[polly] r248697 - [tests] Add memory writes to make this scop not trivially empty

[Tobias Grosser via llvm-commits]

On 09/28/2015 09:46 AM, Johannes Doerfert wrote:
> Does that change anything?

Not in the scop detection. However, when looking at -polly-scops I was just
puzzeled to not see the inner loop. Hence, I ensured it does at least something
in case somebody else has a look.

I looked at this test in response to your last change. Maybe it is worth to
add this test to test/ScopInfo to ensure the scop is properly modeled?

>Was the test coupled with this benchmark
> needed after all?

I do not understand this sentence. I think the test is useful to cover the situation we
previously bailed out explicitly for.

Best,
Tobias