tree-optimization/107302 - fix vec_perm placement for recurrence vect

Message ID 20221018104758.20724139D2@imap2.suse-dmz.suse.de
State Repeat Merge
Headers
Series tree-optimization/107302 - fix vec_perm placement for recurrence vect |

Checks

Context Check Description
snail/gcc-patch-check warning Git am fail log

Commit Message

Richard Biener Oct. 18, 2022, 10:47 a.m. UTC
  The following fixes the VEC_PERM_EXPR placement when the latch
definition is a PHI node.

Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed.

	PR tree-optimization/107302
	* tree-vect-loop.cc (vectorizable_recurrence): Fix vec_perm
	placement for a PHI latch def.

	* gcc.dg/vect/pr107302.c: New testcase.
---
 gcc/testsuite/gcc.dg/vect/pr107302.c | 13 +++++++++++++
 gcc/tree-vect-loop.cc                | 12 +++++++++---
 2 files changed, 22 insertions(+), 3 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/vect/pr107302.c
  

Patch

diff --git a/gcc/testsuite/gcc.dg/vect/pr107302.c b/gcc/testsuite/gcc.dg/vect/pr107302.c
new file mode 100644
index 00000000000..293f7e4067a
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr107302.c
@@ -0,0 +1,13 @@ 
+/* { dg-do compile } */
+/* { dg-additional-options "-fno-tree-pre" } */
+
+int a[2000];
+int s292_im1;
+
+void
+s292() {
+  for (int i = 0; i < 2000; i++) {
+    a[i] = s292_im1;
+    s292_im1 = i;
+  }
+}
diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
index 63e86540d12..92790bd8095 100644
--- a/gcc/tree-vect-loop.cc
+++ b/gcc/tree-vect-loop.cc
@@ -8485,9 +8485,15 @@  vectorizable_recurr (loop_vec_info loop_vinfo, stmt_vec_info stmt_info,
      second and later operands are tentative and will be updated when we have
      vectorized the latch definition.  */
   edge le = loop_latch_edge (LOOP_VINFO_LOOP (loop_vinfo));
-  gimple_stmt_iterator gsi2
-    = gsi_for_stmt (SSA_NAME_DEF_STMT (PHI_ARG_DEF_FROM_EDGE (phi, le)));
-  gsi_next (&gsi2);
+  gimple *latch_def = SSA_NAME_DEF_STMT (PHI_ARG_DEF_FROM_EDGE (phi, le));
+  gimple_stmt_iterator gsi2;
+  if (is_a <gphi *> (latch_def))
+    gsi2 = gsi_after_labels (gimple_bb (latch_def));
+  else
+    {
+      gsi2 = gsi_for_stmt (latch_def);
+      gsi_next (&gsi2);
+    }
 
   for (unsigned i = 0; i < ncopies; ++i)
     {