Optimizing of explicit temporary storage

[Mark Mitchell]

Robert Dewar wrote:

> Mark Mitchell wrote:
>
>> In C, even given just:
>>
>>  (void) malloc(16);
>>
>> it would be surprising to most programmers to optimize away the call 
>> because malloc *does* have side-effects on real systems.  For 
>> example, on UNIX, it is likely to call sbrk, which result in 
>> observable changes in the process state.
>
>
> Yes, but the condition of not surprising programmers is different from
> the condition of conforming to the standard. Strictly from a standards
> conformance point of view, these kinds of optimization seem perfectly
> valid, but of course the criterion of not surprising programmers is
> indeed a reasonable one!

I'm not sure stanadards are all that relevant here, since most real C 
programs go beyond the set of library functions standardized by ISO, or 
even POSIX.

The as-if rule says that the compiler can perform the optimization only 
if the program can't tell the difference.  I don't think it means "can't 
tell the difference using functions from the ISO C standard library;" 
instead, I think it means "can't tell the difference in the current 
environment."  Certainly, calling malloc has observable side effects on 
most UNIX systems; I can see whether or not processes are calling malloc 
by running "top".

-- 
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com