Re: Ada GC (was about Java VM)

dewar@cs.nyu.edu (Robert Dewar) writes:


>"As a non-expert in hard real-time systems, the last sentence strikes
>me as very strange. There are well-known GC algorithms that guarantee
>hard real-time bounds, as well as reasonable (linear) space bounds.
>Henry is the inventor of one."


>During the design of Ada 95, it became abundantly clear that this view
>is NOT shared by most experienced practioners in the real time field.


It seems to me that in order to make progress in this discussion we
need an example of a hard real-time system that uses a general purpose
allocator with explicit deallocation. I claimed it can't be done if
you need reasonable space bounds. (Cf. Knuth, vol.1, 2nd ed, p. 255,
and the answers in the back. He cites a 1971 paper by Robson.) I have
also never seen a general purpose allocator/deallocator that supplied
time bounds adequate for a hard real-time system. (A buddy system
allocator without coalescing could probably do OK, but it's allocation
time is still quite variable, and its space utilization is
suboptimal.)


My suspicion is that we're not really talking about systems with
provable time and space bounds, but about something less stringent,
with less well-defined constraints. I'd like to understand what it
is.


Hans-J. Boehm
(boehm@parc.xerox.com)
--

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