Day 100: Constrained randomization I: rand and constraint blocks
Constrained randomization I
Instead of hand-writing every test, you describe the legal input space and let the solver generate it. Mark fields `rand`, then write `constraint` blocks expressing the rules (valid ranges, relationships, weightings). Calling randomize() produces a random-but-legal value each time. Thousands of these explore corners you'd never enumerate by hand — the essence of constrained-random verification.
class axi_txn;
rand bit [31:0] addr;
rand bit [31:0] data;
rand bit is_write;
rand int len;
constraint c_addr { addr inside {[32'h1000:32'h5FFF]}; } // valid range
constraint c_align { addr[1:0] == 2'b00; } // word-aligned
constraint c_len { len inside {[1:16]}; } // burst length
// usage: axi_txn t = new(); assert(t.randomize());
endclassConstraints encode the spec
A good constraint set is a machine-checkable encoding of the interface spec: 'addresses are word-aligned and in range', 'bursts are 1–16 beats'. The solver then generates only *legal* stimulus, but with far more variety than a human. When a constraint is too loose you generate illegal inputs; too tight and you miss corners — tuning that balance is a real DV skill.
Key terms
- rand
- A modifier marking a class field for randomization by randomize().
- constraint block
- A declarative set of rules the solver must satisfy when randomizing.
- randomize()
- The method that assigns rand fields a random solution meeting all constraints.
- inside
- A constraint operator restricting a value to a set or range.
Before moving on, you should be able to
What is the core idea of constrained-random verification?
