Thanks for your info. I've read the current implementation of hash table. Now I konw the point is to be scalability on multicore systems, and make it non-blocking and use progressive synchronization mechanisms such as RCU (Read-Copy-Update).
Nexr step I'll look at the paper you mentioned, that's interesting:-) Thanks for your info again:-) 2012/3/20 Martin Decky <[email protected]> > Hello, > > > On 20-03-2012 04:51, yang kidz wrote: > >> As hash table implementation is network related, I'm also familiar with >> the hash table >> programming and have been using hash and know some basic about it. >> > > Thanks for your interest. Just a small note: HelenOS has a microkernel > design, thus the hash table in our kernel is not actually being used for > networking (the kernel does not know anything about networking, it does not > know anything about file systems and other user space stuff for that > matter). > > On the other hand, the hash table has many uses in our kernel (e.g. in the > IPC) and a very similar implementation of a hash table is being used in > user space, too. There it might be indeed used for networking. > > > I have great interest in the kernel and I would like to improve the hash >> table implementation. >> And I'm going to search more info for this idea,and >> anaylise the linux source code of the related functions. >> > > OK, learning from Linux certainly cannot hurt. Just remember that the goal > here is not to port the Linux implementation to HelenOS. > > I should even warn you not to directly copy any code from Linux to > HelenOS: If one wants to integrate a piece of GPL code into a BSD codebase, > one needs to contact the original author of the GPL code and ask him/her > for the permission to re-license the particular piece of code. > > > Speaking about the hash table: The basic outline and requirements of the > topic can be found in the ticket [1]. I would add that one of the key > improvements should be scalability on multicore systems, which brings us to > non-blocking algorithms and progressive synchronization mechanisms such as > RCU (Read-Copy-Update). > > I have supervised an excellent master thesis of my colleague Andrej > Podzimek [2] (in English) which describes a novel implementation of RCU in > the Solaris kernel. It also deals with the implementation of a non-blocking > hash table. I suggest this text as a great source of inspiration. > > [1] http://trac.helenos.org/**ticket/398<http://trac.helenos.org/ticket/398> > [2] > https://andrej.podzimek.org/**thesis.pdf<https://andrej.podzimek.org/thesis.pdf> > > > M.D. > > ______________________________**_________________ > HelenOS-devel mailing list > [email protected] > http://lists.modry.cz/cgi-bin/**listinfo/helenos-devel<http://lists.modry.cz/cgi-bin/listinfo/helenos-devel> >
_______________________________________________ HelenOS-devel mailing list [email protected] http://lists.modry.cz/cgi-bin/listinfo/helenos-devel
