[CVS ci] more PIR codegen improvements
Leopold Toetsch
* in @MAIN P1 isn't moved around
* -Oc: replace savetop with up to 3 pushtopX ops
It's a good idea to declare the main function:
.sub _main @MAIN
...
.end
leo
Sterling Hughes
On Mar 22, 2004, at 11:42 AM, Leopold Toetsch wrote:
> * end is now inserted in @MAIN if missing
perhaps this should be a compiler warning instead?
-sterling
Leopold Toetsch
If both C<@MAIN> and C<end> are missing, we could compile a module. So
sometimes you'll get a warning and sometimes not. Not good IMHO.
Inserting C<end> in C<@main> matches the automatically inserted
C<invoke P1> in subs.
> -sterling
leo
Dan Sugalski
>Sterling Hughes <ster...@php.net> wrote:
>
>> On Mar 22, 2004, at 11:42 AM, Leopold Toetsch wrote:
>
>>> * end is now inserted in @MAIN if missing
>
>> perhaps this should be a compiler warning instead?
>
>If both C<@MAIN> and C<end> are missing, we could compile a module. So
>sometimes you'll get a warning and sometimes not. Not good IMHO.
Yeah. And IMCC's really a low-level thing--if you get the code to it
wrong, then you've got a busted compiler and are likely in lots of
trouble as it is. (Not that it makes *writing* a compiler any easier,
though...)
I think I'd as soon leave out this check by default, and enable it
(and other compiler-writer-friendly operations) with a --picky
option. That way we don't pay the cost normally, but it's there to be
paid if we really need it. (Or enable it with a #define that's
normally not defined, or something)
--
Dan
--------------------------------------"it's like this"-------------------
Dan Sugalski even samurai
d...@sidhe.org have teddy bears and even
teddy bears get drunk