Evict on write, a management strategy for a prefetch unit and/or first level cache in a multiprocessor system with speculative execution
View Patent ↗In a multiprocessor system with at least two levels of cache, a speculative thread may run on a core processor in parallel with other threads. When the thread seeks to do a write to main memory, this access is to be written through the first level cache to the second level cache. After the write though, the corresponding line is deleted from the first level cache and/or prefetch unit, so that any further accesses to the same location in main memory have to be retrieved from the second level cache. The second level cache keeps track of multiple versions of data, where more than one speculative thread is running in parallel, while the first level cache does not have any of the versions during speculation. A switch allows choosing between modes of operation of a speculation blind first level cache.
1. In a parallel processing system comprising a plurality of cores configured to run speculative threads in parallel, and at least one each of first and second level caches, a method comprising carrying out operations, the operations comprising:
determining whether a speculative thread seeks to write;
upon a positive determination, writing from the speculative thread through the first level cache to the second level cache;
evicting a line from the first level cache and/or prefetch unit corresponding to the writing; and
resolving speculation downstream from the first level cache.
2. The method of claim 1 , wherein the operations comprise, subsequent to evicting, if a speculative thread seeks to access an address corresponding to the line in the first level cache, retrieving an appropriate version of data from the second level cache.
3. The method of claim 2 , wherein the operations comprise tagging data retrieved from the second level cache, when there are multiple versions.
4. The method of claim 3 , wherein the operations comprise not caching responsive to such tagging.
5. The method of claim 1 , wherein the operations comprise, when a speculative thread reads from the first level cache, notifying the second level cache.
6. The method of claim 1 , wherein the operations comprise:
determining whether a speculation has completed; and
if speculation has completed, allowing the first level cache to store results.
7. The method of claim 5 , wherein the second level cache checks its coherence data when receiving an L1 hit notification for whether the L1 could have had a stale copy, and if so, subsequently invalidates an associated speculation.