Dec 30 patch causes large performance regression on ia64

[Bernd Schmidt]

On Sat, 17 Mar 2001, Richard Henderson wrote:

> (A) Teach reload about cond_exec, so that the spill/fill instructions
>     it generates are protected by the same predicate.  Then we can do
>     if-conversion before register allocation and then reload will see
>     that the predicate register is live across a call and put it in a
>     call-saved hard reg.

I've thought about that too; the thought of having to fix the post-reload
cse/combine/whatnot passes is the only thing holding me back.

> (B) Hack pre-reload ifcvt to notice when we'd like to if-convert a
>     block and emit a USE at the end.  Now the pseudo is live across
>     the call (and conflicts with other predicates used within the
>     block).  This might let us if-convert more often.

Hmm, that sounds interesting.  Bit of a hack, though.


Bernd