llvm dev - Feb 2010 - [LLVMdev] SelectionDAG legality: isel creating cycles

On Feb 22, 2010, at 8:41 AM, David Greene wrote:
> On Monday 22 February 2010 10:31:24 David Greene wrote:
> 
>> The fundamental issue is that the DAG originally looked like this:
>> 
>> MIN
>>  LOAD B
>>    PREFETCH
>>      Chain from LOAD A
>>  LOAD A
> 
> Actually, it looked like this:
> 
> MIN
>  LOAD B
>    Chain from PREFETCH
>      Chain from LOAD A
>    NULL
>  LOAD A
>    Chain from ENTRY
>    Some Addressing
> 
> LOAD B is from NULL because this is a bugpoint-reduced testcase, so
that's not
> an issue.
> 
> Just wanted to clarify in case someone was wondering about this.
I'm currently working in this area.  What pattern is causing the cycle?  Can
I get a testcase?

-Chris

On Monday 22 February 2010 13:06:39 Chris Lattner wrote:
> > Just wanted to clarify in case someone was wondering about this.
>
> I'm currently working in this area.  What pattern is causing the cycle?
> Can I get a testcase?
I'll see if I can generate one and file a PR.

                                        -Dave

On Monday 22 February 2010 13:26:54 David Greene wrote:
> On Monday 22 February 2010 13:06:39 Chris Lattner wrote:
> > > Just wanted to clarify in case someone was wondering about this.
> >
> > I'm currently working in this area.  What pattern is causing the
cycle?
> > Can I get a testcase?
>
> I'll see if I can generate one and file a PR.
Ah, isLegalToFold saves us on trunk.  But we lose folding due to prefetching,
which is unfortunate.

I am seeing the error with 2.5 (yes, we are upgrading!).

I guess I'll have to backport some of the isLogalToFold logic.

                                              -Dave