Lock-free reference counting with RCU
Joe Seigh
http://www.cs.rochester.edu/~saurabh/reports/sadeltcrep.pdf
The compare and swap version seems kind of expensive. It
would have helped if I had known about the paper when I
was arguing about using a wait-free version in LKML.
--
Joe Seigh
When you get lemons, you make lemonade.
When you get hardware, you make software.
Joe Seigh
> Here's a paper on the various techniques in the linux kernel.
> http://www.cs.rochester.edu/~saurabh/reports/sadeltcrep.pdf
>
Anyone know where this reference paper can be found?
1. Sarma, D., Vaddagiri, S., Soni, M., Thirumalai, R., McKenney, P., “Reference
Counting of Lock-free Data Structures in Linux Kernel”
Robert James Shelton
> Here's a paper on the various techniques in the linux kernel.
> http://www.cs.rochester.edu/~saurabh/reports/sadeltcrep.pdf
Looks interesting, thanks for the post.
Rob.
--
Department of Computer Science and Software Engineering.
The University of Melbourne, Australia.
Chris Thomasson
news:l4WdnbVkYL2...@comcast.com...
> Here's a paper on the various techniques in the linux kernel.
> http://www.cs.rochester.edu/~saurabh/reports/sadeltcrep.pdf
>
> The compare and swap version seems kind of expensive. It
> would have helped if I had known about the paper when I
> was arguing about using a wait-free version in LKML.
Yes. The CAS is way expensive, imho. I use groups of per-thread reference
counters to get the job done. No atomic ops/membars... I think SUN "might
of" liked this because I received an e-mail stating that my library was
selected to be one of the four potential winners of the first round of their
CoolThreads contest. It had to be my proxy collector or my memory allocator
that caught one of the judges eyes... Anyway, if all goes well I will soon
be developing lock-free algorithms on a SUNFire T2000.
BTW, did you enter decide to enter that CoolThreads contest? It would be
just my luck to find out that you were one of the four first round winners!
Whoa, what a competition that would be! I don't think I would be able to get
any sleep...
:)
Joe Seigh
was announced so I didn't have a Solaris box to develop on.
You can make money doing lock-free? I didn't think that was possible. :)
Chris Thomasson
> Chris Thomasson wrote:
>> "Joe Seigh" <jsei...@xemaps.com> wrote in message
>> news:l4WdnbVkYL2...@comcast.com...
>>
>>>Here's a paper on the various techniques in the linux kernel.
>>>http://www.cs.rochester.edu/~saurabh/reports/sadeltcrep.pdf
>>>
>>>The compare and swap version seems kind of expensive. It
>>>would have helped if I had known about the paper when I
>>>was arguing about using a wait-free version in LKML.
>>
>>
>> Yes. The CAS is way expensive, imho. I use groups of per-thread reference
>> counters to get the job done. No atomic ops/membars... I think SUN "might
>> of" liked this because I received an e-mail stating that my library was
>> selected to be one of the four potential winners of the first round of
>> their CoolThreads contest. It had to be my proxy collector or my memory
>> allocator that caught one of the judges eyes... Anyway, if all goes well
>> I will soon be developing lock-free algorithms on a SUNFire T2000.
>>
>> BTW, did you enter decide to enter that CoolThreads contest? It would be
>> just my luck to find out that you were one of the four first round
>> winners! Whoa, what a competition that would be! I don't think I would be
>> able to get any sleep...
>>
> You're in luck.
Holy Crap! My luck is usually so go% dam%n bad... My nerves are wound so
tight that I am not too sure I could of survived a competition with you!
FWIW, I have really enjoyed and seriously appreciated "all of the help" you
have provided me with over the years; oh yeah, you haven't flamed me at all!
Thank you for your kindness Joe. Humm, I wonder what would the world be like
without just plain good people such as yourself, most of the people that
frequent this group, and, "hopefully" me?
[...]
> You can make money doing lock-free? I didn't think that was possible. :)
Wow. ;)
I am currently waiting for my signed affidavit to be confirmed. Humm, I am
still a little bit perplexed by the "language" used in the notification I
received from SUN; It clearly states "potential winner". I assume that I am
potentially on of the four finalists, until, my affidavit gets "completely
confirmed/verified". Anyway, again, if all goes well, I am really looking
forward to "super-sizing" my vzoom library up to UltraSPARC T1 standards...
So, I haven't won/received any prize yet, as for making money, I am crossing
my fingers so tight that they are turning black-and-blue!!!
Yikes!!!
Chris Thomasson
news:GtmdnbSPK8gSpz3Z...@comcast.com...
> "Joe Seigh" <jsei...@xemaps.com> wrote in message
> news:9dednTeK-NfUogbZ...@comcast.com...
>> Chris Thomasson wrote:
Anyway, again, if all goes well, I am really looking
> forward to "super-sizing" my vzoom library up to UltraSPARC T1
> standards...
The server just arrived at my home about an hour ago. So, I made through to
the second round... Now, things are dead serious!
:O