Re: fwprop fix for PR79405

[Richard Biener]

On Thu, Feb 23, 2017 at 11:17 PM, Jeff Law  wrote:
> On 02/23/2017 01:57 AM, Richard Biener wrote:
>>
>> On Wed, Feb 22, 2017 at 6:19 PM, Jeff Law  wrote:
>>>
>>> On 02/16/2017 12:41 PM, Bernd Schmidt wrote:


 We have two registers being assigned to each other:

  (set (reg 213) (reg 209))
  (set (reg 209) (reg 213))

 These being the only definitions, we are happy to forward propagate reg
 209 for reg 213 into a third insn, making a new use for reg 209. We are
 then happy to forward propagate reg 213 for it in the same insn...
 ending up in an infinite loop.

 I don't really see an elegant way to prevent this, so the following just
 tries to detect the situation (and more general ones) by brute force.
 Bootstrapped and tested on x86_64-linux, verified that the test passes
 with a ppc cross, ok?


 Bernd


 79405.diff


 PR rtl-optimization/79405
 * fwprop.c (forward_propagate_into): Detect potentially cyclic
 replacements and bail out for them.

 PR rtl-optimization/79405
 * gcc.dg/torture/pr79405.c: New test.
>>>
>>>
>>> OK.
>>
>>
>> Err - this looks quite costly done for each fwprop.  And placing it before
>> less costly bailouts even...
>>
>> See my discussion with Bernd anyway.
>
> I read your last message as being OK with Bernd's approach?  Did I
> mis-understand?

No, I was refering to the following with "walking stmts"

---
But isn't the issue that we are walking "all uses" (in random order) rather than
only processing each stmt once?  That is,

  /* Go through all the uses.  df_uses_create will create new ones at the
 end, and we'll go through them as well.

 Do not forward propagate addresses into loops until after unrolling.
 CSE did so because it was able to fix its own mess, but we are not.  */

  for (i = 0; i < DF_USES_TABLE_SIZE (); i++)
{
  df_ref use = DF_USES_GET (i);
  if (use)
if (DF_REF_TYPE (use) == DF_REF_REG_USE
|| DF_REF_BB (use)->loop_father == NULL
/* The outer most loop is not really a loop.  */
|| loop_outer (DF_REF_BB (use)->loop_father) == NULL)
  forward_propagate_into (use);
}

if that were simply walking all instructions, doing forward_propagat_into on
each use on an instruction we'd avoid the cycle (because we stop propagating).

Because when propagating DF_USES_TABLE changes.
---

Richard.

> jeff
>

Re: fwprop fix for PR79405

[Jeff Law]

On 02/23/2017 01:57 AM, Richard Biener wrote:

On Wed, Feb 22, 2017 at 6:19 PM, Jeff Law  wrote:

On 02/16/2017 12:41 PM, Bernd Schmidt wrote:


We have two registers being assigned to each other:

 (set (reg 213) (reg 209))
 (set (reg 209) (reg 213))

These being the only definitions, we are happy to forward propagate reg
209 for reg 213 into a third insn, making a new use for reg 209. We are
then happy to forward propagate reg 213 for it in the same insn...
ending up in an infinite loop.

I don't really see an elegant way to prevent this, so the following just
tries to detect the situation (and more general ones) by brute force.
Bootstrapped and tested on x86_64-linux, verified that the test passes
with a ppc cross, ok?


Bernd


79405.diff


PR rtl-optimization/79405
* fwprop.c (forward_propagate_into): Detect potentially cyclic
replacements and bail out for them.

PR rtl-optimization/79405
* gcc.dg/torture/pr79405.c: New test.


OK.


Err - this looks quite costly done for each fwprop.  And placing it before
less costly bailouts even...

See my discussion with Bernd anyway.
I read your last message as being OK with Bernd's approach?  Did I 
mis-understand?


jeff

Re: fwprop fix for PR79405

[Richard Biener]

On Wed, Feb 22, 2017 at 6:19 PM, Jeff Law  wrote:
> On 02/16/2017 12:41 PM, Bernd Schmidt wrote:
>>
>> We have two registers being assigned to each other:
>>
>>  (set (reg 213) (reg 209))
>>  (set (reg 209) (reg 213))
>>
>> These being the only definitions, we are happy to forward propagate reg
>> 209 for reg 213 into a third insn, making a new use for reg 209. We are
>> then happy to forward propagate reg 213 for it in the same insn...
>> ending up in an infinite loop.
>>
>> I don't really see an elegant way to prevent this, so the following just
>> tries to detect the situation (and more general ones) by brute force.
>> Bootstrapped and tested on x86_64-linux, verified that the test passes
>> with a ppc cross, ok?
>>
>>
>> Bernd
>>
>>
>> 79405.diff
>>
>>
>> PR rtl-optimization/79405
>> * fwprop.c (forward_propagate_into): Detect potentially cyclic
>> replacements and bail out for them.
>>
>> PR rtl-optimization/79405
>> * gcc.dg/torture/pr79405.c: New test.
>
> OK.

Err - this looks quite costly done for each fwprop.  And placing it before
less costly bailouts even...

See my discussion with Bernd anyway.

Richard.

> jeff
>

Re: fwprop fix for PR79405

[Jeff Law]

On 02/16/2017 12:41 PM, Bernd Schmidt wrote:

We have two registers being assigned to each other:

 (set (reg 213) (reg 209))
 (set (reg 209) (reg 213))

These being the only definitions, we are happy to forward propagate reg
209 for reg 213 into a third insn, making a new use for reg 209. We are
then happy to forward propagate reg 213 for it in the same insn...
ending up in an infinite loop.

I don't really see an elegant way to prevent this, so the following just
tries to detect the situation (and more general ones) by brute force.
Bootstrapped and tested on x86_64-linux, verified that the test passes
with a ppc cross, ok?


Bernd


79405.diff


PR rtl-optimization/79405
* fwprop.c (forward_propagate_into): Detect potentially cyclic
replacements and bail out for them.

PR rtl-optimization/79405
* gcc.dg/torture/pr79405.c: New test.

OK.
jeff

Re: fwprop fix for PR79405

[Richard Biener]

On February 20, 2017 2:58:54 PM GMT+01:00, Bernd Schmidt  
wrote:
>On 02/17/2017 10:11 AM, Richard Biener wrote:
>> Index: gcc/fwprop.c
>> ===
>> --- gcc/fwprop.c(revision 245501)
>> +++ gcc/fwprop.c(working copy)
>> @@ -1478,7 +1478,8 @@ fwprop (void)
>>   Do not forward propagate addresses into loops until after
>unrolling.
>>   CSE did so because it was able to fix its own mess, but we are
>not.  */
>>
>> -  for (i = 0; i < DF_USES_TABLE_SIZE (); i++)
>> +  unsigned sz = DF_USES_TABLE_SIZE ();
>> +  for (i = 0; i < sz; i++)
>>  {
>>df_ref use = DF_USES_GET (i);
>>if (use)
>>
>> might work?  (not knowing too much about this detail of the DF data
>> structures - can the table shrink?)
>
>This would probably work to fix the bug, but this behaviour is 
>explicitly documented as intentional (in the comment the second half of
>
>which you've quoted). I assume it enables additional substitutions.

Hmm, this means the walking-stmts solution sounds more correct and also gets 
these second-level opportunities.

Richard.

>
>Bernd

Re: fwprop fix for PR79405

[Bernd Schmidt]

On 02/17/2017 10:11 AM, Richard Biener wrote:

Index: gcc/fwprop.c
===
--- gcc/fwprop.c(revision 245501)
+++ gcc/fwprop.c(working copy)
@@ -1478,7 +1478,8 @@ fwprop (void)
  Do not forward propagate addresses into loops until after unrolling.
  CSE did so because it was able to fix its own mess, but we are not.  */

-  for (i = 0; i < DF_USES_TABLE_SIZE (); i++)
+  unsigned sz = DF_USES_TABLE_SIZE ();
+  for (i = 0; i < sz; i++)
 {
   df_ref use = DF_USES_GET (i);
   if (use)

might work?  (not knowing too much about this detail of the DF data
structures - can the table shrink?)


This would probably work to fix the bug, but this behaviour is 
explicitly documented as intentional (in the comment the second half of 
which you've quoted). I assume it enables additional substitutions.



Bernd

Re: fwprop fix for PR79405

[Richard Biener]

On Fri, Feb 17, 2017 at 10:07 AM, Richard Biener
 wrote:
> On Thu, Feb 16, 2017 at 8:41 PM, Bernd Schmidt  wrote:
>> We have two registers being assigned to each other:
>>
>>  (set (reg 213) (reg 209))
>>  (set (reg 209) (reg 213))
>>
>> These being the only definitions, we are happy to forward propagate reg 209
>> for reg 213 into a third insn, making a new use for reg 209. We are then
>> happy to forward propagate reg 213 for it in the same insn... ending up in
>> an infinite loop.
>>
>> I don't really see an elegant way to prevent this, so the following just
>> tries to detect the situation (and more general ones) by brute force.
>> Bootstrapped and tested on x86_64-linux, verified that the test passes with
>> a ppc cross, ok?
>
> But isn't the issue that we are walking "all uses" (in random order) rather 
> than
> only processing each stmt once?  That is,
>
>   /* Go through all the uses.  df_uses_create will create new ones at the
>  end, and we'll go through them as well.
>
>  Do not forward propagate addresses into loops until after unrolling.
>  CSE did so because it was able to fix its own mess, but we are not.  */
>
>   for (i = 0; i < DF_USES_TABLE_SIZE (); i++)
> {
>   df_ref use = DF_USES_GET (i);
>   if (use)
> if (DF_REF_TYPE (use) == DF_REF_REG_USE
> || DF_REF_BB (use)->loop_father == NULL
> /* The outer most loop is not really a loop.  */
> || loop_outer (DF_REF_BB (use)->loop_father) == NULL)
>   forward_propagate_into (use);
> }
>
> if that were simply walking all instructions, doing forward_propagat_into on
> each use on an instruction we'd avoid the cycle (because we stop propagating).
>
> Because when propagating DF_USES_TABLE changes.

Which either means we might even miss visiting some uses or a fix as simple as

Index: gcc/fwprop.c
===
--- gcc/fwprop.c(revision 245501)
+++ gcc/fwprop.c(working copy)
@@ -1478,7 +1478,8 @@ fwprop (void)
  Do not forward propagate addresses into loops until after unrolling.
  CSE did so because it was able to fix its own mess, but we are not.  */

-  for (i = 0; i < DF_USES_TABLE_SIZE (); i++)
+  unsigned sz = DF_USES_TABLE_SIZE ();
+  for (i = 0; i < sz; i++)
 {
   df_ref use = DF_USES_GET (i);
   if (use)

might work?  (not knowing too much about this detail of the DF data
structures - can
the table shrink?)

Richard.

> Richard.
>
>
>>
>>
>> Bernd
>>

Re: fwprop fix for PR79405

[Richard Biener]

On Thu, Feb 16, 2017 at 8:41 PM, Bernd Schmidt  wrote:
> We have two registers being assigned to each other:
>
>  (set (reg 213) (reg 209))
>  (set (reg 209) (reg 213))
>
> These being the only definitions, we are happy to forward propagate reg 209
> for reg 213 into a third insn, making a new use for reg 209. We are then
> happy to forward propagate reg 213 for it in the same insn... ending up in
> an infinite loop.
>
> I don't really see an elegant way to prevent this, so the following just
> tries to detect the situation (and more general ones) by brute force.
> Bootstrapped and tested on x86_64-linux, verified that the test passes with
> a ppc cross, ok?

But isn't the issue that we are walking "all uses" (in random order) rather than
only processing each stmt once?  That is,

  /* Go through all the uses.  df_uses_create will create new ones at the
 end, and we'll go through them as well.

 Do not forward propagate addresses into loops until after unrolling.
 CSE did so because it was able to fix its own mess, but we are not.  */

  for (i = 0; i < DF_USES_TABLE_SIZE (); i++)
{
  df_ref use = DF_USES_GET (i);
  if (use)
if (DF_REF_TYPE (use) == DF_REF_REG_USE
|| DF_REF_BB (use)->loop_father == NULL
/* The outer most loop is not really a loop.  */
|| loop_outer (DF_REF_BB (use)->loop_father) == NULL)
  forward_propagate_into (use);
}

if that were simply walking all instructions, doing forward_propagat_into on
each use on an instruction we'd avoid the cycle (because we stop propagating).

Because when propagating DF_USES_TABLE changes.

Richard.


>
>
> Bernd
>