Subj : ANN: AppCore updates and performance enhancements added...
To   : comp.programming.threads
From : Chris Thomasson
Date : Mon May 02 2005 07:43 pm

I have restructured some of the code and added some performance enhancements 
to AppCore.




Here are some of the changes:


- Made the doubly-linked list API public and added some memory barrier 
variants of the push/pop functions so that it can support modification under 
lock-free reads.


- Added HyperThread friendly spinlock yield functions.


- Added a simple HyperThread friendly threading API that automatically 
offsets the thread stacks in order to get around the aliasing problem 
inherit in some IA32 processors.


- Aligned the global lock-free node cache anchor on a cache-line boundary. 
Adds marked performance improvement.


- Added minor tweaks to the assembly files. Stuff like changing "movl %0, 
%eax" to "xorl %eax, %eax", ect...


- Added a lock-free stack pop function that does not depend on Hazard 
Pointers.


- Removed a couple of interlocked operations from the lock-free atomic 
reference count API. This provides a marked performance improvement.


- Added versioning. AppCore is now at version: 0.0.1 ( pre-alpha )




http://appcore.home.comcast.net/appcore_0-0-1-i686.zip
( entire library. )


http://appcore.home.comcast.net/ac_src_index.html
( you can browse the projects file index )




Any questions and/or comments?


-- 
http://appcore.home.comcast.net/
(portable lock-free data-structures)

.