Dear Hayato, > There is a similar idea which parallelize the initialization. Not sure, is > there > an interactions or comparison between them? Your patch seems to divide a > transaction into several parts but everything is done by the single thread, > whereas proposed in [1] is to dispatch to other threads.
You're absolutely right. Mircea (author of multithread patch #6242 [1]) contacted me with a question if our patches could be merged. At that point in time none of our patches had reviewers and sheer size of each one of them was quite big (~500 lines of new code). Therefore we decided to leave our patches as a separate commits and see what community will say. It's quite clear that there is certain synergy between them and perhaps it would make perfect sense to introduce them together in same release of PostgreSQL. I'm not familiar with intricate details of process that takes place after CommitFest, especially if several reviewed commits in "Ready to go" status touch same area of the product. I believe that my patch shouldn't require too many changes in order to be merged with Mircea's one since core logic to initialize set of rows in three main tables for one particular "scale" is already in place. Thank you for your interest in my commit and Best regards, Boris
