Moore's law and why persistence may not be necessary.

John M McIntosh johnmci at smalltalkconsulting.com
Fri Jan 25 05:04:41 UTC 2002


>Marcus Denker <marcus at ira.uka.de> said:
>>Hey, this is a computer, can't it be programmed to organize it's memory?
>>
>I suddenly recalled there being an experimental(?) mmap-based memory allocator
>for Linux in the Unix VM glue code. It would be trivial to try to mmap the
>image into memory, wouldn't it? (at the moment, it mmaps /dev/zero with
>copy-on-write set into memory, so that writes to memory don't affect the
>underlying file). Load the (platform-independent) base image, mmap a nice
>chunk from a platform-dependent work image, and never bother again with
>saving.
>

I added anonymous mmap to the Mac OS-X version (3.2.2), and wrote 
some code to enable it for Mac OS 9.1 (not yet implemented). However 
it's unclear to me if one wants the feature you mention?

a) In case of an issue you can't kill the VM, go back and restore 
from a change set because changes you make *are* persistent.

b) If your machine crashes I'm not quite sure what the state of your 
image would be, are there pages outstanding to be written not yet 
written?

c) If you wander around writing bits, how much IO would that 
generate, especially if you do a full GC and compaction.

However if you make a a copy of the VM at startup and work with it 
and on a save, then overwrite the original (safely) that might solve 
the issues of (a) and (b), but I wonder what issues (c) gives you.



-- 
--
===========================================================================
John M. McIntosh <johnmci at smalltalkconsulting.com> 1-800-477-2659
Corporate Smalltalk Consulting Ltd.  http://www.smalltalkconsulting.com
===========================================================================



More information about the Squeak-dev mailing list