[PATCH] D20315: [LV] For some induction variables, use vector phis instead of widening the scalar in the loop body

[Wei Mi via llvm-commits]

wmi added a comment.

> I think this is actually a weakness in register coalescing. I already

>  have a similar testcase which probably have the same cause. It is good

>  to have another one now. It shows the problem may be more general than

>  I thought and justifies more work to improve it. Will file a separate

>  bug for it.

> 

> Thanks,

>  Wei.


Looks like phabricator did some massage for my mail and made it hard to read.

Filed a bug for the partial redundent mov problem:
https://llvm.org/bugs/show_bug.cgi?id=27827

Thanks,
Wei.


http://reviews.llvm.org/D20315