tree-optimization/112505 - bit-precision induction vectorization

Message ID 20240111141328.AB719386186A@sourceware.org
State Accepted
Headers
Series tree-optimization/112505 - bit-precision induction vectorization |

Checks

Context Check Description
snail/gcc-patch-check success Github commit url

Commit Message

Richard Biener Jan. 11, 2024, 2:07 p.m. UTC
  Vectorization of bit-precision inductions isn't implemented but we
don't check this, instead we ICE during transform.

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

	PR tree-optimization/112505
	* tree-vect-loop.cc (vectorizable_induction): Reject
	bit-precision induction.

	* gcc.dg/vect/pr112505.c: New testcase.
---
 gcc/testsuite/gcc.dg/vect/pr112505.c | 14 ++++++++++++++
 gcc/tree-vect-loop.cc                |  9 +++++++++
 2 files changed, 23 insertions(+)
 create mode 100644 gcc/testsuite/gcc.dg/vect/pr112505.c
  

Patch

diff --git a/gcc/testsuite/gcc.dg/vect/pr112505.c b/gcc/testsuite/gcc.dg/vect/pr112505.c
new file mode 100644
index 00000000000..56546c1095a
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/vect/pr112505.c
@@ -0,0 +1,14 @@ 
+/* { dg-do compile } */
+/* { dg-additional-options "-O3" } */
+
+short int w9;
+struct T {
+  short a : 14;
+  int b;
+};
+struct T v;
+void zc()
+{
+  for(int i = 0; i < 4; i ++)
+    w9 *= v.b ? v.a-- < 0 : 0;
+}
diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
index 1bdad0fbe0f..38bd8267ee1 100644
--- a/gcc/tree-vect-loop.cc
+++ b/gcc/tree-vect-loop.cc
@@ -10037,6 +10037,15 @@  vectorizable_induction (loop_vec_info loop_vinfo,
 
   step_expr = STMT_VINFO_LOOP_PHI_EVOLUTION_PART (stmt_info);
   gcc_assert (step_expr != NULL_TREE);
+  if (INTEGRAL_TYPE_P (TREE_TYPE (step_expr))
+      && !type_has_mode_precision_p (TREE_TYPE (step_expr)))
+    {
+      if (dump_enabled_p ())
+	dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
+			 "bit-precision induction vectorization not "
+			 "supported.\n");
+      return false;
+    }
   tree step_vectype = get_same_sized_vectype (TREE_TYPE (step_expr), vectype);
 
   /* Check for backend support of PLUS/MINUS_EXPR. */