llvm dev - Feb 2017 - Problem ScheduleDAG on PowerPC, X86 works fine.

Long story short: https://llvm.org/bugs/show_bug.cgi?id=31890

The backend fails to schedule a given DAG, the reason being that there is
an instruction and it glue that needs to be broken apart as they can't be
scheduled consecutively. See attached file for a picture of the DAG.

Not sure what's the best course of action is, and not sure why this
isn't a
problem for the X86 backend either. I'm looking for advice on the best
course of actions. As I see it, the option are:

1/ add extra logic in the DAGCombiner to make sure this doesn't happen. I
don't see a way this could be done cheaply and overall I don't think
this
is the best option/
2/ Have the ScheduleDAG machinery detect this case and break up the glue,
for instance via breaking up (adde X, Y, Carry) into (add (add X, Y)n (adde
0, 0, Carry)) or something alike when the situation present itself.
3/ Do whatever the X86 backend does, which I'm not sure what it is.

Advice ?

Thanks in advance,

Amaury SECHET
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170207/6808301d/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dagisel.png
Type: image/png
Size: 251938 bytes
Desc: not available
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170207/6808301d/attachment-0001.png>

That's seems really odd that ADDC/ADDE uses glue there, instead of a plain
value.

The x86 backend has code that converts the glue into a value, which is why
it wasn't affected.... (LowerADDC_ADDE_SUBC_SUBE).

On Tue, Feb 7, 2017 at 2:48 PM, Amaury SECHET via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> Long story short: https://llvm.org/bugs/show_bug.cgi?id=31890
>
> The backend fails to schedule a given DAG, the reason being that there is
> an instruction and it glue that needs to be broken apart as they can't
be
> scheduled consecutively. See attached file for a picture of the DAG.
>
> Not sure what's the best course of action is, and not sure why this
isn't
> a problem for the X86 backend either. I'm looking for advice on the
best
> course of actions. As I see it, the option are:
>
> 1/ add extra logic in the DAGCombiner to make sure this doesn't happen.
I
> don't see a way this could be done cheaply and overall I don't
think this
> is the best option/
> 2/ Have the ScheduleDAG machinery detect this case and break up the glue,
> for instance via breaking up (adde X, Y, Carry) into (add (add X, Y)n (adde
> 0, 0, Carry)) or something alike when the situation present itself.
> 3/ Do whatever the X86 backend does, which I'm not sure what it is.
>
> Advice ?
>
> Thanks in advance,
>
> Amaury SECHET
>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170207/6e0811a6/attachment.html>

Would it not make sense to refactor the code so those don't use glue rather
than emitting them with glue and then getting rid of it. There are times
when we would like to emit these in separate blocks but can't (presumably
because of the glue).

On Tue, Feb 7, 2017 at 9:15 PM, James Y Knight via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> That's seems really odd that ADDC/ADDE uses glue there, instead of a
plain
> value.
>
> The x86 backend has code that converts the glue into a value, which is why
> it wasn't affected.... (LowerADDC_ADDE_SUBC_SUBE).
>
> On Tue, Feb 7, 2017 at 2:48 PM, Amaury SECHET via llvm-dev <
> llvm-dev at lists.llvm.org> wrote:
>
>> Long story short: https://llvm.org/bugs/show_bug.cgi?id=31890
>>
>> The backend fails to schedule a given DAG, the reason being that there
is
>> an instruction and it glue that needs to be broken apart as they
can't be
>> scheduled consecutively. See attached file for a picture of the DAG.
>>
>> Not sure what's the best course of action is, and not sure why this
isn't
>> a problem for the X86 backend either. I'm looking for advice on the
best
>> course of actions. As I see it, the option are:
>>
>> 1/ add extra logic in the DAGCombiner to make sure this doesn't
happen. I
>> don't see a way this could be done cheaply and overall I don't
think this
>> is the best option/
>> 2/ Have the ScheduleDAG machinery detect this case and break up the
glue,
>> for instance via breaking up (adde X, Y, Carry) into (add (add X, Y)n
(adde
>> 0, 0, Carry)) or something alike when the situation present itself.
>> 3/ Do whatever the X86 backend does, which I'm not sure what it is.
>>
>> Advice ?
>>
>> Thanks in advance,
>>
>> Amaury SECHET
>>
>>
>> _______________________________________________
>> LLVM Developers mailing list
>> llvm-dev at lists.llvm.org
>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>>
>>
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20170207/f85015b5/attachment.html>