re PR tree-optimization/81082 (Failure to vectorise after reassociating index computation)

2018-01-26  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/81082
	* fold-const.c (fold_plusminus_mult_expr): Do not perform the
	association if it requires casting to unsigned.
	* match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
	from fold_plusminus_mult_expr to catch important cases late when
	range info is available.

	* gcc.dg/vect/pr81082.c: New testcase.
	* gcc.dg/tree-ssa/loop-15.c: XFAIL the (int)((unsigned)n + -1U) * n + n
	simplification to n * n.

From-SVN: r257077

gcc/ChangeLog

+2018-01-26  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/81082
+	* fold-const.c (fold_plusminus_mult_expr): Do not perform the
+	association if it requires casting to unsigned.
+	* match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
+	from fold_plusminus_mult_expr to catch important cases late when
+	range info is available.
+
 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 
 	* config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.

gcc/fold-const.c

@@ -7097,7 +7097,7 @@ fold_plusminus_mult_expr (location_t loc, enum tree_code code, tree type,
 
   /* Same may be zero and thus the operation 'code' may overflow.  Likewise
      same may be minus one and thus the multiplication may overflow.  Perform
-     the operations in an unsigned type.  */
+     the sum operation in an unsigned type.  */
   tree utype = unsigned_type_for (type);
   tree tem = fold_build2_loc (loc, code, utype,
 			      fold_convert_loc (loc, utype, alt0),
@@ -7110,9 +7110,9 @@ fold_plusminus_mult_expr (location_t loc, enum tree_code code, tree type,
     return fold_build2_loc (loc, MULT_EXPR, type,
 			    fold_convert (type, tem), same);
 
-  return fold_convert_loc (loc, type,
-			   fold_build2_loc (loc, MULT_EXPR, utype, tem,
-					    fold_convert_loc (loc, utype, same)));
+  /* Do not resort to unsigned multiplication because
+     we lose the no-overflow property of the expression.  */
+  return NULL_TREE;
 }
 
 /* Subroutine of native_encode_expr.  Encode the INTEGER_CST

gcc/match.pd

@@ -1939,6 +1939,37 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
      (minus (convert (view_convert:stype @1))
 	    (convert (view_convert:stype @2)))))))
 
+/* (A * C) +- (B * C) -> (A+-B) * C and (A * C) +- A -> A * (C+-1).
+    Modeled after fold_plusminus_mult_expr.  */
+(if (!TYPE_SATURATING (type)
+     && (!FLOAT_TYPE_P (type) || flag_associative_math))
+ (for plusminus (plus minus)
+  (simplify
+   (plusminus (mult:cs @0 @1) (mult:cs @0 @2))
+   (if (!ANY_INTEGRAL_TYPE_P (type)
+        || TYPE_OVERFLOW_WRAPS (type)
+        || (INTEGRAL_TYPE_P (type)
+	    && tree_expr_nonzero_p (@0)
+	    && expr_not_equal_to (@0, wi::minus_one (TYPE_PRECISION (type)))))
+    (mult (plusminus @1 @2) @0)))
+  /* We cannot generate constant 1 for fract.  */
+  (if (!ALL_FRACT_MODE_P (TYPE_MODE (type)))
+   (simplify
+    (plusminus @0 (mult:cs @0 @2))
+    (if (!ANY_INTEGRAL_TYPE_P (type)
+	 || TYPE_OVERFLOW_WRAPS (type)
+	 || (INTEGRAL_TYPE_P (type)
+	     && tree_expr_nonzero_p (@0)
+	     && expr_not_equal_to (@0, wi::minus_one (TYPE_PRECISION (type)))))
+     (mult (plusminus { build_one_cst (type); } @2) @0)))
+   (simplify
+    (plusminus (mult:cs @0 @2) @0)
+    (if (!ANY_INTEGRAL_TYPE_P (type)
+	 || TYPE_OVERFLOW_WRAPS (type)
+	 || (INTEGRAL_TYPE_P (type)
+	     && tree_expr_nonzero_p (@0)
+	     && expr_not_equal_to (@0, wi::minus_one (TYPE_PRECISION (type)))))
+     (mult (plusminus @2 { build_one_cst (type); }) @0))))))
 
 /* Simplifications of MIN_EXPR, MAX_EXPR, fmin() and fmax().  */
 

gcc/testsuite/ChangeLog

+2018-01-26  Richard Biener  <rguenther@suse.de>
+
+	PR tree-optimization/81082
+	* gcc.dg/vect/pr81082.c: New testcase.
+	* gcc.dg/tree-ssa/loop-15.c: XFAIL the (int)((unsigned)n + -1U) * n + n
+	simplification to n * n.
+
 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
 
 	* gcc.target/i386/mcount_pic.c: Only xfail get_pc_thunk scan on

gcc/testsuite/gcc.dg/tree-ssa/loop-15.c

@@ -19,7 +19,7 @@ int bla(void)
 }
 
 /* Since the loop is removed, there should be no addition.  */
-/* { dg-final { scan-tree-dump-times " \\+ " 0 "optimized" } } */
+/* { dg-final { scan-tree-dump-times " \\+ " 0 "optimized" { xfail *-*-* } } } */
 /* { dg-final { scan-tree-dump-times " \\* " 1 "optimized" } } */
 
 /* The if from the loop header copying remains in the code.  */

gcc/testsuite/gcc.dg/vect/pr81082.c

+/* { dg-do compile } */
+/* { dg-require-effective-target vect_int } */
+
+int
+f (int *x, int b1, int b2, int b3)
+{
+  int foo = 0;
+  for (int i1 = 0; i1 < b1; ++i1)
+    for (int i2 = 0; i2 < b2; ++i2)
+      for (int i3 = 0; i3 < b3; ++i3)
+	foo += x[i1 * b2 * b3 + i2 * b3 + (i3 - 1)];
+  return foo;
+}
+
+/* { dg-final { scan-tree-dump "vectorized 1 loops in function" "vect" } } */