[PATCH] Allow to model non-affine control flow in a SCoP.

Tobias Grosser tobias at grosser.es
Fri Feb 20 01:03:09 PST 2015

On 20.02.2015 09:20, Johannes Doerfert wrote:
> Codegen wasn't finished yesterday afternoon but it is now. I would have
> commited it in two patches/reviews but I can merge them too if you'd
> like.

I am a little confused. You do not seem to have an option that disables 
the non-affine control flow support. Would the patch as submitted then 
not break as soon as we encounter such non-affine control flow cases? I 
probably missed something, but in case the first patch would break 
Polly, I do not think we should commit it alone.


More information about the llvm-commits mailing list