More about the scalability...

[Ramine]

Hello,


I was simulating the scalability of my new Scalable Distributed
Reader-Writer Mutex version 1.34 on an x86 Quadcore, and i have noticed
that it can scale up to 40X on 40 cores with a reader section of ~10 CPU
cycles and with 0.1% to 0.2% of writers, so i think that in average it
can scale up to 40X on 40 cores with 0.1% to 02% of writers, so if you
want to scale much more that that you have to use my scalable SeqlockX.

If you ask me why it can not scale more than 40X..

The answer is on the writer side of my scalable distributed
reader-writer mutex, the serial part of the Amdahl's law of the writer side
of my scalable distributed reader-writer mutex will get much bigger when
you add more and more threads on more and more cores because you have to
transfer cache-lines on the writer side of each RWLock of the Scalable
distributed Reader-Writer Mutex...so you can not avoid this weakness
that limit the scalability to an average of 40X.


You can download the new updated version 1.34 from:

https://sites.google.com/site/aminer68/scalable-distributed-reader-writer-mutex


Thank you for your time.


Amine Moulay Ramdane.