Clarification needed on mix between bit bash sequence and single register write

[Bas Arts]

Consider the following code snippet from uvm-tests/tests/examples/simple/registers/models/aliasing/tb_run.sv:

uvm_reg_sequence seq;

seq = uvm_reg_bit_bash_seq::type_id::create("seq");

seq.model = env.regmodel;

seq.start(env.bus.sqr);

seq.wait_for_sequence_state(UVM_FINISHED);

`uvm_info("Test", "Verifying aliasing...", UVM_NONE);

env.regmodel.Ra.write(status, 32'hDEADBEEF, .parent(seq));

 

I'm trying to understand a few things here.

- the bit bash sequence is in state "UVM_FINISHED" at the moment it is used as a parent sequence for the regmodel register write; should that have a consequence for the write (e.g., the sequence cannot be used anymore because it is finished), or can the register write "reuse" the sequence independent of its state?

- what does it actually mean when a regmodel register write is writing a fixed value 32'hDEADBEEF into a register while using a bit bashing sequence as a parent? Is that possible at all?

 

Thanks for giving any clarification on this!

[Bas Arts]

I'd appreciate if some UVM expert or someone involved in the UVM working group could take a look at this. Until now, there has been no response to the question here nor to the same question asked through the reflector at uvm@lists.accellera.org.

[Mark Strickland]

We looked at this in the working group meeting.  It is strange code in that waiting for UVM_FINISHED is redundant as the start() has returned, but a register access needs a uvm_sequence passed as parent just to get access to a sequencer to run on.  The state of that parent sequence and any purpose associated with the body of that sequence is not relevant.  In this example, the only sequence present is the finished bit_bash one, so that is used for convenience.

[Bas Arts]

Thanks for the clarification!