Dynamic Arrays, Queues, and Associative Arrays

Choose a testbench collection for variable batches, transaction queues, and indexed results.

Collections for the testbench

Dynamic collections manage a number of transactions that is unknown at compile time. They belong mainly in verification and high-level models rather than synthesizable RTL.

Three forms serve different needs:

• a dynamic array for a size allocated at run time;
• a queue for easy insertion and removal at its ends;
• an associative array for elements indexed by sparse keys.

Dynamic array

int samples[];
 
samples = new[128];
foreach (samples[i])
  samples[i] = i * 2;
 
$display("count=%0d", samples.size());
samples.delete();

new[128] allocates 128 elements. A new allocation can preserve part of the old content with new[new_size](samples). delete() releases the array.

This format fits a data block whose size is known at the beginning of an operation and which needs frequent direct indexing.

Queue

typedef logic [31:0] word_t;
word_t expected_q[$];
 
expected_q.push_back(32'h1234_5678);
expected_q.push_back(32'hCAFE_BABE);
 
if (expected_q.size() != 0) begin
  word_t expected = expected_q.pop_front();
  check_word(expected, observed);
end

[$] declares an unbounded queue. push_back and pop_front provide FIFO behavior. Other methods include push_front, pop_back, insert, and delete.

A queue works well in a scoreboard when results leave in the same order as commands.

Associative array

logic [31:0] expected_by_id[int unsigned];
 
expected_by_id[transaction_id] = predicted_value;
 
if (expected_by_id.exists(response_id)) begin
  assert (observed == expected_by_id[response_id]);
  expected_by_id.delete(response_id);
end else begin
  $error("Unexpected response id %0d", response_id);
end

Only used keys allocate elements. This format fits responses that may return in a different order, provided each has a unique identifier.

The first, next, last, and prev methods iterate over keys. Their order follows the index type rather than insertion order.

Choose the simplest matching structure

A collection should reflect the real protocol. A queue is enough for an ordered flow. An associative array is useful for reordered responses. Adding complex lookup when a FIFO is sufficient makes the scoreboard harder to trust than the DUT.

Check for an empty collection before pop_front, and verify that collections are empty at the end of the test. A remaining expected item often means a missing response.

Key points

• Dynamic collections mainly belong in testbenches.
• A dynamic array provides resizable, densely indexed storage.
• A queue naturally matches an ordered flow.
• An associative array finds a result by key.
• Queues and tables should be empty or explained at the end of a test.

📝 Test your knowledge - Chapter quiz