diff mbox series

tree-optimization/105276 - Retain existing range knowledge when prefilling statements.

Message ID 393cebab-d163-4572-afa9-5c97e320b2bc@redhat.com
State New
Headers show
Series tree-optimization/105276 - Retain existing range knowledge when prefilling statements. | expand

Commit Message

Andrew MacLeod April 21, 2022, 3:59 p.m. UTC
When range_of_stmt was adjusted to avoid large recursion depth, I added 
code to precalculate the dependencies without recursion.

This patch adjusted that pre-fill code to intersect the current known 
range with the newly calculated one before setting the global range.  
this Is what range_of_stmt does, and was missed.

Its running thru testing now, but I expect no issues with either 
bootstrap nor testsuite regressions.

Do you want me to check this when testing is complete, or hold off?

Andrew

Comments

Richard Biener April 22, 2022, 5:51 a.m. UTC | #1
On Thu, Apr 21, 2022 at 5:59 PM Andrew MacLeod <amacleod@redhat.com> wrote:
>
> When range_of_stmt was adjusted to avoid large recursion depth, I added
> code to precalculate the dependencies without recursion.
>
> This patch adjusted that pre-fill code to intersect the current known
> range with the newly calculated one before setting the global range.
> this Is what range_of_stmt does, and was missed.
>
> Its running thru testing now, but I expect no issues with either
> bootstrap nor testsuite regressions.
>
> Do you want me to check this when testing is complete, or hold off?

OK if testing looks good.

Richard.

> Andrew
Andrew MacLeod April 25, 2022, 1:59 p.m. UTC | #2
On 4/22/22 01:51, Richard Biener wrote:
> On Thu, Apr 21, 2022 at 5:59 PM Andrew MacLeod <amacleod@redhat.com> wrote:
>> When range_of_stmt was adjusted to avoid large recursion depth, I added
>> code to precalculate the dependencies without recursion.
>>
>> This patch adjusted that pre-fill code to intersect the current known
>> range with the newly calculated one before setting the global range.
>> this Is what range_of_stmt does, and was missed.
>>
>> Its running thru testing now, but I expect no issues with either
>> bootstrap nor testsuite regressions.
>>
>> Do you want me to check this when testing is complete, or hold off?
> OK if testing looks good.
>
> Richard.
>
passed everything fine.  committed.

Andrew.
diff mbox series

Patch

commit 035a365988628794311c2d706d0e20e39c5fd261
Author: Andrew MacLeod <amacleod@redhat.com>
Date:   Thu Apr 21 11:10:30 2022 -0400

    Retain existing range knowledge when prefilling statements.
    
    When range_of_stmt was adjusted to avoid large recursion depth, we need to
    intersect the calculated range whth the any known range to avoid losing
    info.  Range_of_stmt does this, but the new prefill code missed it.
    
            PR tree-optimization/105276
            gcc/
            * gimple-range.cc (gimple_ranger::prefill_stmt_dependencies): Include
            existing global range with calculated value.
            gcc/testsuite/
            * g++.dg/pr105276.C: New.

diff --git a/gcc/gimple-range.cc b/gcc/gimple-range.cc
index f41a989d617..f5e9e77bc71 100644
--- a/gcc/gimple-range.cc
+++ b/gcc/gimple-range.cc
@@ -389,6 +389,10 @@  gimple_ranger::prefill_stmt_dependencies (tree ssa)
 	      // Fold and save the value for NAME.
 	      stmt = SSA_NAME_DEF_STMT (name);
 	      fold_range_internal (r, stmt, name);
+	      // Make sure we don't lose any current global info.
+	      int_range_max tmp;
+	      m_cache.get_global_range (tmp, name);
+	      r.intersect (tmp);
 	      m_cache.set_global_range (name, r);
 	    }
 	  continue;
diff --git a/gcc/testsuite/g++.dg/pr105276.C b/gcc/testsuite/g++.dg/pr105276.C
new file mode 100644
index 00000000000..ad0e9dd7e09
--- /dev/null
+++ b/gcc/testsuite/g++.dg/pr105276.C
@@ -0,0 +1,18 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O2 -fdump-tree-optimized" } */
+
+bool
+foo(unsigned i)
+{
+  bool result = true;
+  while (i)
+    {
+      i = i % 3;
+      i = i - (i == 2 ? 2 : i ? 1 : 0);
+      result = !result;
+    }
+  return result;
+}
+
+/* We should be able to eliminate the i - operation.  */
+/* { dg-final { scan-tree-dump-not "i_.* - " "optimized" } } */