I looked at the squeak identityHash in squeak a few years back (a) how efficient was the algorithm? Attempted at one time to replace with a fixed array of numbers. There was some benefit on slow machines. (b) only calculate the identity hash when required, but then how do you mark the oops with that fact when all the identityHash values are in use in the image. The hassle of fixing an image to use the new hash logic wasn't worth the gain which was difficult to measure. In both cases I was more interested in the cpu usage on those 68K machines. On Wed, Oct 28, 2009 at 7:14 PM, Andres Valloud < avalloud@smalltalk.comcastbiz.net> wrote:
Martin,
One of the constituencies I thought of when I decided to leave identityHash alone was folks like you.
-- =========================================================================== John M. McIntosh <johnmci@smalltalkconsulting.com> Corporate Smalltalk Consulting Ltd. http://www.smalltalkconsulting.com ===========================================================================