On 6/4/06, Leopold Toetsch <[EMAIL PROTECTED]> wrote:
On Jun 3, 2006, at 23:45, Charles Reiss wrote:
> <stm_frontend.pod>
<cite>
=item new Px, .STMRef
Creates a new transactionally managed variable which initially
contains
a clone of Px.
</cite>
The C<new> opcode is defined as:
new(out PMC, in INT)
That means that it is creating ("out") a new destination variable. it
can't take an existing C<Px> PMC, because the register allocator
diregards any existing register Px due to the "out" semantics of the
opcode.
I'm not sure why I didn't catch this earlier; I certainly didn't want
to specify such silly semantics. I think the most logical thing would
be to default to the null PMC instead.
-- Charles Reiss
