------- Comment #24 from pcarlini at suse dot de 2006-05-22 14:44 ------- Humm, maybe I (we ;) was wrong: I'm doing some benchmarks and it looks like the cost of checking __gthread_active_p() compared to that of an atomic operation is very small... More later.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24704
