Knowing a cache's state letter is only useful if you can say what the core may do, where the newest value is, and who must preserve it. These four lessons separate those questions before following short transactions on the two-core course model. This makes a stale memory value understandable without excusing a stale load.
Lesson | After it you can... |
|---|---|
Classify a copy without confusing readable data with write permission. | |
Follow the newest value from a dirty cache through H to memory and a new reader. | |
Explain why an MOESI O holder preserves dirty data but cannot write. | |
Locate the boundary between readers and a writer and justify the acknowledgement wait. |
Suggested order: read top to bottom, keeping the Course Model beside the snapshots. We use C0 and C1 with private write-back caches, home H and memory; X is one location inside a 64-byte line. Initially memory holds X = 0 and neither cache holds the line; later examples name their warm starting snapshot.
No. Both can be read-only sharers in a read epoch. Equal values do not confer write permission; a writer must first exclude independent readers.