Re: Argument passing conventions and optimization

Robert A Duff <bobduff@shell01.TheWorld.com> wrote:
>pradeep tumati <ptumati@yahoo.com> writes:
>
>> Does this case even apply to parameter passing in the case of static
>> functions (in C)? Logically, we can perform optimization, but I am not
>> sure what kind of side effects this causes... The same can be the case
>> with functions calling other functions within the same C++ Namespace.
>
>One side effect might be that the debugger is utterly confused.
>Presuming one is trying to use a so-called "language-independent"
>debugger.


Right in one.


>It seems to me that a compiler for a particular language ought to be
>allowed to choose its own calling conventions in at least *some*
>circumstances. (Like, no debugging, the call is not inter-langage, the
>call is not calling a library routine).


They can, and always have done. It was near-universal under IBM
MVT and MVS, where the standard linkage was (a) very slow and
(b) poorly suited to many languages.


>> There could be a scenario where this kind of optimization can be
>> entrusted to the linker, but I guess its going to be a costly
>> stuff. Any thoughts...?
>
>Have we got to the point where link-time optimizations are worthwhile?


Many HPC vendors (system manufacturers and third-party) think so.


Regards,
Nick Maclaren.

Return to the comp.compilers page.
Search the comp.compilers archives again.