When encountering a control block (e.g. JOIN), writing to the system and stack columns is handled by calling self.execute_op(Operation::Noop). This is unintuitive, makes debugging harder (e.g. a print inside execute_op() prints a bunch of NOOP operations that are not actually in the trace). Also, DYNCALL was the first control block where its effect on the stack and system columns doesn't correspond to any existing operation.
The real purpose of "executing NOOPs" is to
ensure the trace capacity
copy the stack columns over to the next row
advance clock
We should introduce an abstraction that does this cleanly instead of reusing execute_op().
When encountering a control block (e.g.
JOIN), writing to the system and stack columns is handled by callingself.execute_op(Operation::Noop). This is unintuitive, makes debugging harder (e.g. a print insideexecute_op()prints a bunch ofNOOPoperations that are not actually in the trace). Also,DYNCALLwas the first control block where its effect on the stack and system columns doesn't correspond to any existing operation.The real purpose of "executing NOOPs" is to
We should introduce an abstraction that does this cleanly instead of reusing
execute_op().Context: https://github.com/0xPolygonMiden/miden-vm/pull/1535#discussion_r1807544164