On Thu, Jul 22, 2021 at 9:21 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Yeah, I should have said "2GB plus palloc slop". It doesn't surprise
> me a bit that we seem to be eating another 20% on top of the nominal
> limit.
MAX_KILOBYTES is the max_val for the work_mem GUC itself, and has been
for many years. The function get_hash_mem() returns a work_mem-style
int that callers refer to as hash_mem -- the convention is that
callers pretend that there is a work_mem style GUC (called hash_mem)
that they must access by calling get_hash_mem().
I don't see how it's possible for get_hash_mem() to be unable to
return a hash_mem value that could be represented by work_mem
directly. MAX_KILOBYTES is an annoyingly low limit on Windows, where
sizeof(long) is 4. But that's nothing new.
--
Peter Geoghegan
