Heikki Linnakangas <hlinnakangas@vmware.com> writes:
> Maybe. Or maybe the heuristic read ahead isn't significant/helpful, when you're prefetching with posix_fadvise
anyway.
Yeah. If we're not reading consecutive blocks, readahead is unlikely
to do anything anyhow.
Claudio's comments do suggest that it might be a bad idea to issue a
posix_fadvise when the next block to be examined *is* adjacent to the
current one, though.
regards, tom lane