[LLVMdev] Virtual register def doesn't dominate all uses

[Tim Northover]

Hi Boris,

On 5 November 2014 06:47, Boris Boesler <baembel at gmx.de> wrote:
>  Cross-check: Move the MV-patterns from the beginning to the end. Voila, it works.

It would be better to delete those patterns entirely. They'll always
match (if asked) and never give a sane result.

>  I assume this is a bug/imperfection in the table match generator.

It possibly should be a compile-time error, if we can detect it
reliably (maybe look for at least one real node being consumed?). But
the kind of pattern it's trying to represent just shouldn't exist.

Cheers.

Tim.