Re: [v3] avoid alignment of static variables affecting stack's

[Bernd Schmidt]

On 12/10/2015 05:07 PM, Jan Beulich wrote:

If not reaching

   if (TREE_CODE (origvar) == SSA_NAME)
 {
   gcc_assert (TREE_CODE (var) != VAR_DECL
  || (!DECL_EXTERNAL (var)
  && !DECL_HAS_VALUE_EXPR_P (var)
  && !TREE_STATIC (var)
  && TREE_TYPE (var) != error_mark_node
  && !DECL_HARD_REGISTER (var)
  && really_expand));
 }

in that case is not a problem, this would make for even simpler a
change. Please let me know.


I think that's fine. BTW there's a is_global_var predicate that 
{c,sh}ould be used for your test.



Bernd

Re: [v3] avoid alignment of static variables affecting stack's

[Bernd Schmidt]

On 12/10/2015 01:38 PM, Jan Beulich wrote:


* cfgexpand.c (expand_one_var): Exclude static and external
variables when adjusting stack alignment related state.

gcc/testsuite/
2015-12-10  Jan Beulich  

* gcc.c-torture/execute/stkalign.c: New.

--- 2015-12-09/gcc/cfgexpand.c
+++ 2015-12-09/gcc/cfgexpand.c
@@ -1544,12 +1544,15 @@ static HOST_WIDE_INT
  expand_one_var (tree var, bool toplevel, bool really_expand)
  {
unsigned int align = BITS_PER_UNIT;
+  bool stack = true;
tree origvar = var;

var = SSAVAR (var);

if (TREE_TYPE (var) != error_mark_node && TREE_CODE (var) == VAR_DECL)
  {
+  stack = !TREE_STATIC (var) && !DECL_EXTERNAL (var);
+
/* Because we don't know if VAR will be in register or on stack,
 we conservatively assume it will be on stack even if VAR is
 eventually put into register after RA pass.  For non-automatic
@@ -1578,7 +1581,8 @@ expand_one_var (tree var, bool toplevel,
align = POINTER_SIZE;
  }

-  record_alignment_for_reg_var (align);
+  if (stack)
+record_alignment_for_reg_var (align);


A bit further down we have

  else if (DECL_EXTERNAL (var))
;
  else if (DECL_HAS_VALUE_EXPR_P (var))
;
  else if (TREE_STATIC (var))
;
[]
  return 0;

so I'm thinking the function doesn't do anything for DECL_EXTERNAL or 
TREE_STATIC vars. You're still computing alignment for them but not 
using it. I suggest just doing an early return for non-stack vars.



Bernd

Re: [v3] avoid alignment of static variables affecting stack's

[Jan Beulich]

>>> On 10.12.15 at 14:53,  wrote:
> On 12/10/2015 01:38 PM, Jan Beulich wrote:
>> --- 2015-12-09/gcc/cfgexpand.c
>> +++ 2015-12-09/gcc/cfgexpand.c
>> @@ -1544,12 +1544,15 @@ static HOST_WIDE_INT
>>   expand_one_var (tree var, bool toplevel, bool really_expand)
>>   {
>> unsigned int align = BITS_PER_UNIT;
>> +  bool stack = true;
>> tree origvar = var;
>>
>> var = SSAVAR (var);
>>
>> if (TREE_TYPE (var) != error_mark_node && TREE_CODE (var) == VAR_DECL)
>>   {
>> +  stack = !TREE_STATIC (var) && !DECL_EXTERNAL (var);
>> +
>> /* Because we don't know if VAR will be in register or on stack,
>>   we conservatively assume it will be on stack even if VAR is
>>   eventually put into register after RA pass.  For non-automatic
>> @@ -1578,7 +1581,8 @@ expand_one_var (tree var, bool toplevel,
>>  align = POINTER_SIZE;
>>   }
>>
>> -  record_alignment_for_reg_var (align);
>> +  if (stack)
>> +record_alignment_for_reg_var (align);
> 
> A bit further down we have
> 
>else if (DECL_EXTERNAL (var))
>  ;
>else if (DECL_HAS_VALUE_EXPR_P (var))
>  ;
>else if (TREE_STATIC (var))
>  ;
> []
>return 0;
> 
> so I'm thinking the function doesn't do anything for DECL_EXTERNAL or 
> TREE_STATIC vars. You're still computing alignment for them but not 
> using it. I suggest just doing an early return for non-stack vars.

If not reaching

  if (TREE_CODE (origvar) == SSA_NAME)
{
  gcc_assert (TREE_CODE (var) != VAR_DECL
  || (!DECL_EXTERNAL (var)
  && !DECL_HAS_VALUE_EXPR_P (var)
  && !TREE_STATIC (var)
  && TREE_TYPE (var) != error_mark_node
  && !DECL_HARD_REGISTER (var)
  && really_expand));
}

in that case is not a problem, this would make for even simpler a
change. Please let me know.

Jan