From 36b022c8a0f74a037c01c085810365b20830ba34 Mon Sep 17 00:00:00 2001 From: Andres Freund Date: Tue, 12 Jul 2016 01:01:28 -0700 Subject: [PATCH 1/2] Perform one only projection to compute agg arguments. Previously we did a ExecProject() for each individual aggregate argument. Doing so is quite a bit cheaper because ExecProject's fastpath can do the work at once in a relatively tight loop, and because it allows to easily slot_getsomeattrs the right amount of columns for all the aggregates. --- src/backend/executor/nodeAgg.c | 136 ++++++++++++++++++++++++++++++----------- src/include/nodes/execnodes.h | 4 ++ 2 files changed, 105 insertions(+), 35 deletions(-) diff --git a/src/backend/executor/nodeAgg.c b/src/backend/executor/nodeAgg.c index 28c15ba..3ed6fb2 100644 --- a/src/backend/executor/nodeAgg.c +++ b/src/backend/executor/nodeAgg.c @@ -213,6 +213,9 @@ typedef struct AggStatePerTransData */ int numInputs; + /* offset of input columns in AggState->evalslot */ + int inputoff; + /* * Number of aggregated input columns to pass to the transfn. This * includes the ORDER BY columns for ordered-set aggs, but not for plain @@ -297,7 +300,6 @@ typedef struct AggStatePerTransData * well go whole hog and use ExecProject too. */ TupleDesc evaldesc; /* descriptor of input tuples */ - ProjectionInfo *evalproj; /* projection machinery */ /* * Slots for holding the evaluated input arguments. These are set up @@ -847,14 +849,20 @@ advance_aggregates(AggState *aggstate, AggStatePerGroup pergroup) int setno = 0; int numGroupingSets = Max(aggstate->phase->numsets, 1); int numTrans = aggstate->numtrans; + TupleTableSlot *slot = aggstate->evalslot; + AggStatePerTrans pertrans; - for (transno = 0; transno < numTrans; transno++) + /* compute input for all aggregates */ + if (aggstate->evalproj) + aggstate->evalslot = ExecProject(aggstate->evalproj, NULL); + + for (transno = 0, pertrans = aggstate->pertrans; transno < numTrans; + transno++, pertrans++) { - AggStatePerTrans pertrans = &aggstate->pertrans[transno]; ExprState *filter = pertrans->aggfilter; int numTransInputs = pertrans->numTransInputs; int i; - TupleTableSlot *slot; + int inputoff = pertrans->inputoff; /* Skip anything FILTERed out */ if (filter) @@ -868,13 +876,10 @@ advance_aggregates(AggState *aggstate, AggStatePerGroup pergroup) continue; } - /* Evaluate the current input expressions for this aggregate */ - slot = ExecProject(pertrans->evalproj, NULL); - if (pertrans->numSortCols > 0) { /* DISTINCT and/or ORDER BY case */ - Assert(slot->tts_nvalid == pertrans->numInputs); + Assert(slot->tts_nvalid >= pertrans->numInputs); /* * If the transfn is strict, we want to check for nullity before @@ -887,7 +892,7 @@ advance_aggregates(AggState *aggstate, AggStatePerGroup pergroup) { for (i = 0; i < numTransInputs; i++) { - if (slot->tts_isnull[i]) + if (slot->tts_isnull[i + inputoff]) break; } if (i < numTransInputs) @@ -899,10 +904,25 @@ advance_aggregates(AggState *aggstate, AggStatePerGroup pergroup) /* OK, put the tuple into the tuplesort object */ if (pertrans->numInputs == 1) tuplesort_putdatum(pertrans->sortstates[setno], - slot->tts_values[0], - slot->tts_isnull[0]); + slot->tts_values[inputoff], + slot->tts_isnull[inputoff]); else - tuplesort_puttupleslot(pertrans->sortstates[setno], slot); + { + /* + * Copy slot contents, starting from inputoff, into sort + * slot. + */ + ExecClearTuple(pertrans->evalslot); + memcpy(pertrans->evalslot->tts_values, + &slot->tts_values[inputoff], + pertrans->numInputs * sizeof(Datum)); + memcpy(pertrans->evalslot->tts_isnull, + &slot->tts_isnull[inputoff], + pertrans->numInputs * sizeof(bool)); + pertrans->evalslot->tts_nvalid = pertrans->numInputs; + ExecStoreVirtualTuple(pertrans->evalslot); + tuplesort_puttupleslot(pertrans->sortstates[setno], pertrans->evalslot); + } } } else @@ -915,8 +935,8 @@ advance_aggregates(AggState *aggstate, AggStatePerGroup pergroup) Assert(slot->tts_nvalid >= numTransInputs); for (i = 0; i < numTransInputs; i++) { - fcinfo->arg[i + 1] = slot->tts_values[i]; - fcinfo->argnull[i + 1] = slot->tts_isnull[i]; + fcinfo->arg[i + 1] = slot->tts_values[i + inputoff]; + fcinfo->argnull[i + 1] = slot->tts_isnull[i + inputoff]; } for (setno = 0; setno < numGroupingSets; setno++) @@ -943,20 +963,25 @@ combine_aggregates(AggState *aggstate, AggStatePerGroup pergroup) { int transno; int numTrans = aggstate->numtrans; + TupleTableSlot *slot = NULL; + AggStatePerTrans pertrans; /* combine not supported with grouping sets */ Assert(aggstate->phase->numsets == 0); - for (transno = 0; transno < numTrans; transno++) - { - AggStatePerTrans pertrans = &aggstate->pertrans[transno]; - AggStatePerGroup pergroupstate = &pergroup[transno]; - TupleTableSlot *slot; - FunctionCallInfo fcinfo = &pertrans->transfn_fcinfo; + /* compute input for all aggregates */ + if (aggstate->evalproj) + slot = ExecProject(aggstate->evalproj, NULL); + + for (transno = 0, pertrans = aggstate->pertrans; transno < numTrans; + transno++, pertrans++) + { + AggStatePerGroup pergroupstate = &pergroup[transno]; + FunctionCallInfo fcinfo = &pertrans->transfn_fcinfo; + int inputoff = pertrans->inputoff; - /* Evaluate the current input expressions for this aggregate */ - slot = ExecProject(pertrans->evalproj, NULL); Assert(slot->tts_nvalid >= 1); + Assert(slot->tts_nvalid + inputoff >= 1); /* * deserialfn_oid will be set if we must deserialize the input state @@ -965,18 +990,18 @@ combine_aggregates(AggState *aggstate, AggStatePerGroup pergroup) if (OidIsValid(pertrans->deserialfn_oid)) { /* Don't call a strict deserialization function with NULL input */ - if (pertrans->deserialfn.fn_strict && slot->tts_isnull[0]) + if (pertrans->deserialfn.fn_strict && slot->tts_isnull[inputoff]) { - fcinfo->arg[1] = slot->tts_values[0]; - fcinfo->argnull[1] = slot->tts_isnull[0]; + fcinfo->arg[1] = slot->tts_values[inputoff]; + fcinfo->argnull[1] = slot->tts_isnull[inputoff]; } else { FunctionCallInfo dsinfo = &pertrans->deserialfn_fcinfo; MemoryContext oldContext; - dsinfo->arg[0] = slot->tts_values[0]; - dsinfo->argnull[0] = slot->tts_isnull[0]; + dsinfo->arg[0] = slot->tts_values[inputoff]; + dsinfo->argnull[0] = slot->tts_isnull[inputoff]; /* Dummy second argument for type-safety reasons */ dsinfo->arg[1] = PointerGetDatum(NULL); dsinfo->argnull[1] = false; @@ -995,8 +1020,8 @@ combine_aggregates(AggState *aggstate, AggStatePerGroup pergroup) } else { - fcinfo->arg[1] = slot->tts_values[0]; - fcinfo->argnull[1] = slot->tts_isnull[0]; + fcinfo->arg[1] = slot->tts_values[inputoff]; + fcinfo->argnull[1] = slot->tts_isnull[inputoff]; } advance_combine_function(aggstate, pertrans, pergroupstate); @@ -2928,6 +2953,53 @@ ExecInitAgg(Agg *node, EState *estate, int eflags) aggstate->numaggs = aggno + 1; aggstate->numtrans = transno + 1; + /* + * Build a single projection computing the aggregate arguments for all + * aggregates at once, that's considerably faster than doing it separately + * for each. + */ + { + List *inputeval = NIL; + int offset = 0; + + for (transno = 0; transno < aggstate->numtrans; transno++) + { + AggStatePerTrans pertrans = &pertransstates[transno]; + ListCell *arg; + + pertrans->inputoff = offset; + + /* + * Adjust resno in a copied target entry, to point into the + * combined slot. + */ + foreach(arg, pertrans->aggref->args) + { + TargetEntry *tle; + + Assert(IsA(lfirst(arg), TargetEntry)); + tle = copyObject(lfirst(arg)); + tle->resno += offset; + + inputeval = lappend(inputeval, tle); + } + + offset += list_length(pertrans->aggref->args); + } + + aggstate->evaldesc = ExecTypeFromTL(inputeval, false); + + aggstate->evalslot = ExecInitExtraTupleSlot(estate); + + inputeval = (List *) ExecInitExpr((Expr *) inputeval, + (PlanState *) aggstate); + aggstate->evalproj = ExecBuildProjectionInfo(inputeval, + aggstate->tmpcontext, + aggstate->evalslot, + NULL); + ExecSetSlotDescriptor(aggstate->evalslot, aggstate->evaldesc); + } + return aggstate; } @@ -3127,12 +3199,6 @@ build_pertrans_for_aggref(AggStatePerTrans pertrans, (errcode(ERRCODE_GROUPING_ERROR), errmsg("aggregate function calls cannot be nested"))); - /* Set up projection info for evaluation */ - pertrans->evalproj = ExecBuildProjectionInfo(pertrans->args, - aggstate->tmpcontext, - pertrans->evalslot, - NULL); - /* * If we're doing either DISTINCT or ORDER BY for a plain agg, then we * have a list of SortGroupClause nodes; fish out the data in them and diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h index f6f73f3..30e1aee 100644 --- a/src/include/nodes/execnodes.h +++ b/src/include/nodes/execnodes.h @@ -1863,6 +1863,10 @@ typedef struct AggState List *hash_needed; /* list of columns needed in hash table */ bool table_filled; /* hash table filled yet? */ TupleHashIterator hashiter; /* for iterating through hash table */ + /* support for evaluation of agg inputs */ + TupleTableSlot *evalslot; /* slot for agg inputs */ + ProjectionInfo *evalproj; /* projection machinery */ + TupleDesc evaldesc; /* descriptor of input tuples */ } AggState; /* ---------------- -- 2.10.2