[LLVMdev] EFLAGS and MVT::Glue

[Anton Korobeynikov]

Hello Edmund,

> Am I right in thinking that one of the problems with MVT::Glue is that
> it is hard to guarantee that other instructions won't come between the
> two instructions that are glued together?
No. This is actually why MVT::Glue was introduced: to always glue
stuff altogether.

> * How hard is it to replace Glue in other back ends by something like
>  EFLAGS? Should we be doing that?
This is not possible for ARM right now, because latency-modelling scheduler
cannot handle physreg deps (there is PR for this). But in general, adding more
"degrees of freedom" is a generic goodness.

-- 
With best regards, Anton Korobeynikov
Faculty of Mathematics and Mechanics, Saint Petersburg State University