Compute information about EEOP_*_FETCHSOME at expression init time.

Previously this information was computed when JIT compiling an
expression.  But the information is useful for assertions in the
non-JIT case too (for assertions), therefore it makes sense to move
it.

This will, in a followup commit, allow to treat different slot types
differently. E.g. for virtual slots there's no need to generate a JIT
function to deform the slot.

Author: Andres Freund
Discussion: https://postgr.es/m/20181105210039.hh4vvi4vwoq5ba2q@alap3.anarazel.de

diff --git a/src/backend/executor/execExpr.c b/src/backend/executor/execExpr.c
index cd5ee91cd435db556b34f94ef1feaa70b3b5d66b..d64dab43b8df70c3f62e4a3d009766de8874b18f 100644 (file)
--- a/src/backend/executor/execExpr.c
+++ b/src/backend/executor/execExpr.c
@@ -65,6 +65,7 @@ static void ExecInitFunc(ExprEvalStep *scratch, Expr *node, List *args,
 static void ExecInitExprSlots(ExprState *state, Node *node);
 static void ExecPushExprSlots(ExprState *state, LastAttnumInfo *info);
 static bool get_last_attnums_walker(Node *node, LastAttnumInfo *info);
+static void ExecComputeSlotInfo(ExprState *state, ExprEvalStep *op);
 static void ExecInitWholeRowVar(ExprEvalStep *scratch, Var *variable,
                    ExprState *state);
 static void ExecInitArrayRef(ExprEvalStep *scratch, ArrayRef *aref,
@@ -2288,21 +2289,30 @@ ExecPushExprSlots(ExprState *state, LastAttnumInfo *info)
    {
        scratch.opcode = EEOP_INNER_FETCHSOME;
        scratch.d.fetch.last_var = info->last_inner;
+       scratch.d.fetch.fixed = false;
+       scratch.d.fetch.kind = NULL;
        scratch.d.fetch.known_desc = NULL;
+       ExecComputeSlotInfo(state, &scratch);
        ExprEvalPushStep(state, &scratch);
    }
    if (info->last_outer > 0)
    {
        scratch.opcode = EEOP_OUTER_FETCHSOME;
        scratch.d.fetch.last_var = info->last_outer;
+       scratch.d.fetch.fixed = false;
+       scratch.d.fetch.kind = NULL;
        scratch.d.fetch.known_desc = NULL;
+       ExecComputeSlotInfo(state, &scratch);
        ExprEvalPushStep(state, &scratch);
    }
    if (info->last_scan > 0)
    {
        scratch.opcode = EEOP_SCAN_FETCHSOME;
        scratch.d.fetch.last_var = info->last_scan;
+       scratch.d.fetch.fixed = false;
+       scratch.d.fetch.kind = NULL;
        scratch.d.fetch.known_desc = NULL;
+       ExecComputeSlotInfo(state, &scratch);
        ExprEvalPushStep(state, &scratch);
    }
 }
@@ -2355,6 +2365,94 @@ get_last_attnums_walker(Node *node, LastAttnumInfo *info)
                                  (void *) info);
 }
 
+/*
+ * Compute additional information for EEOP_*_FETCHSOME ops.
+ *
+ * The goal is to determine whether a slot is 'fixed', that is, every
+ * evaluation of the the expression will have the same type of slot, with an
+ * equivalent descriptor.
+ */
+static void
+ExecComputeSlotInfo(ExprState *state, ExprEvalStep *op)
+{
+   PlanState *parent = state->parent;
+   TupleDesc   desc = NULL;
+   const TupleTableSlotOps *tts_ops = NULL;
+   bool isfixed = false;
+
+   if (op->d.fetch.known_desc != NULL)
+   {
+       desc = op->d.fetch.known_desc;
+       tts_ops = op->d.fetch.kind;
+       isfixed = op->d.fetch.kind != NULL;
+   }
+   else if (!parent)
+   {
+       isfixed = false;
+   }
+   else if (op->opcode == EEOP_INNER_FETCHSOME)
+   {
+       PlanState  *is = innerPlanState(parent);
+
+       if (parent->inneropsset && !parent->inneropsfixed)
+       {
+           isfixed = false;
+       }
+       else if (parent->inneropsset && parent->innerops)
+       {
+           isfixed = true;
+           tts_ops = parent->innerops;
+       }
+       else if (is)
+       {
+           tts_ops = ExecGetResultSlotOps(is, &isfixed);
+           desc = ExecGetResultType(is);
+       }
+   }
+   else if (op->opcode == EEOP_OUTER_FETCHSOME)
+   {
+       PlanState  *os = outerPlanState(parent);
+
+       if (parent->outeropsset && !parent->outeropsfixed)
+       {
+           isfixed = false;
+       }
+       else if (parent->outeropsset && parent->outerops)
+       {
+           isfixed = true;
+           tts_ops = parent->outerops;
+       }
+       else if (os)
+       {
+           tts_ops = ExecGetResultSlotOps(os, &isfixed);
+           desc = ExecGetResultType(os);
+       }
+   }
+   else if (op->opcode == EEOP_SCAN_FETCHSOME)
+   {
+       desc = parent->scandesc;
+
+       if (parent && parent->scanops)
+           tts_ops = parent->scanops;
+
+       if (parent->scanopsset)
+           isfixed = parent->scanopsfixed;
+   }
+
+   if (isfixed && desc != NULL && tts_ops != NULL)
+   {
+       op->d.fetch.fixed = true;
+       op->d.fetch.kind = tts_ops;
+       op->d.fetch.known_desc = desc;
+   }
+   else
+   {
+       op->d.fetch.fixed = false;
+       op->d.fetch.kind = NULL;
+       op->d.fetch.known_desc = NULL;
+   }
+}
+
 /*
  * Prepare step for the evaluation of a whole-row variable.
  * The caller still has to push the step.
@@ -3255,12 +3353,18 @@ ExecBuildGroupingEqual(TupleDesc ldesc, TupleDesc rdesc,
    /* push deform steps */
    scratch.opcode = EEOP_INNER_FETCHSOME;
    scratch.d.fetch.last_var = maxatt;
+   scratch.d.fetch.fixed = false;
    scratch.d.fetch.known_desc = ldesc;
+   scratch.d.fetch.kind = lops;
+   ExecComputeSlotInfo(state, &scratch);
    ExprEvalPushStep(state, &scratch);
 
    scratch.opcode = EEOP_OUTER_FETCHSOME;
    scratch.d.fetch.last_var = maxatt;
+   scratch.d.fetch.fixed = false;
    scratch.d.fetch.known_desc = rdesc;
+   scratch.d.fetch.kind = rops;
+   ExecComputeSlotInfo(state, &scratch);
    ExprEvalPushStep(state, &scratch);
 
    /*

diff --git a/src/backend/jit/llvm/llvmjit_expr.c b/src/backend/jit/llvm/llvmjit_expr.c
index 0dbc1e41062def6c2928aca102d5d8db5ae1a6a7..be9b2aecffe0103ce74def93e80e3dc0caa6a210 100644 (file)
--- a/src/backend/jit/llvm/llvmjit_expr.c
+++ b/src/backend/jit/llvm/llvmjit_expr.c
@@ -276,6 +276,8 @@ llvm_compile_expr(ExprState *state)
                    LLVMValueRef v_slot;
                    LLVMBasicBlockRef b_fetch;
                    LLVMValueRef v_nvalid;
+                   LLVMValueRef l_jit_deform = NULL;
+                   const TupleTableSlotOps *tts_ops = NULL;
 
                    b_fetch = l_bb_before_v(opblocks[i + 1],
                                            "op.%d.fetch", i);
@@ -283,40 +285,22 @@ llvm_compile_expr(ExprState *state)
                    if (op->d.fetch.known_desc)
                        desc = op->d.fetch.known_desc;
 
-                   if (opcode == EEOP_INNER_FETCHSOME)
-                   {
-                       PlanState  *is = innerPlanState(parent);
+                   if (op->d.fetch.fixed)
+                       tts_ops = op->d.fetch.kind;
 
+                   if (opcode == EEOP_INNER_FETCHSOME)
                        v_slot = v_innerslot;
-
-                       if (!desc &&
-                           is &&
-                           is->ps_ResultTupleSlot &&
-                           TTS_FIXED(is->ps_ResultTupleSlot))
-                           desc = is->ps_ResultTupleSlot->tts_tupleDescriptor;
-                   }
                    else if (opcode == EEOP_OUTER_FETCHSOME)
-                   {
-                       PlanState  *os = outerPlanState(parent);
-
                        v_slot = v_outerslot;
-
-                       if (!desc &&
-                           os &&
-                           os->ps_ResultTupleSlot &&
-                           TTS_FIXED(os->ps_ResultTupleSlot))
-                           desc = os->ps_ResultTupleSlot->tts_tupleDescriptor;
-                   }
                    else
-                   {
                        v_slot = v_scanslot;
-                       if (!desc && parent)
-                           desc = parent->scandesc;
-                   }
 
                    /*
                     * Check if all required attributes are available, or
                     * whether deforming is required.
+                    *
+                    * TODO: skip nvalid check if slot is fixed and known to
+                    * be a virtual slot.
                     */
                    v_nvalid =
                        l_load_struct_gep(b, v_slot,
@@ -336,19 +320,21 @@ llvm_compile_expr(ExprState *state)
                     * function specific to tupledesc and the exact number of
                     * to-be-extracted attributes.
                     */
-                   if (desc && (context->base.flags & PGJIT_DEFORM))
+                   if (tts_ops && desc && (context->base.flags & PGJIT_DEFORM))
                    {
-                       LLVMValueRef params[1];
-                       LLVMValueRef l_jit_deform;
-
                        l_jit_deform =
                            slot_compile_deform(context, desc,
                                                op->d.fetch.last_var);
+                   }
+
+                   if (l_jit_deform)
+                   {
+                       LLVMValueRef params[1];
+
                        params[0] = v_slot;
 
                        LLVMBuildCall(b, l_jit_deform,
                                      params, lengthof(params), "");
-
                    }
                    else
                    {

diff --git a/src/include/executor/execExpr.h b/src/include/executor/execExpr.h
index ac53935d700b686debea905894da4b93e0991dec..194bf46e0f570c596ab82ddd2fef967560787884 100644 (file)
--- a/src/include/executor/execExpr.h
+++ b/src/include/executor/execExpr.h
@@ -262,7 +262,12 @@ typedef struct ExprEvalStep
        {
            /* attribute number up to which to fetch (inclusive) */
            int         last_var;
+           /* will the type of slot be the same for every invocation */
+           bool        fixed;
+           /* tuple descriptor, if known */
            TupleDesc   known_desc;
+           /* type of slot, can only be relied upon if fixed is set */
+           const TupleTableSlotOps *kind;
        }           fetch;
 
        /* for EEOP_INNER/OUTER/SCAN_[SYS]VAR[_FIRST] */