Re: Compile Time Garbage Collection impossible?

torbenm@app-3.diku.dk (Torben gidius Mogensen) schreibt:


> Florian Liekweg <liekweg@ipd.info.uni-karlsruhe.de> writes:
>
>> Torben gidius Mogensen wrote:
>
>>> [...] the programmer may have to program with regions
>>> in mind, sometimes doing things somewhat differently than he would
>>> with RTGC.
>>
>> Programming "with regions in mind" is very much like programming with
>> explicit, manual memory management in mind, except that you don't get
>> to deallocate single objects (which, come to think of it, isn't much
>> of a restriction).


[...]


>> Programming with "what will the compiler do with this program" in
>> mind would seem to be more of a torture, though.
>
> I think most programmers do exactly that.


If there's any truth to the statement that programs spend 90% of their
time in 10% of the code, they shouldn't. If you find your finished
program is too slow, you should use a profiler. There are exceptions
(interpreters, OSes, etc., when you know you need all the performance
you can get), but most programs just need to be fast enough.


> And you can argue that you need to have a good model of what the
> compiler does to understand what happens in, say, virtual method
> calls.


Yes, but understanding language features and having to worry about the
compiler all the time are different things. Coming back to memory
management, the fact that Java is garbage-collected is probably one of
the main reasons why it has become so popular. After all, apart from
not having delete, Java is not *that* different from C++.




Wolfram

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