How to deal with initial reset of DUT? Use of Reset phase or Sequence

UVM
1

Hi all,

I want to finish resetting my DUT before allowing my driver and monitor to start their work. Can someone show me how you could use reset phase or sequence? Sequence is prefered because UVM cookbook said the reset phase may not exist in the future UVM release.

begin
reset = 1’b1;
repeat (10) @ (posedge clk); // Hold Reset for 10 clocks for DUT’s power up configuration.
reset = 1’b0; // Release DUT from its reset state.
end

Thank you
Michael

2

One suggestion would be to have a separate agent control the reset signal, and use a virtual sequence to send a reset_sequence first followed by your other sequences.

3

In reply to dave_59:

Hi Dave,

Do you mean like this?

Env – agent_i (normal sequence)
– agent_o (only monitor)
– agent_r (reset sequence)

virtual sequence will have both reset and normal sequences. How will order of the sequence be controlled? I like reset sequence to go first before normal sequence. How will monitor in agent_o avoid checking any RTL output activities until the reset sequence be completed?

Thank you,
Michael

4

In reply to mikelisfbay:

You use begin/end blocks to start sequences in order, and fork/join blocks to start sequences in parallel. i.e.

task reset_seq::body();
fork // parallel sequences
  seqA_reset.start(a_seqr,this);
  seqB_reset.start(b_seqr,this);
  seqC_reset.start(c_seqr,this);
join
endtask

task normal_seq::body();
begin // serial sequences
  reset_seq.start(null,this);
  normal_seq.start(null,this);
  shutdown_seq.start(null,this);
end
endtask

If you want to wait for the reset signal in a monitor, you can use the mechanism described here: Stimulus/Signal Wait | Verification Academy

5

In reply to dave_59:

Hi Dave,

I like this simple illustration. The link is very helpful. I see the syntax that put the monitor to wait and then start. I look forward to try it out.

Thank you for writing to me.
Michael

6

In reply to dave_59:

Hi Dave/Micheal

I have this specific requirement in the below code.

task normal_seq::body();
begin
  reset_seq.start(null,this); // reset sequence
  begin
    normal_seq1.start(null,this); // sequences following reset sequence. 
    normal_seq2.start(null,this);
    normal_seq3.start(null,this);
    normal_seq4.start(null,this);
    normal_seq5.start(null,this);
  end
end
endtask

I wanted to have multiple resets during simulation with random delays. and if reset is detected, then normal_seq1 should start executing after each reset.

task normal_seq::body();
fork
  repeat(5) begin
  #($urandom_range(25,20)) 
  reset_seq.start(null,this); // reset sequence
  end
  begin
    normal_seq1.start(null,this); // sequences following reset sequence. 
    normal_seq2.start(null,this);
    normal_seq3.start(null,this);
    normal_seq4.start(null,this);
    normal_seq5.start(null,this);
  end
end
endtask

Problem is i am not finding a way to return my normal sequence flow to normal_seq1 each time reset is detected. Could you please suggest a way to do it.

Thanks

7

In reply to 8Blades:
Why not put the repeat outside the fork, and find a way to cleanly abort the currently running sequences.