of left shift except when the MSB is 1. When the MSB is 1, its value is copied to the LSB.

A following test bench is used to display the values 

`timescale 1ns / 1ps
module stimulus;
	// Inputs
	reg clock;
	reg reset;
	// Outputs
  wire[3:0] q;
	// Instantiate the Unit Under Test (UUT)
	four_bit_ring_counter r1 (
      .clock(clock), 
      .reset(reset), 
      .q(q)
	);
 
  always #10 clock = ~clock;
 
	initial begin
		// Initialize Inputs
    clock = 0;  
	reset = 0;
 
	#5 reset = 1;
	#20 reset = 0; 
	#100 $finish;
    end  
 
		initial begin
          $monitor($time, " clock=%1b,reset=%1b,q=%4b",clock,reset,q);
		 end
 
endmodule


It will give the following output 


		   0 clock=0,reset=0,q=xxxx
           5 clock=0,reset=1,q=xxxx
           10 clock=1,reset=1,q=0001
           20 clock=0,reset=1,q=0001
           25 clock=0,reset=0,q=0001
           30 clock=1,reset=0,q=0010
           40 clock=0,reset=0,q=0010
           50 clock=1,reset=0,q=0100
           60 clock=0,reset=0,q=0100
           70 clock=1,reset=0,q=1000
           80 clock=0,reset=0,q=1000
           90 clock=1,reset=0,q=0001
           100 clock=0,reset=0,q=0001
           110 clock=1,reset=0,q=0010
           120 clock=0,reset=0,q=0010


The most important line to see is 


        a <=  a<<1; // Notice the non blocking assignment
        a[0] <=a[3];
For example consider the case when the a has the value 1000. Usind the non blocking statement a <= a<<1, it is shifted towards left by 1. But before this execution blocks the execution of the next statement 

        a[0] <=a[3];
we want to assign [a3]'s value to a[0].

If we write the blocking statements in place of non blocking as in 

        a =  a<<1; 
        a[0] =a[3];
The output as given by icarus compiler was 


		   0 clock=0,reset=0,q=xxxx
                   5 clock=0,reset=1,q=xxxx
                  10 clock=1,reset=1,q=0001
                  20 clock=0,reset=1,q=0001
                  25 clock=0,reset=0,q=0001
                  30 clock=1,reset=0,q=0010
                  40 clock=0,reset=0,q=0010
                  50 clock=1,reset=0,q=0100
                  60 clock=0,reset=0,q=0100
                  70 clock=1,reset=0,q=1001
                  80 clock=0,reset=0,q=1001
                  90 clock=1,reset=0,q=0010
                 100 clock=0,reset=0,q=0010
                 110 clock=1,reset=0,q=0100
                 120 clock=0,reset=0,q=0100


In the blocking case, the two statements execute sequantially. And therefore, when a is 0100 , and the first blocking assignment a = a << 1; makses is 1000. The next assignment statement a[0] =a[3]; executes next SEQUENTIALLY and a[0] gets assigned as 1. Hence the unexpected result of 1001.