zgornel
commented
4 years ago It is unclear whether this is actually a desirable feature. Multi-threading can be used in:
- data reading/parsing - already the case (individual loaders can be multi-threaded)
- document embedding
- searching (at searcher level or index level)
- live re-indexing (during operation) - already the case
While document embedding and searching can be paralelized, their advantages are somewhat debatable: the added complexity in managing thread occupancy is not warranted by the increase in speed. Also, except for very low memory situations, parallelization though multiple engine instances is desirable.
At this point this task can be closed and should be re-opened only if additional paralellization of the engine operations becomes required.
The cc-multithreading contains a multi-threaded implementation of search, document embedding.
Add multi-threading support where needed and bump julia dependency to v1.3+