Robin Green writes: > On Mon, Aug 15, 2005 at 11:25:53AM +0100, Andrew Haley wrote: > > > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23367 > > > > > > I'm told that this bug only affects the interpreter. > > > > Can you please tell us how this should be changed? Not necessarily a > > patch, but an outline of one. > > The function in question needs to be synchronized in some way, assuming > you're going to keep the cache. However, Tom Tromey speculated that removing > the cache might be more efficient than synchronizing on every cache lookup. Yes, it might indeed be. We should perhaps just move the cache to thread-local storage, and if the target has no TLS, we can disable the feature. We already have a configure check for TLS in gcc, so it might be pretty easy. Andrew.
