https://gcc.gnu.org/g:44d84db11ab724c34a8b1f8c0e06da1cc78439a2

commit r13-8735-g44d84db11ab724c34a8b1f8c0e06da1cc78439a2
Author: Jakub Jelinek <ja...@redhat.com>
Date:   Wed May 8 10:17:32 2024 +0200

    reassoc: Fix up optimize_range_tests_to_bit_test [PR114965]
    
    The optimize_range_tests_to_bit_test optimization normally emits a range
    test first:
              if (entry_test_needed)
                {
                  tem = build_range_check (loc, optype, unshare_expr (exp),
                                           false, lowi, high);
                  if (tem == NULL_TREE || is_gimple_val (tem))
                    continue;
                }
    so during the bit test we already know that exp is in the [lowi, high]
    range, but skips it if we have range info which tells us this isn't
    necessary.
    Also, normally it emits shifts by exp - lowi counter, but has an
    optimization to use just exp counter if the mask isn't a more expensive
    constant in that case and lowi is > 0 and high is smaller than prec.
    
    The following testcase is miscompiled because the two abnormal cases
    are triggered.  The range of exp is [43, 43][48, 48][95, 95], so we on
    64-bit arch decide we don't need the entry test, because 95 - 43 < 64.
    And we also decide to use just exp as counter, because the range test
    tests just for exp == 43 || exp == 48, so high is smaller than 64 too.
    Because 95 is in the exp range, we can't do that, we'd either need to
    do a range test first, i.e.
    if (exp - 43U <= 48U - 43U) if ((1UL << exp) & mask1))
    or need to subtract lowi from the shift counter, i.e.
    if ((1UL << (exp - 43)) & mask2)
    but can't do both unless r.upper_bound () is < prec.
    
    The following patch ensures that.
    
    2024-05-08  Jakub Jelinek  <ja...@redhat.com>
    
            PR tree-optimization/114965
            * tree-ssa-reassoc.cc (optimize_range_tests_to_bit_test): Don't try 
to
            optimize away exp - lowi subtraction from shift count unless entry
            test is emitted or unless r.upper_bound () is smaller than prec.
    
            * gcc.c-torture/execute/pr114965.c: New test.
    
    (cherry picked from commit 9adec2d91e62a479474ae79df5b455fd4b8463ba)

Diff:
---
 gcc/testsuite/gcc.c-torture/execute/pr114965.c | 30 ++++++++++++++++++++++++++
 gcc/tree-ssa-reassoc.cc                        |  3 ++-
 2 files changed, 32 insertions(+), 1 deletion(-)

diff --git a/gcc/testsuite/gcc.c-torture/execute/pr114965.c 
b/gcc/testsuite/gcc.c-torture/execute/pr114965.c
new file mode 100644
index 000000000000..89d68e187015
--- /dev/null
+++ b/gcc/testsuite/gcc.c-torture/execute/pr114965.c
@@ -0,0 +1,30 @@
+/* PR tree-optimization/114965 */
+
+static void
+foo (const char *x)
+{
+
+  char a = '0';
+  while (1)
+    {
+      switch (*x)
+       {
+       case '_':
+       case '+':
+         a = *x;
+         x++;
+         continue;
+       default:
+         break;
+       }
+      break;
+    }
+  if (a == '0' || a == '+')
+    __builtin_abort ();
+}
+
+int
+main ()
+{
+  foo ("_");
+}
diff --git a/gcc/tree-ssa-reassoc.cc b/gcc/tree-ssa-reassoc.cc
index c5020465c2b3..d8f5471951af 100644
--- a/gcc/tree-ssa-reassoc.cc
+++ b/gcc/tree-ssa-reassoc.cc
@@ -3411,7 +3411,8 @@ optimize_range_tests_to_bit_test (enum tree_code opcode, 
int first, int length,
             We can avoid then subtraction of the minimum value, but the
             mask constant could be perhaps more expensive.  */
          if (compare_tree_int (lowi, 0) > 0
-             && compare_tree_int (high, prec) < 0)
+             && compare_tree_int (high, prec) < 0
+             && (entry_test_needed || wi::ltu_p (r.upper_bound (), prec)))
            {
              int cost_diff;
              HOST_WIDE_INT m = tree_to_uhwi (lowi);

Reply via email to