William Coleda <[EMAIL PROTECTED]> wrote: > Is it worth trying to optimize the mapping of $P2 to P5 there to avoid > the extra assignment? possible?
"parrot -Oc" did that until a few days ago. It was broken. The problem
arises with code like:
foo(a,b) # P5, P6
foo(b,a) # P6, P5 different mapping in the caller
.sub foo
.param pmc x
.param pmc y
.return (y,x) # different mapping in sub
At some point the (new) register allocator will use hints for simple
cases along with register renaming.
leo
