Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.programming.threads > #2646 > unrolled thread

Scalable RWLocks

Back to article view | Back to comp.programming.threads

Contents

  Scalable RWLocks Ramine <ramine@1.1> - 2014-10-31 14:59 -0700

#2646 — Scalable RWLocks

Hello,


I have just read the following PhD paper about NUMA-Aware Reader-Writer 
Locks...

Here it is:

http://mcg.cs.tau.ac.il/papers/ppopp2013-rwlocks.pdf


You will notice that it is scaling better than the  DV Read-Writer lock 
cause i think there is less NUMA nodes inthis Reader-Writer lock than 
CPU ids in the DV side, so it makes the DV side more expensive in the 
writer side. so it makes the serial part in the Amdahl equation bigger , 
so it makes the DV Reader-Writer lock less scalable... but the DV 
Reader-Writer lock is good also.


But i have also invented and inplemented scalable RWLocks , here they are:

https://sites.google.com/site/aminer68/scalable-rwlock


The scalability of my scalable RWLocks is good at 1% to 3% of writes, 
even more, i have invented scalable and starvation-free RWLocks, you 
will find them inside the same zipfile and this makes my scalable 
RWLocks attractive and powerful i think.

Please notice that the Reader-Writer lock that is included inside
the  Omnithread library is using an expensive atomic operation,
on the reader side , so it makes it less scalable than my scalable 
RWLocks, so i advice you to use my scalable RWLocks cause they are more 
powerful and they are really good.


You can download my scalable RWLocks:

https://sites.google.com/site/aminer68/scalable-rwlock




Thank you,
Amine Moulay Ramdane.

[toc] | [standalone]

Back to top | Article view | comp.programming.threads

csiph-web