Re: [PATCH PR64434]

[Richard Biener <richard.guenther@gmail.com>]

On Thu, Jan 15, 2015 at 10:39 AM, Yuri Rumyantsev <ysrumyan@gmail.com> wrote:
> Hi All,
>
> I did a change proposed by Richard - unconditionally allocate from the heap.
>
> Bootstrap and regression testing did not show any new failures.
>
> Is it OK for trunk?

+      if (!is_gimple_assign (stmt)
+         || gimple_has_volatile_ops (stmt))
+       continue;
+      code = gimple_assign_rhs_code (stmt);
+      if (!commutative_tree_code (code))
+       continue;
+      gcc_assert (gimple_num_ops (stmt) == 3);
+      op0 = gimple_op (stmt, 1);
+      op1 = gimple_op (stmt, 2);
+      if (op0 == NULL_TREE || op1 == NULL_TREE
+         || TREE_CODE (op0) != SSA_NAME
+         || TREE_CODE (op1) != SSA_NAME)
+       continue;

you can simplify this to

         if (!is_gimple_assign (stmt)
            || !commutative_tree_code (gimple_assign_rhs_code (stmt)))
           continue;
         op0 = gimple_assign_rhs1 (stmt);
         op1 = gimple_assign_rhs2 (Stmt);
         if (TREE_CODE (op0) != SSA_NAME
             || TREE_CODE (op1) != SSA_NAME)
           continue;

Please output the computed costs for both operands in

+         if (dump_file && (dump_flags & TDF_DETAILS))
+           {
+             fprintf (dump_file, "Swap operands in stmt:\n");
+             print_gimple_stmt (dump_file, stmt, 0, TDF_SLIM);
+           }

as that will help debugging if odd things happen.

Ok with that changes.  I belive this may fix some duplicate bugs
as well.

Thanks,
Richard.


> ChangeLog
>
> 2015-01-15  Yuri Rumyantsev  <ysrumyan@gmail.com>
>
> PR tree-optimization/64434
> * cfgexpand.c (reorder_operands): New function.
> (expand_gimple_basic_block): Insert call of reorder_operands if
> optimized is true.
>
> gcc/testsuite/ChangeLog
> * gcc.dg/torture/pr64434.c: New test.
>
> 2015-01-14 17:07 GMT+03:00 Yuri Rumyantsev <ysrumyan@gmail.com>:
>> Jakub,
>>
>> I did all changes requested by you.
>>
>> Here is updated patch.
>>
>> BTW I thought that gcc performs splitting of blocks with huge  size.
>>
>>
>> 2015-01-14 16:33 GMT+03:00 Jakub Jelinek <jakub@redhat.com>:
>>> On Wed, Jan 14, 2015 at 04:28:42PM +0300, Yuri Rumyantsev wrote:
>>>> Hi All,
>>>>
>>>> I did all changes proposed by Richard and delete check on def in the
>>>> same block as Jakub proposed.
>>>> I also moved check  on optimization to call site..
>>>>
>>>> I also checked that bootstrap and regression testing did not show any
>>>> new failures.
>>>>
>>>> Is it OK for trunk?
>>>
>>> The  | SSA_OP_VUSE is still in there, the testcase is still executable,
>>> still doesn't end with newline, and I really think you should replace
>>>   lattice = XALLOCAVEC (unsigned int, n);
>>> with something like:
>>>   if (n >= 100000)
>>>     lattice = XNEWVEC (unsigned int, n);
>>>   else
>>>     lattice = XALLOCAVEC (unsigned int, n);
>>> ...
>>>   if (n >= 100000)
>>>     XDELETE (lattice);
>>> or similar.
>>>
>>>         Jakub