Am So., 3. Dez. 2023 um 11:49 Uhr schrieb John Cowan <[email protected]>:

> I don't see the analogy.  Anyway, the point is that no terminating program
> can *rely* on its garbage being collected.
>

The guarantee is the following one: Given a program with a space complexity
of O(1) (according to the Clinger definition) and time complexity of O(N),
there is an M independent of N such that in a VM with memory M the program
won't run out of memory (for any N) until it halts.



>
> On Sun, Dec 3, 2023 at 4:58 AM Marc Nieper-Wißkirchen <
> [email protected]> wrote:
>
>> Am So., 3. Dez. 2023 um 06:13 Uhr schrieb John Cowan <[email protected]>:
>>
>>> Marc Nieper-Wißkirchen scripsit:
>>>
>>> SRFI 124 is wrong here; by Scheme's proper tail call guarantee, defined
>>>> in "Will Clinger's Proper Tail Recursion and Space Efficiency," no longer
>>>> used locations must be garbage collected (up to some "O(1)").
>>>
>>>
>>>  Not necessarily.  If a program is known to terminate and doesn't
>>> require more locations (naively) than the machine makes available, then no
>>> garbage collection is required.
>>>
>>
>> This is like saying that bubble sort is an O(1) algorithm for any program
>> that terminates.
>>
>

Reply via email to