Constrained Pseudo-Random Stimulus Generation
Context in the Curriculum
Constrained pseudo-random stimulus generation is identified in the Stimuli Generation lecture material as a topic belonging to Part II: Test Automation, together with the more general topic of Randomness [1]. This places it after Part I, which surveys foundational issues in stimuli generation, including:
- How many generators?
- Level of abstraction
- Online vs. offline generation
- Dynamic vs. static generation
- Test length
The detailed treatment of constrained pseudo-random stimulus generation itself is deferred to Part II of the lecture series [1].
Related Concepts Discussed in the Same Material
Although the lecture slides included in the evidence only briefly name "constrained pseudo-random stimulus generation" as a forthcoming topic, the surrounding material describes several adjacent ideas that situate the concept:
- Dynamic vs. static generation — Dynamic test generation is based on reaction, whereas static test generation is based on planning. The lecture notes that reaction is generally harder than planning because of timing factors and unexpected events, and that most generators use dynamic features only lightly — for example, by observing and reacting to shallow or stable states of the Design Under Verification (DUV) such as architectural registers or the occupancy state of a FIFO [1].
- Offline dynamic generation — Dynamic and static generation should not be confused with online and offline generation. An offline generator can still employ dynamic features by consulting a reference model whose abstraction level and accuracy determine the level and accuracy of information available about the DUV [1].
- Test length trade-offs — Two extreme approaches are considered: short tests (the shortest tests that fulfill a verification-plan requirement, e.g., a two-instruction test for an instruction-pair requirement) and long tests (combining many requirements in a single test wrapped with initial and end sequences). Short tests are easy to create, debug, maintain, and quick to simulate; long tests need fewer total tests, reduce redundant initialization time, exercise states far from the initial state, traverse less-traveled paths, and often reach verification targets in unexpected ways [1].
- Risk of dynamic reaction failing — A worked example notes that a dynamic stimulus intended to trigger forwarding may fail because of the time-distance between instruction entry into the processor and the dispatch queue; intervening events such as pipeline-flushing exceptions can cause the interesting condition to disappear before the relevant pipeline stage is reached [1].
These issues (reacting to DUV state, balancing planning against reaction, and choosing test length) constitute the practical landscape into which constrained pseudo-random stimulus generation is introduced as an automation technique.
Summary
Based on the provided evidence, constrained pseudo-random stimulus generation is presented as a Part II test-automation topic within the COMS30026 Design Verification: Stimuli Generation lecture material. The slides do not elaborate the technique itself in the excerpts available; they only enumerate it alongside Randomness under the Part II heading [1].