On Mon, Jul 13, 2020 at 12:47:36PM -0400, Alvaro Herrera wrote: > On 2020-Jul-13, Jeff Davis wrote: > > > On Tue, 2020-07-14 at 02:25 +1200, David Rowley wrote: > > > Updated summary: > > > * For hash_mem = Tomas [7], Justin [16] > > > * For hash_mem_multiplier with a default > 1.0 = DavidG [21] > > > * For hash_mem_multiplier with default = 1.0 = PeterG [15][0], Tom > > > [20][24] > > > > I am OK with these options, but I still prefer a simple escape hatch. > > I'm in favor of hash_mem_multiplier. I think a >1 default is more > sensible than =1 in the long run, but if strategic vote is what we're > doing, then I support the =1 option.
I recanted and support hash_mem_multiplier (or something supporting that behavior, even if it also supports an absolute/scalar value). https://www.postgresql.org/message-id/20200703145620.gk4...@telsasoft.com 1.0 (or -1) is fine, possibly to be >= 1.0 in master at a later date. -- Justin
