Spin locks

Ulrich Drepper drepper@redhat.com
Fri May 5 08:47:00 GMT 2000


tb@MIT.EDU (Thomas Bushnell, BSG) writes:

> For things like this, my preference is to just key on __OPTIMIZE__.
> We shouldn't burden the users with a gajillion optimization options.

Well, normally this is a good thing but not here I'd say.  gcc
generates really lousy code when not optimizing, especially the dead
code analysis is missing.  Now you suspect a problem with the
spinlocks.  Your only alternative would be to recompile everything
without optimization.  Yuck.  Instead it should be possible to prevent
the inling during debugging (to get the extra checks) and still have
the dead code being removed.

But since it's hurd I let you guys make the decision.  I think this
feature is important enough having used it on Linux (for mutexes)
several times in the past.

-- 
---------------.      drepper at gnu.org  ,-.   1325 Chesapeake Terrace
Ulrich Drepper  \    ,-------------------'   \  Sunnyvale, CA 94089 USA
Red Hat          `--' drepper at redhat.com   `------------------------


More information about the Libc-hacker mailing list