Add simulation delay after UVM_MAX_QUIT_COUNT reached

UVM
,
1

Hi,
Is there a way to add a predetermined simulation time delay after UVM_MAX_QUIT_COUNT has reached its value?
Thanks

2

In reply to shaygueta:

How about

UVM_MAX_QUIT_COUNT = [your desired quit count] + 1?

Sort of work-around…

Srini

3

In reply to shaygueta:

The only way I can think of is to turn off MAX_QUIT_COUNT by setting it to 0, and then writing a report catcher that reads the error count and sets up a timeout with the pre-determined exit delay.

4

In reply to dave_59:

I do not believe the Simulator is immediatly stopping when MAX_QUIT_COUNT has been reached. I think then it goes through all the steps to stop the Simulation in a regular way. I’s try to set a specific drain_time.

5

In reply to dave_59:

Hi,
Thanks. Can you add an example code for such a report catcher?
The errors can originate from various components in the environmdnt.

6

In reply to shaygueta:
Untested:

class my_report_catcher extends uvm_report_catcher;
  int unsigned max_error = 1; // limit
  time fatal_timeout = 20ns; // amount to delay before exiting
  function new (string name = ""); super.new(name); endfunction
  function action_e catch; 
    if (get_severity() == UVM_ERROR  && --max_error == 0) // short-circuit && operator
      fork 
        #fatal_timeout get_client().die();
      join_none
    return THROW;
  endfunction
endclass : my_report_catcher

Then in your test

my_report_catcher rc_h;
rc_h = new("rc");
rc_h.max_error = 10;
rc_h.fatal_timeout = 100ms;
uvm_report_cb::add(null, rc_h); // all errors in all contexts

P.S. Christoph - UVM calls die() when reaching MAX_QUIT_COUNT.

7

In reply to dave_59:

Nice trick Dave. I would like to extend this to add a command line arg (like UVM_MAX_QUIT_COUNT) and make it little easier to deploy (Didn’t think through it yet fully). I hope you won’t mind me copying this code into a Go2UVM app soon (We already have some form of this - to print a summary of error IDs etc. as part go “go2uvm debug layer” - not released yet though). I will make sure to give the credits to this thread and our “Walking LRM” (As you are referred to on social media).

Having said that, I am bit worried that this may hurt some performance as it is a callback on every message. Maybe we can improve that, will delve deeper later.

Thanks
Srini

8

In reply to Srini @ CVCblr.com:
No, in such a case the simulator does not stop