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

Groups > comp.programming.threads > #2174

As you have noticed

Cross-posted to 2 groups.

Show all headers | View raw

Hello,

As you have noticed on my previous post called:
"Weakness of lockfree loops" , i have again looked
at the performance of the lockfree loops from the cache=line transfer 
point of view , you see ? This has confirmed what i have told you before...


Here is what i have told you, please reread...

I have come to another subject, you have seen me on
this forums inventing scalable RWLocks and scalable FIFO fair locks,
and speaking about waitfree and lockfree FIFO queues.. and i was
thinking more and saying to myself what is the very important
thing to know when designing such scalable algorithms? i think
the very important thing to know about those algorithms is there speed
and scalability, but when you you want to optimize those algorithms for 
speed you have to concentrate on cache-lines transfers, you have to 
reduce cache-lines transfers between caches and that's the very 
important thing to do to speed your synchronization algorithms or
concurrent FIFO queues algorithms, and the other very important thing to 
know is how to achieve scalability with those algorithms? it's the same 
answer, you have to eliminate cache-lines transfers between caches this 
is how you can achieve scalability, so from now on you have to look at 
those synchronization algorithms and concurrent FIFO queues algorithms 
from the cache-lines transfers point of view to judge or to achieve 
speed and scalability.


Thank you,
Amine Moulay Ramdane.

Back to comp.programming.threads | Previous | Next | Find similar | Unroll thread

Thread

As you have noticed aminer <aminer@toto.net> - 2014-04-14 20:32 -0700

csiph-web