Re: Segment Register thread descriptor (was Re: Jv_AllocBytesChec ked)

[Benjamin Kosnik <bkoz at redhat dot com>]

> Hans> (See stl_threads.h in libstdc++-v3. It actually looks like there
> Hans> is a second implementation in libstc++-v3 in atomicity.h.  What
> Hans> a mess.)

it would be nice to unify the thread bits for c++. That's supposed to
happen. Note that stl_threads and the atomicity bits are different ways of
implementing this: of the two, I think the atomicity.h implementation is
cleaner. Anyway. That's a discussion that's considerably off topic for
this list.

> If we can't get it into glibc then maybe we can at least share these
> implementations.  We can just start a mini-library (maybe just a
> header file or something) that is incorporated where needed.

Note we had to abstract out this interface on a per-os as well as per-cpu
way. Notably, AIX and Irix require this. Others too.

One of these days the java and libstdc++ people will actually get together
and resolve this thread stuff.

> It looks like at least some of the atomicity.h files come from glibc.
> Is this ok to do?  It seems like this would make part of libstdc++
> LGPL.

I'll look into this. The original work came from glibc, then it was
changed around and some of the ports (arm) went back into glibc. 

Richard Henderson wants to put some of this stuff into the compiler as
builtins. I believe the IA64 port already does this.

benjamin