"Guido van Rossum" <[EMAIL PROTECTED]> writes: > Also, > the argument about dirtying cache lines seem to be relevant to SMP > architectures and probably doesn't apply much to Python as long as we > have the GIL (which is not going away).
I guess the primary cost would probably be incurring the cache miss and resulting stall, but it does also pollute the cache a bit to have lots of 128 or 64-byte lines around for the single refcount field contained therein... But this is just armchair talk, some numbers would be nice. Maybe one could measure it by making the refcounting macros real non-inline functions and seeing how they show up in eg oprofile... -- erno _______________________________________________ Python-3000 mailing list Python-3000@python.org http://mail.python.org/mailman/listinfo/python-3000 Unsubscribe: http://mail.python.org/mailman/options/python-3000/archive%40mail-archive.com
