Tech Tip: Determining When a Sequence Has Finished

Imagine the complex scenario whereby you start the *same* sequence on multiple sub-drivers.  Naturally each started sequence has its own execution thread, so how do you refer to all the started sequences, and/or know when they have finished? 

Consider the following steps, followed by a code example that references the AXI UVC.

Step 1 - Add "has_started" and "has_ended" flags to "any_sequence", check why aren’t they there already?

Step 2 - Gen the child sequence, add it to a list of running seq’s, start the child sequence.

Step 3 - Wait to see a child sequence started using the "has_started" flag,

Step 4 - Wait for all child sequences to end

And now, the corresponding code example:

extend any_sequence {
   !has_started : bool;
   on_started() is also { has_started = TRUE; };
   !has_ended : bool;
   on_ended() is also { has_ended = TRUE; };
};

extend MAIN virtual_seq {
   !running_seqs : list of any_seq;
   !child_seq : my_seq vr_axi_master_seq;
   drivers : list of vr_axi_master_driver;
      keep drivers == get_sub_drivers().all(it is a vr_axi_master_driver);
 
   body()@driver.clock is only {
      driver.raise_objection(TEST_DONE);
      foreach in drivers {
          gen child_seq keeping {.driver == it;};
          running_seqs.add(child_seq);
          child_seq.start_sequence()
      };
    
      -- wait for a child sequence to start
      wait true(not running_seqs.is_empty() and running_seqs.has(it.has_started));
      -- allow the traffic to complete before dropping the TEST_DONE objection
      wait true(not running_seqs.has(it.has_ended == FALSE));
      driver.drop_objection(TEST_DONE);
   };
};

Make sense?  Don't be shy about commenting / improving upon this example ...

Happy Sequencing!

Team Specman