pmw@cunixb.cc.columbia.edu (Peter M Williams) (04/29/91)
I'm wondering if there is a way to turn off the garbage collection process in MIT scheme version 7.1.0. It seems to me that suppressing garbage collection until program completion may shorten the execution time for programs. Please let me know if you have any information on this. Thanks for any information. -Peter, pmw@cunixb.cc.columbia.edu
jinx@zurich.ai.mit.edu (Guillermo J. Rozas) (04/30/91)
In article <1991Apr28.193104.2627@cunixf.cc.columbia.edu> pmw@cunixb.cc.columbia.edu (Peter M Williams) writes:
Path: ai-lab!snorkelwacker.mit.edu!paperboy!think.com!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!cunixf.cc.columbia.edu!cunixb.cc.columbia.edu!pmw
From: pmw@cunixb.cc.columbia.edu (Peter M Williams)
Newsgroups: comp.lang.scheme
Date: 28 Apr 91 19:31:04 GMT
Sender: usenet@cunixf.cc.columbia.edu (The Network News)
Reply-To: pmw@cunixb.cc.columbia.edu
Organization: Columbia University
Lines: 10
Originator: pmw@cunixb.cc.columbia.edu
Nntp-Posting-Host: cunixb.cc.columbia.edu
I'm wondering if there is a way to turn off the garbage collection
process in MIT scheme version 7.1.0. It seems to me that suppressing
garbage collection until program completion may shorten the execution time
for programs. Please let me know if you have any information on this.
Thanks for any information.
-Peter, pmw@cunixb.cc.columbia.edu
There is a way. If you are not using the SICP compatibility package,
you can use
(with-absolutely-no-interrupts <thunk>)
which will invoke thunk disallowing all interrupts, including GC
(which is treated as an interrupt in our system).
I don't recommend that you use this, however, unless you know that
<thunk> will not cons at all (or the system on its behalf). Otherwise
you will cause the system to crash.
MIT Scheme does not collect garbage spuriously, but only when it is
out of storage. If it is out of storage, what would you have it do?
Note that if your code is running interpreted, typically 90% of all
consing is done by the interpreter, rather than the program you are
running, and thus the frequency of garbage collection may be greater
than what you would expect. In particular, programs that do not
apparently cons will cause the system to GC when they are run
interpreted.
The compiler changes things significantly, but it is still possible
for your program to cause garbage collections even if it does not
explicitly invoke any allocation procedures. Passing and returning
procedure objects often causes allocation, and the garbage collector
must be run eventually.
A completely different question is whether you can collect garbage
pre-emptively. For example, the system could invoke the collector
every time it types a prompt at a terminal, so that if the user takes
any time to think about the next expression to type, the garbage
collection will have been done in the interim.
You can play with the global variable HOOK/CMDL-PROMPT, and the global
procedure GC-FLIP and get this to work, and decide whether you like it
better than the default.