[PATCH] ScopDetection: Only detect scops that have at least one read and one write

[Tobias Grosser]

On 19.02.2015 15:22, Johannes Doerfert wrote:
> Hey Tobias,
>
> do you kow how much we actually gain by your change?

For certain benchmarks we see large (up to 70%, mostly around 20%) 
reductions in compile time.

 > Why shouldn't we
> keep (and delete) SCoPs that have no stores (or at least delete the
> loads and let the other passes delete the rest)? Why shouldn't we look at
> generating memset intrinsics for loops that only write?

In some case we possibly could. However, I think it is important to make 
Polly by default use little compile time. We can then on a case-by-case 
basis decide if a certain optimization should be performed by Polly and 
if the involved run-time benefit is worth the compile-time cost involved.

> Regarding the patch, could you please comment your member variables
> (even though it's pretty obvious).

r229854

Thank you,
Tobias