Hrvoje Niksic <hniksic(a)iskon.hr> writes:
It seems that our relocating allocator attempts to do what Doug Lea
malloc() already does.
IIRC both our current collectors are older versions of the allocator
that Doug Lea's allocator replaced in libg++/glibc. Using the Doug Lea
version should normally be preferred I think..
The system-provided mmap-based allocation
seems a better idea in XEmacs because it's performed for *all* large
allocations (large vectors, strings, frob blocks etc.) rather than
only for buffers, as our ralloc does.
Shouldn't we be using ralloc for frob blocks anyway?
Does this mean we ought to turn off relalloc on Linux for 21.2?
I don't know. My problem would be the fact that the triggering is
based on allocation size.. What happens if you alloc+free
- 1000, 1K strings
- 1000, 1k buffers
- 1 MB worth of cons cells.
I am not sure either the current system or dlmalloc works in this
case. Along the same lines I have been wondering whether the frobblock
size shouldn't be upped?
Jan