z/Architecture Storage Hierarchy and Cache Notes
Scope
The CPU model must not invent a universal cache size, associativity, replacement policy, or latency for z/Architecture processors. Cache and storage-hierarchy behavior varies by processor generation and implementation.
Modeling levels
- Architectural level: model memory ordering and storage semantics required by the architecture.
- Performance abstraction: optional latency model parameterized by documented processor and memory information.
- Microarchitectural detail: only implement cache geometry, sharing, prefetching, and maintenance behavior when reliable target-specific data is available.
Main storage
Represent main storage and address translation separately from any cache-performance abstraction. Storage protection, translation, and access exceptions must remain correct even when cache timing is not modeled.
Evidence and reporting
Every timing or cache parameter carries an evidence label. Unknown values remain unspecified. Do not expose guessed cache geometry as an architectural guarantee.