re PR tree-optimization/69157 (ICE in vect_transform_stmt, at tree-vect-stmts.c:8176)

2016-01-12  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/69157
	* tree-vect-stmts.c (vectorizable_mask_load_store): Check
	stmts def type only during analyze phase.
	(vectorizable_call): Likewise.
	(vectorizable_simd_clone_call): Likewise.
	(vectorizable_conversion): Likewise.
	(vectorizable_assignment): Likewise.
	(vectorizable_shift): Likewise.
	(vectorizable_operation): Likewise.
	(vectorizable_store): Likewise.
	(vectorizable_load): Likewise.

	* gcc.dg/torture/pr69157.c: New testcase.

2016-01-12  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/69174
	* tree-vect-stmts.c (vect_mark_relevant): Remove excessive vertical
	space.
	(vectorizable_load): Properly compute the number of loads needed
	for permuted strided SLP loads and do not spuriously assign
	to SLP_TREE_VEC_STMTS.

	* gcc.dg/torture/pr69174.c: New testcase.

From-SVN: r232260

5 files changed, 103 insertions, 14 deletions

diff --git a/gcc/ChangeLog b/gcc/ChangeLog
index f8784f0..6ccdbf8 100644
--- a/gcc/ChangeLog
+++ b/gcc/ChangeLog
@@ -1,3 +1,26 @@
+2016-01-12  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/69157
+	* tree-vect-stmts.c (vectorizable_mask_load_store): Check
+	stmts def type only during analyze phase.
+	(vectorizable_call): Likewise.
+	(vectorizable_simd_clone_call): Likewise.
+	(vectorizable_conversion): Likewise.
+	(vectorizable_assignment): Likewise.
+	(vectorizable_shift): Likewise.
+	(vectorizable_operation): Likewise.
+	(vectorizable_store): Likewise.
+	(vectorizable_load): Likewise.
+
+2016-01-12  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/69174
+	* tree-vect-stmts.c (vect_mark_relevant): Remove excessive vertical
+	space.
+	(vectorizable_load): Properly compute the number of loads needed
+	for permuted strided SLP loads and do not spuriously assign
+	to SLP_TREE_VEC_STMTS.
+
 2016-01-12  Andris Pavenis <andris.pavenis@iki.fi>
 
 	* config/i386/djgpp.h (PREFERRED_DEBUGGING_TYPE): Define to DWARF2
diff --git a/gcc/testsuite/ChangeLog b/gcc/testsuite/ChangeLog
index 1866304..e19ea31 100644
--- a/gcc/testsuite/ChangeLog
+++ b/gcc/testsuite/ChangeLog
@@ -1,3 +1,13 @@
+2016-01-12  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/69157
+	* gcc.dg/torture/pr69157.c: New testcase.
+
+2016-01-12  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/69174
+	* gcc.dg/torture/pr69174.c: New testcase.
+
 2016-01-12  Jakub Jelinek  <jakub@redhat.com>
 
 	PR c++/66808
diff --git a/gcc/testsuite/gcc.dg/torture/pr69157.c b/gcc/testsuite/gcc.dg/torture/pr69157.c
new file mode 100644
index 0000000..e85082e
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/pr69157.c
@@ -0,0 +1,17 @@
+/* { dg-do compile } */
+
+typedef struct {
+    float *data_normal3f;
+    float *data_texcoordtexture2f;
+    float *data_texcoordlightmap2f;
+    float *data_color4f;
+} dp_model_t;
+dp_model_t a;
+float *b;
+void fn1() {
+    int c;
+    a.data_normal3f = b + c * 3;
+    a.data_texcoordtexture2f = a.data_normal3f + c * 3;
+    a.data_texcoordlightmap2f = a.data_color4f =
+	a.data_texcoordlightmap2f + c * 2;
+}
diff --git a/gcc/testsuite/gcc.dg/torture/pr69174.c b/gcc/testsuite/gcc.dg/torture/pr69174.c
new file mode 100644
index 0000000..0866331
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/pr69174.c
@@ -0,0 +1,19 @@
+/* { dg-do compile } */
+
+typedef int pixval;
+typedef struct { pixval r, g, b; } xel;
+int convertRow_sample, convertRaster_col;
+short *convertRow_samplebuf;
+xel *convertRow_xelrow;
+short convertRow_spp;
+void fn1() {
+    int *alpharow;
+    for (; convertRaster_col;
+	 ++convertRaster_col, convertRow_sample += convertRow_spp) {
+	convertRow_xelrow[convertRaster_col].r =
+	    convertRow_xelrow[convertRaster_col].g =
+	    convertRow_xelrow[convertRaster_col].b =
+	    convertRow_samplebuf[convertRow_sample];
+	alpharow[convertRaster_col] = convertRow_samplebuf[convertRow_sample + 3];
+    }
+}
diff --git a/gcc/tree-vect-stmts.c b/gcc/tree-vect-stmts.c
index 465826e..872fa07 100644
--- a/gcc/tree-vect-stmts.c
+++ b/gcc/tree-vect-stmts.c
@@ -190,8 +190,11 @@ vect_mark_relevant (vec<gimple *> *worklist, gimple *stmt,
   gimple *pattern_stmt;
 
   if (dump_enabled_p ())
-    dump_printf_loc (MSG_NOTE, vect_location,
-                     "mark relevant %d, live %d.\n", relevant, live_p);
+    {
+      dump_printf_loc (MSG_NOTE, vect_location,
+		       "mark relevant %d, live %d: ", relevant, live_p);
+      dump_gimple_stmt (MSG_NOTE, TDF_SLIM, stmt, 0);
+    }
 
   /* If this stmt is an original stmt in a pattern, we might need to mark its
      related pattern stmt instead of the original stmt.  However, such stmts
@@ -1757,7 +1760,8 @@ vectorizable_mask_load_store (gimple *stmt, gimple_stmt_iterator *gsi,
   if (!STMT_VINFO_RELEVANT_P (stmt_info))
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   if (!STMT_VINFO_DATA_REF (stmt_info))
@@ -2206,7 +2210,8 @@ vectorizable_call (gimple *gs, gimple_stmt_iterator *gsi, gimple **vec_stmt,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   /* Is GS a vectorizable call?   */
@@ -2811,7 +2816,8 @@ vectorizable_simd_clone_call (gimple *stmt, gimple_stmt_iterator *gsi,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   if (gimple_call_lhs (stmt)
@@ -3669,7 +3675,8 @@ vectorizable_conversion (gimple *stmt, gimple_stmt_iterator *gsi,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   if (!is_gimple_assign (stmt))
@@ -4246,7 +4253,8 @@ vectorizable_assignment (gimple *stmt, gimple_stmt_iterator *gsi,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   /* Is vectorizable assignment?  */
@@ -4462,7 +4470,8 @@ vectorizable_shift (gimple *stmt, gimple_stmt_iterator *gsi,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   /* Is STMT a vectorizable binary/unary operation?   */
@@ -4823,7 +4832,8 @@ vectorizable_operation (gimple *stmt, gimple_stmt_iterator *gsi,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   /* Is STMT a vectorizable binary/unary operation?   */
@@ -5248,7 +5258,8 @@ vectorizable_store (gimple *stmt, gimple_stmt_iterator *gsi, gimple **vec_stmt,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   /* Is vectorizable store? */
@@ -6237,7 +6248,8 @@ vectorizable_load (gimple *stmt, gimple_stmt_iterator *gsi, gimple **vec_stmt,
   if (!STMT_VINFO_RELEVANT_P (stmt_info) && !bb_vinfo)
     return false;
 
-  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def)
+  if (STMT_VINFO_DEF_TYPE (stmt_info) != vect_internal_def
+      && ! vec_stmt)
     return false;
 
   /* Is vectorizable load? */
@@ -6748,9 +6760,16 @@ vectorizable_load (gimple *stmt, gimple_stmt_iterator *gsi, gimple **vec_stmt,
 	  else
 	    ltype = vectype;
 	  ltype = build_aligned_type (ltype, TYPE_ALIGN (TREE_TYPE (vectype)));
-	  ncopies = SLP_TREE_NUMBER_OF_VEC_STMTS (slp_node);
+	  /* For SLP permutation support we need to load the whole group,
+	     not only the number of vector stmts the permutation result
+	     fits in.  */
 	  if (slp_perm)
-	    dr_chain.create (ncopies);
+	    {
+	      ncopies = (group_size * vf + nunits - 1) / nunits;
+	      dr_chain.create (ncopies);
+	    }
+	  else
+	    ncopies = SLP_TREE_NUMBER_OF_VEC_STMTS (slp_node);
 	}
       for (j = 0; j < ncopies; j++)
 	{
@@ -6798,9 +6817,10 @@ vectorizable_load (gimple *stmt, gimple_stmt_iterator *gsi, gimple **vec_stmt,
 
 	  if (slp)
 	    {
-	      SLP_TREE_VEC_STMTS (slp_node).quick_push (new_stmt);
 	      if (slp_perm)
 		dr_chain.quick_push (gimple_assign_lhs (new_stmt));
+	      else
+		SLP_TREE_VEC_STMTS (slp_node).quick_push (new_stmt);
 	    }
 	  else
 	    {