remove unused flags, and remove confusable '*_flags' field

src/codegen.c

@@ -235,10 +235,9 @@ build_composite_devtype_info(TypeCacheEntry *tcache, const char *ext_name)
 	devtype_info  **subtypes = alloca(sizeof(devtype_info *) * tupdesc->natts);
 	devtype_info   *dtype;
 	MemoryContext	oldcxt;
-	uint32_t		extra_flags = DEVKIND__ANY;
-	int				j;
+	uint32_t		subtype_flags = DEVKIND__ANY;
 
-	for (j=0; j < tupdesc->natts; j++)
+	for (int j=0; j < tupdesc->natts; j++)
 	{
 		Form_pg_attribute attr = TupleDescAttr(tupdesc, j);
 
@@ -248,7 +247,7 @@ build_composite_devtype_info(TypeCacheEntry *tcache, const char *ext_name)
 			ReleaseTupleDesc(tupdesc);
 			return NULL;
 		}
-		extra_flags &= dtype->type_flags;
+		subtype_flags &= dtype->type_flags;
 		subtypes[j] = dtype;
 	}
 	ReleaseTupleDesc(tupdesc);
@@ -260,7 +259,7 @@ build_composite_devtype_info(TypeCacheEntry *tcache, const char *ext_name)
 		dtype->type_extension = pstrdup(ext_name);
 	dtype->type_code = TypeOpCode__composite;
 	dtype->type_oid = tcache->type_id;
-	dtype->type_flags = extra_flags | DEVTYPE__USE_KVARS_SLOTBUF;
+	dtype->type_flags = subtype_flags;
 	dtype->type_length = tcache->typlen;
 	dtype->type_align = typealign_get_width(tcache->typalign);
 	dtype->type_byval = tcache->typbyval;
@@ -301,7 +300,7 @@ build_array_devtype_info(TypeCacheEntry *tcache, const char *ext_name)
 		dtype->type_extension = pstrdup(ext_name);
 	dtype->type_code = TypeOpCode__array;
 	dtype->type_oid = tcache->type_id;
-	dtype->type_flags = elem->type_flags | DEVTYPE__USE_KVARS_SLOTBUF;
+	dtype->type_flags = elem->type_flags;
 	dtype->type_length = tcache->typlen;
 	dtype->type_align = typealign_get_width(tcache->typalign);
 	dtype->type_byval = tcache->typbyval;
@@ -1530,7 +1529,7 @@ create_codegen_context(PlannerInfo *root,
 												   pp_info->num_rels + 2]));
 	context->elevel = ERROR;
 	context->root = root;
-	context->required_flags = (pp_info->xpu_task_flags & DEVKIND__ANY);
+	context->xpu_task_flags = pp_info->xpu_task_flags;
 	context->kvecs_ndims = pp_info->num_rels + 1;
 	context->kvecs_usage = 0;
 	context->scan_relid = pp_info->scan_relid;
@@ -1689,7 +1688,7 @@ codegen_var_expression(codegen_context *context,
 		 * on the next depth, and private values are not visible to
 		 * others.
 		 */
-		if ((context->extra_flags & DEVKIND__NVIDIA_GPU) != 0 &&
+		if ((context->xpu_task_flags & DEVKIND__NVIDIA_GPU) != 0 &&
 			kvdef->kv_offset >= 0 &&
 			kvdef->kv_depth != curr_depth)
 			kexp.u.v.var_offset = kvdef->kv_offset;
@@ -1717,7 +1716,7 @@ __codegen_func_expression(codegen_context *context,
 
 	dfunc = pgstrom_devfunc_lookup(func_oid, func_args, func_collid);
 	if (!dfunc ||
-		(dfunc->func_flags & context->required_flags) != context->required_flags)
+		(dfunc->func_flags & context->xpu_task_flags & DEVKIND__ANY) == 0)
 		__Elog("function %s is not supported on the target device",
 			   format_procedure(func_oid));
 	dtype = dfunc->func_rettype;
@@ -3014,7 +3013,7 @@ codegen_build_packed_kvars_move(codegen_context *context, pgstromPlanInfo *pp_in
 	int			gist_depth = context->num_rels + 1;
 
 	/* Only NVIDIA-GPU has kvec-buffer for MoveVars across depth */
-	if ((context->extra_flags & DEVKIND__NVIDIA_GPU) == 0)
+	if ((context->xpu_task_flags & DEVKIND__NVIDIA_GPU) == 0)
 		return;
 
 	sz = MAXALIGN(offsetof(kern_expression,
@@ -3984,7 +3983,7 @@ codegen_build_groupby_actions(codegen_context *context,
  */
 bool
 pgstrom_xpu_expression(Expr *expr,
-					   uint32_t required_xpu_flags,
+					   uint32_t xpu_task_flags,
 					   Index scan_relid,
 					   List *inner_target_list,
 					   int *p_devcost)
@@ -3995,13 +3994,13 @@ pgstrom_xpu_expression(Expr *expr,
 	int			depth;
 	ListCell   *lc;
 
-	Assert((required_xpu_flags & DEVKIND__ANY) == DEVKIND__NVIDIA_GPU ||
-		   (required_xpu_flags & DEVKIND__ANY) == DEVKIND__NVIDIA_DPU);
+	Assert((xpu_task_flags & DEVKIND__ANY) == DEVKIND__NVIDIA_GPU ||
+		   (xpu_task_flags & DEVKIND__ANY) == DEVKIND__NVIDIA_DPU);
 	context = alloca(sz);
 	memset(context, 0, sz);
 	context->elevel = DEBUG2;
 	context->top_expr = expr;
-	context->required_flags = (required_xpu_flags & DEVKIND__ANY);
+	context->xpu_task_flags = xpu_task_flags;
 	context->scan_relid = scan_relid;
 	context->num_rels = num_rels;
 	depth = 1;
@@ -4029,7 +4028,7 @@ estimate_cuda_stack_size(codegen_context *context)
 	ListCell   *lc;
 
 	/* only NVIDIA-GPU needs stack configuration */
-	if ((context->extra_flags & DEVKIND__NVIDIA_GPU) == 0)
+	if ((context->xpu_task_flags & DEVKIND__NVIDIA_GPU) == 0)
 		return 0;
 	/* minimum working area */
 	stack_sz = 3200;

src/gpu_join.c

@@ -1459,7 +1459,7 @@ pgstrom_build_join_tlist_dev(codegen_context *context,
 			if (contain_var_clause((Node *)tle->expr))
 				tlist_dev = __pgstrom_build_tlist_dev_expr(tlist_dev,
 														   (Node *)tle->expr,
-														   context->required_flags,
+														   context->xpu_task_flags,
 														   context->scan_relid,
 														   inner_target_list);
 		}
@@ -1485,7 +1485,7 @@ pgstrom_build_join_tlist_dev(codegen_context *context,
 			if (contain_var_clause(node))
 				tlist_dev = __pgstrom_build_tlist_dev_expr(tlist_dev,
 														   node,
-														   context->required_flags,
+														   context->xpu_task_flags,
 														   context->scan_relid,
 														   inner_target_list);
 		}
@@ -1646,9 +1646,11 @@ PlanXpuJoinPathCommon(PlannerInfo *root,
 	codegen_build_packed_kvars_load(context, pp_info);
 	/* MoveVars for each depth (only GPUs) */
 	codegen_build_packed_kvars_move(context, pp_info);
-
+	/* xpu_task_flags should not be cleared in codege.c */
+	Assert((context->xpu_task_flags &
+			pp_info->xpu_task_flags) == pp_info->xpu_task_flags);
 	pp_info->kvars_deflist = context->kvars_deflist;
-	pp_info->extra_flags = context->extra_flags;
+	pp_info->xpu_task_flags = context->xpu_task_flags;
 	pp_info->extra_bufsz = context->extra_bufsz;
 	pp_info->used_params = context->used_params;
 	pp_info->outer_refs  = outer_refs;

src/gpu_scan.c

@@ -800,9 +800,11 @@ PlanXpuScanPathCommon(PlannerInfo *root,
 	codegen_build_packed_kvars_load(context, pp_info);
 	/* VarMoves for each depth (only GPUs) */
 	codegen_build_packed_kvars_move(context, pp_info);
-
+	/* xpu_task_flags should not be cleared in codege.c */
+	Assert((context->xpu_task_flags &
+			pp_info->xpu_task_flags) == pp_info->xpu_task_flags);
 	pp_info->kvars_deflist = context->kvars_deflist;
-	pp_info->extra_flags = context->extra_flags;
+	pp_info->xpu_task_flags = context->xpu_task_flags;
 	pp_info->extra_bufsz = context->extra_bufsz;
 	pp_info->used_params = context->used_params;
 	pp_info->cuda_stack_size = estimate_cuda_stack_size(context);

src/misc.c

@@ -160,7 +160,6 @@ form_pgstrom_plan_info(CustomScan *cscan, pgstromPlanInfo *pp_info)
 	exprs = lappend(exprs, kvars_deflist_exprs);
 	privs = lappend(privs, makeInteger(pp_info->kvecs_bufsz));
 	privs = lappend(privs, makeInteger(pp_info->kvecs_ndims));
-	privs = lappend(privs, makeInteger(pp_info->extra_flags));
 	privs = lappend(privs, makeInteger(pp_info->extra_bufsz));
 	privs = lappend(privs, makeInteger(pp_info->cuda_stack_size));
 	privs = lappend(privs, pp_info->fallback_tlist);
@@ -272,7 +271,6 @@ deform_pgstrom_plan_info(CustomScan *cscan)
 	}
 	pp_data.kvecs_bufsz = intVal(list_nth(privs, pindex++));
 	pp_data.kvecs_ndims = intVal(list_nth(privs, pindex++));
-	pp_data.extra_flags = intVal(list_nth(privs, pindex++));
 	pp_data.extra_bufsz = intVal(list_nth(privs, pindex++));
 	pp_data.cuda_stack_size = intVal(list_nth(privs, pindex++));
 	pp_data.fallback_tlist = list_nth(privs, pindex++);

src/pg_strom.h

@@ -309,7 +309,6 @@ typedef struct
 	List	   *kvars_deflist;
 	uint32_t	kvecs_bufsz;	/* unit size of vectorized kernel values */
 	uint32_t	kvecs_ndims;
-	uint32_t	extra_flags;
 	uint32_t	extra_bufsz;
 	uint32_t	cuda_stack_size;/* estimated stack consumption */
 	/* fallback projection */
@@ -571,8 +570,7 @@ typedef struct
 	Expr	   *top_expr;
 	PlannerInfo *root;
 	List	   *used_params;
-	uint32_t	required_flags;
-	uint32_t	extra_flags;
+	uint32_t	xpu_task_flags;
 	uint32_t	extra_bufsz;
 	uint32_t	device_cost;
 	uint32_t	kexp_flags;
@@ -619,15 +617,6 @@ extern void		codegen_build_packed_kvars_load(codegen_context *context,
 												pgstromPlanInfo *pp_info);
 extern void		codegen_build_packed_kvars_move(codegen_context *context,
 												pgstromPlanInfo *pp_info);
-
-extern void		codegen_build_packed_xpucode(bytea **p_xpucode,
-											 List *exprs_list,
-											 bool inject_hash_value,
-											 List *input_rels_tlist,
-											 uint32_t *p_extra_flags,
-											 uint32_t *p_extra_bufsz,
-											 uint32_t *p_kvars_nslots,
-											 List **p_used_params);
 extern bool		pgstrom_xpu_expression(Expr *expr,
 									   uint32_t required_xpu_flags,
 									   Index scan_relid,

src/xpu_common.h

@@ -1976,8 +1976,6 @@ typedef struct
 												 * no locale configuration */
 #define DEVKERN__SESSION_TIMEZONE	0x00000200U	/* Device function needs session
 												 * timezone */
-#define DEVTYPE__USE_KVARS_SLOTBUF	0x00000400U	/* Device type uses extra buffer on
-												 * the kvars-slot for LoadVars */
 #define DEVTYPE__HAS_COMPARE		0x00000800U	/* Device type has compare handler */
 #define DEVTASK__SCAN				0x10000000U	/* xPU-Scan */
 #define DEVTASK__JOIN				0x20000000U	/* xPU-Join */

src/xpu_opcodes.h

@@ -24,7 +24,7 @@ TYPE_OPCODE(int8, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(float2, "pg_strom", DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(float4, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(float8, NULL, DEVTYPE__HAS_COMPARE)
-TYPE_OPCODE(numeric,NULL, DEVTYPE__HAS_COMPARE | DEVTYPE__USE_KVARS_SLOTBUF)
+TYPE_OPCODE(numeric,NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(bytea, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(text,  NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(bpchar, NULL, DEVTYPE__HAS_COMPARE)
@@ -33,13 +33,13 @@ TYPE_OPCODE(time, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(timetz, NULL, 0)
 TYPE_OPCODE(timestamp,   NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(timestamptz, NULL, DEVTYPE__HAS_COMPARE)
-TYPE_OPCODE(interval, NULL, DEVTYPE__HAS_COMPARE | DEVTYPE__USE_KVARS_SLOTBUF)
+TYPE_OPCODE(interval, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(money, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(uuid, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(macaddr, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(inet, NULL, DEVTYPE__HAS_COMPARE)
 TYPE_OPCODE(jsonb, NULL, 0)
-TYPE_OPCODE(geometry, "postgis", DEVTYPE__USE_KVARS_SLOTBUF)
+TYPE_OPCODE(geometry, "postgis", 0)
 TYPE_OPCODE(box2df, "postgis", 0)
 TYPE_OPCODE(cube, "cube", 0)