[polly] r212198 - Derive run-time conditions for delinearization

[Tobias Grosser]

On 03/08/2014 04:20, Johannes Doerfert wrote:
> Hey Tobias,
>
> I think this change introduces an unwanted effect.
>
> For one dimensional parametric array accesses we add assumptions about the
> parameters to ensure we access the array only in a positive range. Is this
> intended? I see the point for multidimensional array accesses but why do we
> enforce this for one dimensional accesses too?

What about the attached patch?

Tobias

-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-Do-allow-negative-offsets-in-the-outermost-array-dim.patch
Type: text/x-patch
Size: 9433 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140803/f27713ac/attachment.bin>