[Gc] GC_malloc_atomic overallocation bug

Boehm, Hans hans.boehm at hp.com
Wed Oct 24 17:52:12 PDT 2007


It was pointed out to me that GC 7.0 thread local allocation has a
possibly serious performance bug, in that it passed bytes instead of
granules to the inline allocation macro that does the work.
GC_gcj_malloc had the same problem.  I checked in a fix:

http://bdwgc.cvs.sourceforge.net/bdwgc/bdwgc/thread_local_alloc.c?r1=1.1
2&r2=1.13

I don't think gc6.8 is affected.  This is new code.

Hans




More information about the Gc mailing list