[11] RFR(M): 8203915: Induction variable of over-unrolled loop conflicts with range checks

[Tobias Hartmann]

Hi Roland,

On 12.06.2018 14:28, Roland Westrelin wrote:
> What Opaque1 nodes do you leave in?

I should have added that I'm not copying the skeleton predicate from the pre-loop because the
pre-loop might not longer exist when we perform unrolling of the main loop. Instead, I'm copying a
skeleton predicate which still has the Opaque1Node once to before the main loop that is then updated
during loop unrolling according to the new stride.

When updating the init+stride predicate during loop unrolling, I leave the Opaque1Node that guards
this expression in because otherwise the constant expression might be folded and we cannot update it
during the next iteration of unrolling.

Thanks,
Tobias