id summary reporter owner description type status priority milestone component version resolution keywords cc 3555 Parallelization of PostGIS raster ST_Union aggregate robe Bborie Park "I experimented with this and sadly I think this is a no go for PostgreSQL 9.6, though I'm keeping ourt hope for PostgreSQL 9.7. The reason I think it's a no go (unless we revise postgis raster ST_Union to use similar approach to ST_MemUnion (which might result in more loss than gain) is because I don't think the PostgresQL parallel aggregate infrastructure can support aggregates that use state transitions. Reason I thin that is that function like string_agg, array_agg don't support parallelization in 9.6 http://www.hagander.net/talks/PostgreSQL_9.6.pdf#38 I'm guessing its because they use that internal state transition stuff that we employ with our raster aggregates. I'll try to still experiment with a couple of approaches before I give up on 9.6 parallelization for raster. The ST_MapAlgebra functions look like good candidates for parallelization, so I'll add that aas a separate ticket, as I think that one would be easier to swing." enhancement new medium PostGIS Fund Me raster master