Boehm-gc with -00

Fri May 31 10:51:00 GMT 2002

More on incremantal GC vs. thread-local allocation:
I think there's a way to change this such that GC_local_gcj_malloc just calls GC_gcj_malloc in the incremental case. This adds significant allocation overhead to the incremental case, but very little to the normal case. This requires that incremental collection be turned on immediately when starting a gcj process, e.g. in response to an environment variable. I'll put that in GC version 6.1, since it makes things a bit more robust.
It's not clear that this is a complete solution. It does have the advantage that pointerfree objects continue to use real thread-local allocation. (The same is somwhat true, and could be made more true, in the "all pointers" case. My intuition is that only matters for non-primitve arrays, where I think it's already mostly true.)
Hans
> -----Original Message-----
> From: Boehm, Hans [mailto:hans_boehm@hp.com]
> - You will need to either turn off thread local allocation 
> somehow (probably at build time, yucch), or run the collector 
> in fully conservative mode (possibly at run-time), for this 
> to work. Otherwise there are scenarios in which the marker 
> sees a bad mark descriptor in an object being allocated.
>