Re: Order of argument evaluation in C++, etc.

Stefan Monnier <stefan.monnier@epfl.ch> wrote:
>
>What's wrong with using
>
> tmp1=argexp1;
> tmp2=argexp2;
> ...
> fun(tmp1, tmp2, ...)
>
>instead of
>
> fun(argexp1, argexp2, ..)
>


What about providing more than one invocation mechanism, allowing the
programmer to specify partial ordering constraints? I always thought such
"parallel" invocations would clearly specify a sort of private
barrier-syncronization, particularily when mixed with first-class
continuations, but I wouldn't want to have to manually syncronize all those
other calls...


Or perhaps a meta-object protocol to configure the invocation mechanism in a
particular lexical scope would satisfy both the sequential-thinking
programmer and the parallel architecture?




karl c.


--
Karl Czajkowski
Reply-To: karlcz@uclink.berkeley.edu
--

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