[polly] independent blocks: do not insert stores between phi nodes

[Sebastian Pop]

Tobias Grosser wrote:
> On 05/02/2013 11:05 PM, Sebastian Pop wrote:
> >Hi,
> >
> >When translateScalarToArray tranlates a phi node to an array, it does not make
> >sure that the insertion point is after all the phi nodes of the basic block, and
> >that could lead to a mix of phi nodes and translated stmts: this ICEs one of the
> >verifiers that expects to see all the phi nodes grouped at the beginning of a
> >basic block.  The patch moves the insertion point for the translation of a phi
> >node after all the phi nodes.
> >
> >Ok to commit?
> 
> Looks good, but please include a _reduced_ test case.

I was not able to reproduce this problem in the current tree.  Do you want still
to commit this fix, or you want to keep the patch on the side in case the bug
reappears in the future?

Sebastian
-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation