On Sun, Jul 08, 2001 at 10:18:12AM -0700, dean gaudet wrote: > > > On Sun, 8 Jul 2001, Justin Erenkrantz wrote: > > > Yup. I've brought this up to Sander and David before, but this is how > > pools > > woah. no way really? > > that's not at all how it was in 1.3 or in early 2.0 ... > > in 2.0 as of uh a year ago say, there was one free list per process, > and locks were used to access it. > > no matter where in the tree of pools you tried an allocation, if it > didn't fit into the current block, the allocator would lock and go to > the global free block list and pick up the first fit block. none of this > going up through a chain of pools or anything, that's insane.
No, nothing changed. We just added the chain of pools. =) It's a work in progress. You may call it a "feature" right now. We're not saying it is ready to be activated - just hammered/examined by people on this list. That definitely needs to be cleaned up. We know that. If we switch that global free block list to per-thread (where we are heading), is this allocation algorithm the best one for httpd-style memory usage? If so, then I guess I won't look at other allocation algorithms. I'll defer to people who've studied this a lot longer than I have. -- justin
