acf
Subj : Re: Hazard Pointers w/o memory barrier.
To   : comp.programming.threads
From : Joe Seigh
Date : Tue May 03 2005 08:34 am

On Mon, 2 May 2005 19:07:37 -0700, Chris Thomasson <_no_damn_spam_cristom@_no_damn_comcast.net_spam> wrote:

>> About a 3x
>> slowdown in write access but that's because I don't
>> have the SMR logic in yet plus I have to stay with a
>> slightly crippled version of RCU to avoid imperial
>> entanglement.
>
> We aren't the programmers there looking for...  ;)
>
> Seriously though, this is why I don't provide the public with my user-space
> rcu ( URCU ) implementation. Did you inform McKenney that you were posting
> your algorithm? I am not sure if I need to ask permission...
>

Actually, thinking about it, I'd stay away from RCU until IBM makes its
intentions a little more clear.  IBM does have this Linux strategy.  It's
not a problem for me since I'm not going to publish any code, I'm just
experimenting with it.  If you've been following any of the C++ smart
pointer / shared pointer discussions, you'd realize that nothing short
of a scheme that can GC everything with zero overhead in a multi-threaded
environment with no scalability problems and can deal with untracked raw
pointers will satisfy anyone.  Until everyone becomes more realistic, and
I don't know how long that will take since most of them don't appear to
have even started on their multi-threaded learning curves, there's not
much use trying to create a general usage smart pointer.

Back to IBM again.  They don't seem to publicize any projects that are in
the works, so it's hard to tell what they might be up to here.  I would
guess they might try to put this in the Linux kernel since RCU is already
in place and the hazard pointer stuff would solve some problems running
with preemption enabled and some problems with resource granularity that
a pure proxy GC method would have.  As far as user space is concerned,
they might try to restrict it's usage to Linux as far as it fits in with
their Linux strategy.  That could be a problem however since a lot of
open source projects are platform independent and aren't going to adopt
non-portable api's.

You might run something by IBM's lawyers just to see what they might say.
If you phrase the questions right, you might learn something about IBM's
intentions but being lawyers they'd probably catch on pretty quick.  You'd
have more of a chance if you restricted your stuff to Linux though I don't
know if the GPL will allow patent restrictions of that kind.  I don't know
how they got RCU in the Linux kernel in the first place like they did.


-- 
Joe Seigh

When you get lemons, you make lemonade.
When you get hardware, you make software.

.

0