On 19 January 2016 at 18:04, Tomas Vondra <tomas.von...@2ndquadrant.com> wrote:
> Hi, > > On 01/19/2016 05:00 AM, David Rowley wrote: > >> On 19 January 2016 at 06:03, Pavel Stehule <pavel.steh...@gmail.com >> <mailto:pavel.steh...@gmail.com>> wrote: >> >> ... > >> >> It is strange, why hashaggregate is too slow? >> >> >> Good question. I looked at this and found my VM was swapping like crazy. >> Upon investigation it appears that's because, since the patch creates a >> memory context per aggregated group, and in this case I've got 1 million >> of them, it means we create 1 million context, which are >> ALLOCSET_SMALL_INITSIZE (1KB) in size, which means about 1GB of memory, >> which is more than my VM likes. >> > > Really? Where do we create the memory context? IIRC string_agg uses the > aggcontext directly, and indeed that's what I see in string_agg_transfn and > makeStringAggState. > > Yeah, all this is talk is relating to Robert's expandedstring-v1.patch which changes string_agg to use text and expanded-objects. This also means that a memory context is created per group, which is rather a big overhead. -- David Rowley http://www.2ndQuadrant.com/ PostgreSQL Development, 24x7 Support, Training & Services
