public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r12-7050] Adjust LSHIFT_EXPR handling of multiple_of_p
@ 2022-02-04 11:54 Richard Biener
  0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2022-02-04 11:54 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:24c72fb0eedfe7c67f9b15aa31b474a999cc4723

commit r12-7050-g24c72fb0eedfe7c67f9b15aa31b474a999cc4723
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Jan 24 15:26:46 2022 +0100

    Adjust LSHIFT_EXPR handling of multiple_of_p
    
    This removes the odd check of size_type_node when handling left-shifts
    as multiplications of 1 << N and instead uses the type as specified.
    It also moves left-shift handling next to multiplications where it
    semantically belongs.
    
    2022-01-24  Richard Biener  <rguenther@suse.de>
    
            * fold-const.cc (multiple_of_p): Re-write and move LSHIFT_EXPR
            handling.

Diff:
---
 gcc/fold-const.cc | 33 ++++++++++++++++-----------------
 1 file changed, 16 insertions(+), 17 deletions(-)

diff --git a/gcc/fold-const.cc b/gcc/fold-const.cc
index 8fc01cdfb77..12732d39c79 100644
--- a/gcc/fold-const.cc
+++ b/gcc/fold-const.cc
@@ -14079,7 +14079,7 @@ int
 multiple_of_p (tree type, const_tree top, const_tree bottom)
 {
   gimple *stmt;
-  tree t1, op1, op2;
+  tree op1, op2;
 
   if (operand_equal_p (top, bottom, 0))
     return 1;
@@ -14125,6 +14125,21 @@ multiple_of_p (tree type, const_tree top, const_tree bottom)
       return (multiple_of_p (type, TREE_OPERAND (top, 1), bottom)
 	      || multiple_of_p (type, TREE_OPERAND (top, 0), bottom));
 
+    case LSHIFT_EXPR:
+      /* Handle X << CST as X * (1 << CST) and only process the constant.  */
+      if (TREE_CODE (TREE_OPERAND (top, 1)) == INTEGER_CST)
+	{
+	  op1 = TREE_OPERAND (top, 1);
+	  if (wi::to_widest (op1) < TYPE_PRECISION (type))
+	    {
+	      wide_int mul_op
+		= wi::one (TYPE_PRECISION (type)) << wi::to_wide (op1);
+	      return multiple_of_p (type,
+				    wide_int_to_tree (type, mul_op), bottom);
+	    }
+	}
+      return 0;
+
     case MINUS_EXPR:
       /* It is impossible to prove if op0 - op1 is multiple of bottom
 	 precisely, so be conservative here checking if both op0 and op1
@@ -14144,22 +14159,6 @@ multiple_of_p (tree type, const_tree top, const_tree bottom)
       return (multiple_of_p (type, op1, bottom)
 	      && multiple_of_p (type, TREE_OPERAND (top, 0), bottom));
 
-    case LSHIFT_EXPR:
-      if (TREE_CODE (TREE_OPERAND (top, 1)) == INTEGER_CST)
-	{
-	  op1 = TREE_OPERAND (top, 1);
-	  /* const_binop may not detect overflow correctly,
-	     so check for it explicitly here.  */
-	  if (wi::gtu_p (TYPE_PRECISION (TREE_TYPE (size_one_node)),
-			 wi::to_wide (op1))
-	      && (t1 = fold_convert (type,
-				     const_binop (LSHIFT_EXPR, size_one_node,
-						  op1))) != 0
-	      && !TREE_OVERFLOW (t1))
-	    return multiple_of_p (type, t1, bottom);
-	}
-      return 0;
-
     CASE_CONVERT:
       /* Can't handle conversions from non-integral or wider integral type.  */
       if ((TREE_CODE (TREE_TYPE (TREE_OPERAND (top, 0))) != INTEGER_TYPE)


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2022-02-04 11:54 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-02-04 11:54 [gcc r12-7050] Adjust LSHIFT_EXPR handling of multiple_of_p Richard Biener

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).