On Mon, Nov 09, 2020 at 11:31:15PM -0300, Alvaro Herrera wrote:
> Yeah ... it would be much better if we can make it use atomics instead.
> Currently it's an uint8, and in PGPROC itself it's probably not a big
> deal to enlarge that, but I fear that quadrupling the size of the
> mirroring array in PROC_HDR might be bad for performance. However,
> maybe if we use atomics to access it, then we don't need to mirror it
> anymore? That would need some benchmarking of GetSnapshotData.
Hmm. If you worry about the performance impact here, it is possible
to do a small performance test without this patch. vacuum_rel() sets
the flag for a non-full VACUUM, so with one backend running a manual
VACUUM in loop on an empty relation could apply some pressure on any
benchmark, even a simple pgbench.
--
Michael