From b6b91b6eba0b5bbac0956032e965d294bf84265d Mon Sep 17 00:00:00 2001
From: Sven Verdoolaege <skimo@kotnet.org>
Date: Sun, 20 Oct 2013 17:09:03 +0200
Subject: [PATCH] gpu.c: create_domain_leaf: handle domain leaves not inside
 any kernel

In the next commit, we will introduce the possibility of mapping
only part of the input scop to the device.  We therefore need to
adjust create_domain_leaf to take into account that some user
statements may be generated on the host, outside any kernel.

Signed-off-by: Sven Verdoolaege <skimo@kotnet.org>
---
 gpu.c | 29 +++++++++++++++++++++++------
 1 file changed, 23 insertions(+), 6 deletions(-)

diff --git a/gpu.c b/gpu.c
index b9cfe64..71f3ecf 100644
--- a/gpu.c
+++ b/gpu.c
@@ -1381,12 +1381,14 @@ static int find_array_index(struct ppcg_kernel *kernel, const char *name)
 /* Internal data structure for the index and AST expression transformation
  * callbacks for pet_stmt_build_ast_exprs.
  *
- * "kernel" is the kernel for which are computing AST expressions.
+ * "kernel" is the kernel for which are computing AST expressions and
+ * may be NULL if we are not inside a kernel.
  * "accesses" is the list of gpu_stmt_access in the statement.
  * "iterator_map" expresses the statement iterators in terms of
  * the AST loop iterators.
  * "sched2shared" expresses the outer shared_schedule_dim dimensions of
- * the kernel schedule in terms of the AST loop iterators.
+ * the kernel schedule in terms of the AST loop iterators and
+ * may be NULL if we are not inside a kernel.
  *
  * The following fields are set in transform_index and used in transform_expr.
  * "array" is the array that is being accessed.
@@ -1448,7 +1450,9 @@ static struct gpu_array_ref_group *find_ref_group(
  *
  * We first reformulate "index" in terms of the AST loop iterators.
  * Then we check if we are accessing the global array or
- * a shared/private copy.  In the former case, we simply return
+ * a shared/private copy.  In particular, if we are not inside a kernel
+ * then we must be accessing a global array.
+ * In the former case, we simply return
  * the updated index.  If "index" is an affine expression rather
  * than an array access, then we also return the updated index here.
  *
@@ -1502,6 +1506,9 @@ static __isl_give isl_multi_pw_aff *transform_index(
 	iterator_map = isl_pw_multi_aff_copy(data->iterator_map);
 	index = isl_multi_pw_aff_pullback_pw_multi_aff(index, iterator_map);
 
+	if (!data->kernel)
+		return index;
+
 	access = find_access(data->accesses, ref_id);
 	if (!access)
 		return index;
@@ -1728,6 +1735,7 @@ static __isl_give isl_ast_expr *transform_expr(__isl_take isl_ast_expr *expr,
 
 /* This function is called for each instance of a user statement
  * in the kernel "kernel", identified by "gpu_stmt".
+ * "kernel" may be NULL if we are not inside a kernel.
  *
  * We attach a struct ppcg_kernel_stmt to the "node", containing
  * a computed AST expression for each access.
@@ -1743,21 +1751,29 @@ static __isl_give isl_ast_node *create_domain_leaf(
 {
 	struct ppcg_transform_data data;
 	struct ppcg_kernel_stmt *stmt;
+	isl_ctx *ctx;
 	isl_id *id;
 	isl_pw_multi_aff *sched2shared;
 	isl_map *map;
 	isl_pw_multi_aff *iterator_map;
 	isl_union_map *schedule;
 
-	stmt = isl_calloc_type(kernel->ctx, struct ppcg_kernel_stmt);
+	if (!node)
+		return NULL;
+	ctx = isl_ast_node_get_ctx(node);
+
+	stmt = isl_calloc_type(ctx, struct ppcg_kernel_stmt);
 	if (!stmt)
 		return isl_ast_node_free(node);
 
 	schedule = isl_ast_build_get_schedule(build);
 	map = isl_map_reverse(isl_map_from_union_map(schedule));
 	iterator_map = isl_pw_multi_aff_from_map(map);
-	sched2shared = compute_sched_to_shared(kernel,
+	if (kernel)
+		sched2shared = compute_sched_to_shared(kernel,
 					isl_pw_multi_aff_copy(iterator_map));
+	else
+		sched2shared = NULL;
 
 	stmt->type = ppcg_kernel_domain;
 	stmt->u.d.stmt = gpu_stmt;
@@ -1773,7 +1789,7 @@ static __isl_give isl_ast_node *create_domain_leaf(
 	isl_pw_multi_aff_free(iterator_map);
 	isl_pw_multi_aff_free(sched2shared);
 
-	id = isl_id_alloc(kernel->ctx, NULL, stmt);
+	id = isl_id_alloc(ctx, NULL, stmt);
 	id = isl_id_set_free_user(id, &ppcg_kernel_stmt_free);
 	return isl_ast_node_set_annotation(node, id);
 }
@@ -1884,6 +1900,7 @@ static __isl_give isl_ast_node *create_sync_leaf(
  * "prog" represents the entire scop.
  * "kernel" points to the kernel to which the current schedule node
  * belongs.  It is set by before_mark and reset by after_mark.
+ * It may be NULL if we are outside any kernel.
  */
 struct ppcg_at_domain_data {
 	struct gpu_prog *prog;
-- 
2.11.4.GIT