[PATCH] [Polly] Data flow reduction detection

[Johannes Doerfert]

>>! In D5190#4, @grosser wrote:
> Without reviewing this code already in detail, I think this is very valuable. Specifically, it might allow us to (better) optimize the himeono benchmark (http://accc.riken.jp/2444.htm), which is also available in our test suite: ./SingleSource/Benchmarks/Misc/himenobmtxpa.c

I looked into the source but I cannot find a "complex reduction like access". Which line/variable did you have in mind?

http://reviews.llvm.org/D5190