[llvm] r189281 - LoopVectorize: Implement partial loop unrolling when vectorization is not profitable.

[Nadav Rotem]

On Aug 28, 2013, at 1:58 PM, Eric Christopher <echristo at gmail.com> wrote:

> Hi Nadav,
> 
> Sure, it seems reasonable to me that this should be hoisted out to
> some analysis and then stuck in the general loop unrolling pass. What
> do you think?
> 

Hi Eric, 

Thanks for the suggestion.  No, because this is not the kind of loop-unrolling that the loop-unroller does. The term is overloaded. The loop unroller does not handle reduction variables, it does not check for memory safety across iterations, and it does not model register pressure. This is something completely different. 

Thanks,
Nadav