Re: maybe-finish-mapping in map-allocated-objects must be inline
Raymond Toy <[email protected]>
| Newsgroups | gmane.lisp.cmucl.devel |
|---|---|
| Message-ID | <[email protected]> |
On 8/18/11 1:26 AM, Helmut Eller wrote: > The local function MAYBE-FINISH-MAPPING in MAP-ALLOCATED-OBJECTS must be > inlined otherwise the CURRENT pointer is boxed and moved into a value > cell for every object. That boxing allocates something like 1MB of saps > when (room t) is called with an almost empty heap and much more for if > there is something on the heap. Thanks for the patch. I had noticed this some time ago but never got around to investigating why this was happening. Such a simple fix too! > > There's also a unused variable TESTED. It looks like some parts were > written by a drunk :-) Could be me. I did port some of this from sbcl long ago. Ray _______________________________________________ cmucl-imp mailing list [email protected] http://lists.zs64.net/mailman/listinfo/cmucl-imp