Is there any change on GC algorithm ever cince 19.28?

Dear all, How do you do. My name Is Kenichi Okuyama, and I have just joined this ML few days ago. I'm intrested in xemacs newest version, especially on part of Garbage Collection. And I would like to know if there is any changes made ever since 19.28. Is it re-written ( with basically same algorithm? ), or is it totally new? Long ago, when emacs was 19.28 ( and mule was 2.3 ), I made a little hack on alloc.c . It's still available on http://nagahashi-www.isl.titech.ac.jp/~okuyama/html/program/Mule-2.3-allo... (this is for Mule version of emacs, so it might not work properly on emacs, though I don't beleave so). I wondered if the same hack works on current version of xemacs. What I did was very simple. When you have once used many cons-cells, like when you run GNUS 4.* ( the very old one ), and when you're done with it, what you'll have is pile full of "unused cons-cell". Problem that old emacs-GC had, was that they keep on re-checking the cells that are linked on FREE-LIST. This will: 1)re-load all the swapped out pages that have not been used for long time, just to re-check that they are not being used. 2)by swapping out, they keep the values that have no meaning anymore, which will waste virtual address space. Since cons-cells were malloced per "chunk", I added very little routine that if all the cons-cells in the "chunk" was unused, instead of adding those cons-cells to free-list, simply free() the "chunk". This hack worked good because 1) after first GC is done, chunks are free-ed and will not cause swap-in/swap-out anymore. This was effective when we had little real memory ( like 12Mbytes.. ). 2) since "chunk" are being free-ed, under some malloc system, virtual memory area could be released. I wondered if this hack ( not the same routine, but same idea ) still works or not. I'm asking this because I'm having trouble with accessing to xemacs's newest version, Because 1) I have no idea which one is the newest. ( I didn't even know about xemacs until few month ago ). 2) I can't run CVS* tools, because of firewall. but still, I would like to know if my old idea could be of any help. ,regards ---- Kenichi Okuyama ＠ Tokyo Research Lab. IBM. Co.