Answer
Oracle In-Memory adds a columnar, in-memory store (Enterprise Edition only) that analytical queries can run against several times faster than the row-based store. The inmemory_size database parameter controls the total memory reserved for IM (0 disables it entirely), and inmemory_query can toggle IM query usage on or off at the session or system level. Queries that benefit most tend to have filters, GROUP BY, or inner joins; filter on non-indexed columns; scan many rows while selecting only a few columns; or use aggregate functions like MAX/MIN/SUM. IFS Applications ships pre-defined In-Memory Acceleration Packages (visible under the IM Acceleration Package/Packages screens) that aren't enabled by default — before enabling one, use its Estimate Usage option (or the IM Advisor Lobby) to check its expected memory footprint against your configured inmemory_size, since the whole package needs to fit within that allocation. In practice this has shown a measurable improvement on calculation-heavy screens that load many related records at once, such as Projects with many sub-projects.