Patchwork Fix PR tree-optimization/44152 - ICE in SLP vectorization

login
register
mail settings
Submitter Ira Rosen
Date July 27, 2010, 12:27 p.m.
Message ID <OFB119A189.FBAD0584-ONC225776D.003C540D-C225776D.004477A9@il.ibm.com>
Download mbox | patch
Permalink /patch/59999/
State New
Headers show

Comments

Ira Rosen - July 27, 2010, 12:27 p.m.
Hi,

This patch adds missing checks for complex numbers in SLP vectorization.

Bootstrapped and tested on x86_64-suse-linux.
Committed revision 162564.

Ira

ChangeLog:

	PR tree-optimization/44152
	* tree-vect-slp.c (vect_build_slp_tree): Collect nodes with
	complex numbers for further check.
	(vect_supported_load_permutation_p): Check nodes with
	complex numbers.

testsuite/ChangeLog:

	PR tree-optimization/44152
	* gcc.dg/vect/fast-math-pr44152.c: New test.

+/* { dg-final { cleanup-tree-dump "vect" } } */

Patch

Index: tree-vect-slp.c
===================================================================
--- tree-vect-slp.c     (revision 162526)
+++ tree-vect-slp.c     (working copy)
@@ -318,7 +318,7 @@  vect_build_slp_tree (loop_vec_info loop_
   gimple stmt = VEC_index (gimple, stmts, 0);
   enum vect_def_type first_stmt_dt0 = vect_uninitialized_def;
   enum vect_def_type first_stmt_dt1 = vect_uninitialized_def;
-  enum tree_code first_stmt_code = ERROR_MARK, rhs_code;
+  enum tree_code first_stmt_code = ERROR_MARK, rhs_code = ERROR_MARK;
   tree first_stmt_def1_type = NULL_TREE, first_stmt_def0_type = NULL_TREE;
   tree lhs;
   bool stop_recursion = false, need_same_oprnds = false;
@@ -648,6 +648,13 @@  vect_build_slp_tree (loop_vec_info loop_
             += targetm.vectorize.builtin_vectorization_cost (vec_perm,
NULL, 0)
                * group_size;
         }
+      else
+        {
+          /* We don't check here complex numbers chains, so we keep them
in
+            LOADS for further check in vect_supported_load_permutation_p.
*/
+          if (rhs_code == REALPART_EXPR || rhs_code == IMAGPART_EXPR)
+            VEC_safe_push (slp_tree, heap, *loads, *node);
+        }

       return true;
     }
@@ -868,8 +875,9 @@  vect_supported_load_permutation_p (slp_i
   int i = 0, j, prev = -1, next, k, number_of_groups;
   bool supported, bad_permutation = false;
   sbitmap load_index;
-  slp_tree node;
-  gimple stmt;
+  slp_tree node, other_complex_node;
+  gimple stmt, first = NULL, other_node_first;
+  unsigned complex_numbers = 0;

   /* FORNOW: permutations are only supported in SLP.  */
   if (!slp_instn)
@@ -893,10 +901,71 @@  vect_supported_load_permutation_p (slp_i
   for (i = 0;
        VEC_iterate (slp_tree, SLP_INSTANCE_LOADS (slp_instn), i, node);
        i++)
-    if (VEC_length (gimple, SLP_TREE_SCALAR_STMTS (node))
-        != (unsigned) group_size)
-      return false;
-
+    {
+      if (VEC_length (gimple, SLP_TREE_SCALAR_STMTS (node))
+          != (unsigned) group_size)
+        return false;
+
+      stmt = VEC_index (gimple, SLP_TREE_SCALAR_STMTS (node), 0);
+      if (is_gimple_assign (stmt)
+          && (gimple_assign_rhs_code (stmt) == REALPART_EXPR
+              || gimple_assign_rhs_code (stmt) == IMAGPART_EXPR))
+        complex_numbers++;
+    }
+
+  /* Complex operands can be swapped as following:
+      real_c = real_b + real_a;
+      imag_c = imag_a + imag_b;
+     i.e., we have {real_b, imag_a} and {real_a, imag_b} instead of
+     {real_a, imag_a} and {real_b, imag_b}. We check here that if
interleaving
+     chains are mixed, they match the above pattern.  */
+  if (complex_numbers)
+    {
+      for (i = 0;
+           VEC_iterate (slp_tree, SLP_INSTANCE_LOADS (slp_instn), i,
node);
+           i++)
+        {
+          for (j = 0;
+               VEC_iterate (gimple, SLP_TREE_SCALAR_STMTS (node), j,
stmt);
+               j++)
+            {
+              if (j == 0)
+                first = stmt;
+              else
+                {
+                  if (DR_GROUP_FIRST_DR (vinfo_for_stmt (stmt)) != first)
+                    {
+                      if (complex_numbers != 2)
+                        return false;
+
+                      if (i == 0)
+                        k = 1;
+                      else
+                        k = 0;
+
+                      other_complex_node = VEC_index (slp_tree,
+                                            SLP_INSTANCE_LOADS
(slp_instn), k);
+                      other_node_first = VEC_index (gimple,
+                                SLP_TREE_SCALAR_STMTS
(other_complex_node), 0);
+
+                      if (DR_GROUP_FIRST_DR (vinfo_for_stmt (stmt))
+                          != other_node_first)
+                       return false;
+                    }
+                }
+            }
+        }
+    }
+
+  /* We checked that this case ok, so there is no need to proceed with
+     permutation tests.  */
+  if (complex_numbers == 2)
+    {
+      VEC_free (slp_tree, heap, SLP_INSTANCE_LOADS (slp_instn));
+      VEC_free (int, heap, SLP_INSTANCE_LOAD_PERMUTATION (slp_instn));
+      return true;
+    }
+
   node = SLP_INSTANCE_TREE (slp_instn);
   stmt = VEC_index (gimple, SLP_TREE_SCALAR_STMTS (node), 0);
   /* LOAD_PERMUTATION is a list of indices of all the loads of the SLP
Index: testsuite/gcc.dg/vect/fast-math-pr44152.c
===================================================================
--- testsuite/gcc.dg/vect/fast-math-pr44152.c   (revision 0)
+++ testsuite/gcc.dg/vect/fast-math-pr44152.c   (revision 0)
@@ -0,0 +1,38 @@ 
+/* { dg-do compile } */
+/* { dg-require-effective-target vect_float } */
+
+#include <stdio.h>
+#include <stdarg.h>
+#include "tree-vect.h"
+
+#define N 16
+
+_Complex float a[N]  __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)))
=
+    { 10.0F + 20.0iF, 11.0F + 21.0iF, 12.0F + 22.0iF, 13.0F + 23.0iF,
+      14.0F + 24.0iF, 15.0F + 25.0iF, 16.0F + 26.0iF, 17.0F + 27.0iF,
+      18.0F + 28.0iF, 19.0F + 29.0iF, 20.0F + 30.0iF, 21.0F + 31.0iF,
+      22.0F + 32.0iF, 23.0F + 33.0iF, 24.0F + 34.0iF, 25.0F + 35.0iF };
+
+_Complex float c[N] __attribute__ ((__aligned__(__BIGGEST_ALIGNMENT__)));
+
+
+__attribute__ ((noinline)) _Complex float
+foo (int x)
+{
+  int i;
+  _Complex float *p = a + x;
+  _Complex float sum = 10.0F + 20.0iF;
+
+  for (i = 0; i < N; i++)
+   {
+     sum += *p;
+     p++;
+   }
+
+  c[0] = sum + 66.0F + 86.0iF;
+
+  return 0;
+}
+
+