[gclist] death statistics

James McCartney james@clyde.as.utexas.edu
Thu, 23 Jan 1997 11:10:40 -0700


Are there papers that analyze statistics of the means by which objects
die in an OOPL?

For example:

A. it was an intermediate result on the stack and is consumed by a function.

B. its last reference is overwritten by another pointer.

C. its last reference was directly or indirectly referred to by an object
that died by means A or B

In the GC algorithm I am considering, A and C are cheaper, and B is more
costly.
My intuition is that B is the least frequent means of death.

However that might not be the case in a language with heap allocated
function context frames and there are loops where immutable boxed values
such as double floats are updated each iteration. The updates would produce
a lot of garbage via overwritten pointers in the context frame.

   --- james mccartney     james@clyde.as.utexas.edu   james@lcsaudio.com
If you have a PowerMac check out SuperCollider, a real time synth program:
ftp://mirror.apple.com//mirrors/Info-Mac.Archive/gst/snd/super-collider-demo.hqx