Re: Down with weak lists!

On Wed, 14 Feb 2001 10:48:10 +0100, sperber(a)informatik.uni-tuebingen.de (Michael Sperber [Mr. Preprocessor]) said: I am reminded of the story of the early 70's Fortran optimizer - a programmer tightened up special case code and thereby introduced a bug. Upon analysis, it turned out that the special case code was never *ever* called - it was essentially dead code, but the optimization broke code paths that WERE used. I say implement it slowly but bug-free, if at all. We'll put in caching and the like when somebody notices it's slow. If skipping caching here in code that nobody even *USES* as far as we know can result in benefits for the GC code that *every single user* will depend on *heavily*, that's a win. Even if the weak-list code ends up being 10x slower, if the GC can run 10% faster as a result, we've won. First rule of optimization: Optimize the actual hot spots. -- Valdis Kletnieks Operating Systems Analyst Virginia Tech