[gclist] Memory cheats and VM (was Counting on one thumb)
David Chase
chase@centerline.com
Wed, 15 Jan 97 10:10:13 EST
Two things to note:
1. This stuff has been studied before in other contexts; there's
Stoye's 1-bit hardware reference counts, for instance, and the SETL
project produced a number of papers on analysis and optimization
of copying and reference-counting. If I had the time, I'd yank the
relevant stuff out of my bibliography, but I don't have time today.
It would be good to avoid re-inventing the wheel, and it would be
good to measure things. The point about blindly copying anything
statically known to be small is a good one (with the obvious inference
that the relative costs and frequencies of sharing versus copying
could be measured and tuned, on a size and/or type-specific basis,
and that this would probably also vary depending on the machine where
it was done).
2. I'll be changing my address, "chase@naturalbridge.com" will do
for now. I'll see about relocating the GC FAQ page to someplace
else, and then post when I do that. As usual, if anyone has any
comments on content, I'm happy to listen. It's been a pretty busy
couple of months, and I've let it slide, and the next few months
will be just as busy.
David Chase
CenterLine Software (today)
Natural Bridge (tomorrow)